feat(paper): ETH a tick 5m + tooling per-year/per-tick analysis

scripts/run_paper_trading.py: AssetConfig ETH ora usa timeframe="5m" invece del default 1h. Il winner c04dff7086 e' stato trovato dal GA su dati 5m e a 1h la strategia perde: - ETH @ 5m (native): +359.50% cum 7y, 77% winrate, max DD/yr 19% - ETH @ 1h (precedente): -33.03% cum 7y, 67% winrate, max DD 74% BTC resta a 1h (winner 238e4812 native a 1h, +104% 7y, Sharpe 2+ in 3 anni). Nuovi script di analisi: - scripts/yearly_strategies.py: breakdown per anno (2019-2025) di 4 strategie su tick di discovery (trade/winrate/return/maxDD/Sharpe). - scripts/multi_tick_strategies.py: confronto cross-tick (5m/15m/1h) per i 2 winner correnti. Documenta la divergenza tick-paper di ETH. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
feat(validation): WFA tooling + multi-fold results phase1-100 runs
2026-05-16 22:10:38 +00:00 · 2026-05-16 21:48:55 +00:00 · 2026-05-16 21:48:42 +00:00 · 2026-05-16 13:21:37 +00:00 · 2026-05-16 13:21:08 +00:00 · 2026-05-16 11:07:40 +00:00
36 changed files with 5446 additions and 193 deletions
@@ -12,23 +12,23 @@ git clone ssh://git@git.tielogic.xyz:222/Adriano/Multi_Swarm_Coevolutive.git

 ## Layout monorepo (uv workspace)

-Il repo è un **workspace uv** con due member packages indipendenti:
+Il repo è un **workspace uv** con due member packages indipendenti, principio "**core = framework, strategy = contenuto**":

 ```
 multi_swarm_coevolutive/             repo root (workspace coordinator)
 ├── pyproject.toml                    workspace + dev deps + ruff/mypy/pytest
-├── docker-compose.yml                strategy-crypto-paper + strategy-crypto-gui
+├── docker-compose.yml                3 servizi su immagine condivisa
 ├── Dockerfile                        immagine multi-swarm-coevolutive:dev
 ├── uv.lock                           lock unico del workspace
 ├── data/, series/, state/            cache OHLCV + DB (runtime, gitignored)
 ├── scripts/                          CLI entrypoints (run_phase1, run_paper_trading, ...)
 └── src/
    ├── multi_swarm_core/             WORKSPACE MEMBER (wheel: multi-swarm-core)
-    │   ├── pyproject.toml             deps: pandas, numpy, openai, pydantic, ...
+    │   ├── pyproject.toml             core deps (pandas, numpy, openai, pydantic, nicegui, ...)
    │   ├── multi_swarm_core/          GA + genome + protocol + backtest + cerbero +
    │   │                              data + llm + agents + ga + orchestrator +
-    │   │                              metrics + persistence + config
-    │   ├── tests/                     unit + integration (182 test)
+    │   │                              metrics + persistence + config + dashboard (GA-only)
+    │   ├── tests/                     unit + integration
    │   └── docs/                      design/ + decisions/ + reports/
    │
    └── strategy_crypto/              WORKSPACE MEMBER (wheel: strategy-crypto)
@@ -36,27 +36,73 @@ multi_swarm_coevolutive/             repo root (workspace coordinator)
        ├── README.md                  overview strategia + pattern per nuove strategie
        ├── strategy_crypto/
        │   ├── backend/               paper-trading (executor, portfolio, persistence, schema)
-        │   ├── frontend/               NiceGUI dashboard dual-DB
-        │   └── strategies/             JSON freezate (btc_*.json, eth_*.json)
-        └── tests/                      smoke regression (import + json + schema)
+        │   ├── frontend/              NiceGUI paper-only dashboard
+        │   ├── strategies/            JSON freezate (btc_*.json, eth_*.json)
+        │   └── prompts.json           v3.2: agent_role/pattern_guidance/instruction/
+        │                              domain_warnings/anti_patterns/output_priorities +
+        │                              7 stili cognitive (directive + focus_metrics)
+        └── tests/                     smoke regression
 ```

 **DB separati per dominio:** `state/runs.db` (GA core universale) + `state/strategy_crypto.db` (paper della strategia crypto). Pattern scala a N strategie senza naming collision.

-**Pattern N strategie future:** aggiungere `src/strategy_<asset>/` con lo stesso scheletro (`backend/`, `frontend/`, `strategies/`, `tests/`), DB dedicato `state/strategy_<asset>.db`, servizi Docker `strategy-<asset>-paper` + `strategy-<asset>-gui`, GUI su `/strategy_<asset>_gui`.
+**Pattern N strategie future:** aggiungere `src/strategy_<asset>/` con stesso scheletro (`backend/`, `frontend/`, `strategies/`, `tests/`, `prompts.json`), DB dedicato `state/strategy_<asset>.db`, servizi Docker `strategy-<asset>-paper` + `strategy-<asset>-gui`, GUI su `/strategy_<asset>_gui`. **Zero modifiche al core** richieste.
+
+## Architettura prompt (v3.2)
+
+**Compositor**: il SYSTEM prompt al LLM viene COMPOSTO at-runtime da scaffold core + contenuto strategy:
+
+```
+[1] agent_role           ← strategy (prompts.json — chi è l'agente)
+[2] cognitive_style + directive  ← genome (evoluti dal GA)
+[3] GRAMMAR_SPEC         ← core scaffold (operatori, indicatori, units rules)
+[4] pattern_guidance     ← strategy (forme di curva astratte, no indicatori prescritti)
+[5] domain_warnings      ← strategy (es. "crypto trada 24/7, NON inferire funding rate")
+[6] CONSTRAINTS          ← core scaffold (validator semantics)
+[7] anti_patterns        ← strategy (7 voci: no >4 AND, no chattering, isteresi, ecc.)
+[8] output_priorities    ← strategy (5 voci, #1 coerenza con lente cognitiva)
+[9] EXAMPLE              ← core scaffold
+```
+
+**Input USER (calcolato da `build_market_summary`):**
+- Base (5): mean, std, skew, kurt, vol_regime
+- Regime recente rolling 500 (6): autocorr_lag1 (recent + baseline), hurst, vol_percentile, seasonality (hour + dow)
+- Geometria & frattali (7): efficiency_ratio (Kaufman), tail_index (left + right Hill), structural_uptrend (HH/HL), compression, spectral_entropy, dominant_cycle (gated)
+- Feature accessibili dal genome + lookback_window
+- **Focus per la tua lente**: blocco style-aware (4 metriche prioritarie da `focus_metrics` di prompts.json)
+- Instruction finale (da strategy)
+
+**Grammar protocol JSON (8 indicatori):**
+| Indicatore | Output | Range |
+|------------|--------|-------|
+| `sma(length)` | media mobile | unità prezzo |
+| `sma_pct(length)` | (close-sma)/sma | ±0.1 frazione |
+| `rsi(length)` | oscillator | 0-100 |
+| `atr(length)` | true range | unità prezzo |
+| `atr_pct(length)` | atr/close | 0-0.1 frazione |
+| `realized_vol(window)` | std returns | 0-0.1 frazione |
+| `macd(fast,slow,signal)` | momentum | unità prezzo |
+| `macd_pct(...)` | macd/close | ±0.02 frazione |
+
+**7 stili cognitive** (in `prompts.json`, editable): physicist, biologist, historian, meteorologist, engineer, military_strategist, psychologist. Ognuno con directive 800-950 char, ASCII-strict, archetipo dominante + lookback consigliato + 4 focus_metrics.

 ## Stato del progetto

-**Phase 3 (paper-trading forward-test) in corso** dal 13 maggio 2026 su VPS. Runner `scripts/run_paper_trading.py` long-running in Docker, dashboard NiceGUI su `https://swarm.tielogic.xyz/strategy_crypto_gui/`. Due strategie freezate:
+**Phase 3 (paper-trading forward-test) in corso** dal 13 maggio 2026 su VPS. Runner `scripts/run_paper_trading.py` long-running in Docker, dashboard NiceGUI su `https://swarm.tielogic.xyz/strategy_crypto_gui/`. Due strategie freezate in `src/strategy_crypto/strategy_crypto/strategies/`:

- `strategy_crypto/strategies/btc_fb63e851.json` — BTC-PERPETUAL, RSI estremi + ATR vs SMA + filtro orario 9-17 (Sharpe OOS +0,26 su 7,33 anni).
- `strategy_crypto/strategies/eth_facd6af85d5d.json` — ETH-PERPETUAL, regime-based (Sharpe OOS +0,19 su 6,75 anni).
+- `btc_fb63e851.json` — BTC-PERPETUAL, RSI estremi + ATR vs SMA + filtro orario 9-17 (Sharpe OOS +0,26 su 7,33 anni).
+- `eth_facd6af85d5d.json` — ETH-PERPETUAL, regime-based (con `atr_pct` post-fix bug unità).

-Phase 1 → 2.7 chiuse (30 run GA, $3.74 cumulato LLM).
+Phase 1 → 2.7 chiuse (30 run GA, $3.74 cumulato LLM). Sessione refactor 15 maggio 2026:
+- Split repo invertito, monorepo unificato come uv workspace
+- Family `*_pct` completa (atr_pct, sma_pct, macd_pct) per fix bug unità
+- Dashboard split: core (GA) vs strategy (paper)
+- Prompt architecture compositor + prompts.json v3.2 (vedi decision log)

- [**Stato progetto e roadmap (14 maggio 2026)**](src/multi_swarm_core/docs/reports/2026-05-14-stato-progetto-e-roadmap.md) — riepilogo fasi, decisioni, caveat, roadmap.
- Decision log: [`src/multi_swarm_core/docs/decisions/`](src/multi_swarm_core/docs/decisions/) (gate Phase 1, scelta nemotron).
- Design docs concettuali: [`src/multi_swarm_core/docs/design/`](src/multi_swarm_core/docs/design/).
+Documenti:
+- [**Stato progetto e roadmap (14 maggio 2026)**](src/multi_swarm_core/docs/reports/2026-05-14-stato-progetto-e-roadmap.md)
+- Decision log: [`src/multi_swarm_core/docs/decisions/`](src/multi_swarm_core/docs/decisions/) (gate Phase 1, nemotron, atr_pct fix)
+- Design docs concettuali: [`src/multi_swarm_core/docs/design/`](src/multi_swarm_core/docs/design/)

 Stack: Python 3.13, uv workspace, hatchling, pytest+pytest-mock+responses, openai SDK (verso OpenRouter), requests+tenacity, pandas+numpy+scipy, sqlmodel+sqlite, nicegui+plotly, yfinance.

@@ -65,7 +111,7 @@ Stack: Python 3.13, uv workspace, hatchling, pytest+pytest-mock+responses, opena
 ```bash
 uv sync                           # installa entrambi i workspace member come editable
 cp .env.example .env              # compila CERBERO_*_TOKEN e OPENROUTER_API_KEY
-uv run pytest                     # 186 test attesi (182 core + 4 smoke strategy_crypto)
+uv run pytest                     # 252 test attesi (248 core + 4 smoke strategy_crypto)
 ```

 ### Variabili .env richieste
@@ -84,10 +130,9 @@ OPENROUTER_API_KEY=<sk-or-v1-...>
 GA_DB_PATH=./state/runs.db
 STRATEGY_CRYPTO_DB_PATH=./state/strategy_crypto.db

-# Deploy Docker
+# Deploy Docker — DASHBOARD_ROOT_PATH ora per-servizio (vedi docker-compose.yml)
 DOMAIN_NAME=tielogic.xyz
 SWARM_DASHBOARD_PORT=8080
-DASHBOARD_ROOT_PATH=/strategy_crypto_gui   # subpath traefik per la dashboard
 ```

 Backcompat: `DB_PATH` legacy continua a funzionare come alias di `GA_DB_PATH`.
@@ -105,14 +150,27 @@ uv run mypy src/ scripts/
 # Smoke run (MockLLM + OHLCV sintetico, no API calls)
 uv run python scripts/smoke_run.py

-# Run reale Phase 1/2 (Cerbero + OpenRouter, ~$0.07 per run K=20 10gen)
+# Run reale Phase 1/2 (Cerbero + OpenRouter, ~$0.15-0.25 per run K=20 10gen,
+# ~$0.40-0.55 per run esteso K=40 20gen con WFA OOS).
+# Default --start ora 2018-09-01 (7.3y, copre bear+halving+covid+ATH+winter+ETF).
 uv run python scripts/run_phase1.py \
  --name run-XXX \
  --exchange deribit --symbol BTC-PERPETUAL --timeframe 1h \
-  --start 2024-01-01T00:00:00+00:00 \
-  --end 2026-01-01T00:00:00+00:00 \
  --population-size 20 --n-generations 10 \
-  --prompt-mutation-weight 0.30 --fitness-v2
+  --prompt-mutation-weight 0.30 --fitness-v2 \
+  --llm-concurrency 8                          # 5-8x speedup wall time (default 1)
+# fitness-v2 hardened: hard-kill su {no_trades, degenerate, undertrading,
+# fees_eat_alpha, negative_net_pnl}. Override via --fitness-hard-kill CSV.
+# Default --prompt-library: importlib.resources del package strategy_crypto/prompts.json
+
+# Multi-fold validation di un run esistente (anti-overfit, 7y expanding-window)
+uv run python scripts/validate_run.py \
+  --run-id <run_id> \
+  --top-k 10 --n-folds 4 --train-ratio 0.5 \
+  --start 2018-09-01T00:00:00+00:00 --end 2026-01-01T00:00:00+00:00 \
+  --fitness-v2 \
+  --output-json state/validation-XXX.json
+# Ranking per "robust_score" = min(fitness_oos) su tutti i fold.

 # Backtest standalone di una strategia JSON
 uv run python scripts/backtest_strategy.py \
@@ -123,32 +181,50 @@ uv run python scripts/backtest_strategy.py \
 uv run python scripts/run_paper_trading.py \
  --name phase3-papertrade-XXX \
  --initial-capital 1000 --poll-seconds 300
-# Default --strategies-dir: importlib.resources del package strategy_crypto

-# Dashboard NiceGUI locale
-uv run python -m strategy_crypto.frontend.nicegui_app
-# → http://localhost:8080 (env SWARM_DASHBOARD_PORT)
+# Dashboard NiceGUI locale (2 distinte)
+uv run python -m multi_swarm_core.dashboard.nicegui_app   # GA core (/, /convergence, /genomes)
+uv run python -m strategy_crypto.frontend.nicegui_app     # Strategy crypto (/ paper)
 ```

-## Dashboard
+## Performance & Validation

-NiceGUI dashboard (dark palette) — **dual-DB reader** (GA + paper):
+**Backtest engine vettorializzato** (`backtest/engine.py`): rimosso il loop `pd.iterrows()` a favore di state machine numpy. Speedup misurati:

+| Dataset | Before (iterrows) | After (vectorized) | Speedup |
+|---------|-------------------|--------------------|---------|
+| 2 anni (17545 bar) | 470 ms | **28 ms** | **16.8×** |
+| 7 anni (64297 bar) | 1744 ms | **154 ms** | **11.3×** |
+
+Equivalenza numerica garantita: 5 parity test parametrici vs. reference implementation legacy (`test_backtest_engine_vectorized.py`).
+
+**Parallel propose LLM** (`orchestrator/run.py`): `--llm-concurrency N` lancia N chiamate `hypothesis_agent.propose()` concorrenti per generazione tramite `ThreadPoolExecutor`. OpenRouter qwen-2.5 regge 6-10 concorrenti senza rate-limit. Default 1 = backward-compat.
+
+**Multi-fold validation tool** (`scripts/validate_run.py`): qualunque run completato puo' essere rivalutato post-hoc su N fold expanding-window di un dataset esteso (tipicamente 7 anni). Vital per evitare il single-hold-out overfit: il GA puo' selezionare un genome con `fitness_is` alta che collassa OOS (osservato su `phase1-extended-001`: elite IS Sharpe 1.93, OOS Sharpe -1.00). Ranking finale per `robust_score = min(fitness_oos)`. Output JSON con per-fold breakdown + aggregati mean/min/std.
+
+## Dashboard (split core + strategy)
+
+Due NiceGUI dashboard distinte (dark palette, palette neon):
+
+**Core GA** — `multi_swarm_core.dashboard.nicegui_app` → `https://swarm.tielogic.xyz/multi_swarm_core_gui/`:
 - **Overview** (`/`): lista runs GA, costo cumulato, metriche aggregate evaluations.
 - **GA Convergence** (`/convergence`): fitness median/max/p90 per generazione + entropy.
 - **Genomes** (`/genomes`): top-K ordinati per fitness, ispezione system_prompt + JSON strategy.
- **Paper** (`/paper`): forward-test live con equity curve, posizioni aperte, trade list, tick log.

-In produzione su `https://swarm.tielogic.xyz/strategy_crypto_gui/` (subpath gestito via `DASHBOARD_ROOT_PATH` + Traefik PathPrefix). La root del dominio resta libera per future GUI di altre strategie.
+**Strategy crypto** — `strategy_crypto.frontend.nicegui_app` → `https://swarm.tielogic.xyz/strategy_crypto_gui/`:
+- **Paper** (`/`): forward-test live con equity curve, posizioni aperte, trade list, tick log.
+
+In produzione subpath gestiti via `DASHBOARD_ROOT_PATH` (per-servizio) + Traefik `replacepathregex` (NB: NON `stripprefix`, vedi sezione Deploy). La root del dominio resta libera per future GUI di altre strategie.

 ## Deploy

-`docker-compose.yml` definisce due servizi su immagine `multi-swarm-coevolutive:dev`:
+`docker-compose.yml` definisce 3 servizi su immagine condivisa `multi-swarm-coevolutive:dev`:

 - **`strategy-crypto-paper`** — runner `scripts/run_paper_trading.py` long-running.
- **`strategy-crypto-gui`** — NiceGUI dashboard dietro Traefik su `https://swarm.${DOMAIN_NAME}/strategy_crypto_gui/`.
+- **`strategy-crypto-gui`** — NiceGUI paper dashboard su `https://swarm.${DOMAIN_NAME}/strategy_crypto_gui/`.
+- **`multi-swarm-core-gui`** — NiceGUI GA dashboard su `https://swarm.${DOMAIN_NAME}/multi_swarm_core_gui/`.

-Persistenza via bind mount: `./data/`, `./series/`, `./state/`. Le strategie JSON sono bind-mounted in read-only dal package: `./src/strategy_crypto/strategy_crypto/strategies/`.
+Persistenza via bind mount: `./data/`, `./series/`, `./state/`. Strategie JSON bind-mounted in read-only dal package: `./src/strategy_crypto/strategy_crypto/strategies/`.

 ```bash
 docker compose up -d --build
@@ -158,12 +234,16 @@ docker compose ps

 Note operative:

- Le bind-mount dir devono essere `chown 1000:1000` (uid utente `app` nel container).
+- Bind-mount dir devono essere `chown 1000:1000` (uid utente `app` nel container). **Anche `src/strategy_crypto/strategy_crypto/strategies/`** (creata da `git mv`, default `root:root`).
 - Override del command paper via env (`PAPER_RUN_NAME`, `PAPER_INITIAL_CAPITAL`, ecc.).
 - `SWARM_DASHBOARD_PORT` controlla la porta interna del container (Traefik fa il TLS).
+- **Traefik subpath**: usa `replacepathregex` middleware (NON `stripprefix`) per evitare doppio root_path (uvicorn legge `X-Forwarded-Prefix` da stripprefix + applica `root_path` di NiceGUI = doppio prefix). Vedi commit `436613b`.
+- Dopo cambio label Traefik: `docker restart traefik-traefik-1` per forzare refresh discovery.

 ## Sviluppo

 Conventional commits con prefix `feat:` `fix:` `chore:` `docs:` `refactor:` `test:`. Body italiano. Footer `Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>` su ogni commit collaborativo.

 Branch attuale: `main`. Workspace single-repo, monorepo unificato dal 15 maggio 2026 (split temporaneo monorepo→figlio invertito, vedi tag `v0.1.0-pre-split` come bookmark).
+
+**Modificare il prompt LLM** senza toccare codice: edita `src/strategy_crypto/strategy_crypto/prompts.json`. Schema documentato in `_design_invariants` del JSON stesso. I 3 regression guard test (`test_strategy_crypto_directives_ascii_safe`, `..._have_archetype_marker`, `..._have_lookback_hint`) bloccano regressioni accidentali sulle invarianti di design.
@@ -0,0 +1,125 @@
+"""Analisi per-trade dei top-K candidate del run BTC.
+
+Per ciascun genome top-K, ri-esegue il backtest su ogni fold WFA e raccoglie:
+- n_trades, n_wins, n_losses, win_rate
+- max_drawdown
+- return, sharpe
+- list trade pnl summary
+
+Output stampato a stdout, non scrive su disco.
+"""
+from __future__ import annotations
+
+import argparse
+from datetime import datetime
+
+import pandas as pd  # type: ignore[import-untyped]
+
+from multi_swarm_core.agents.hypothesis import _try_parse
+from multi_swarm_core.backtest.engine import BacktestEngine
+from multi_swarm_core.cerbero.client import CerberoClient
+from multi_swarm_core.config import load_settings
+from multi_swarm_core.data.cerbero_ohlcv import CerberoOHLCVLoader, OHLCVRequest
+from multi_swarm_core.data.splits import expanding_walk_forward
+from multi_swarm_core.metrics.basic import max_drawdown, sharpe_ratio, total_return
+from multi_swarm_core.persistence.repository import Repository
+from multi_swarm_core.protocol.compiler import compile_strategy
+
+
+def main() -> None:
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--run-id", required=True)
+    ap.add_argument("--top-k", type=int, default=10)
+    ap.add_argument("--n-folds", type=int, default=4)
+    ap.add_argument("--train-ratio", type=float, default=0.5)
+    ap.add_argument("--symbol", default="BTC-PERPETUAL")
+    ap.add_argument("--timeframe", default="1h")
+    ap.add_argument("--start", default="2018-09-01T00:00:00+00:00")
+    ap.add_argument("--end", default="2026-01-01T00:00:00+00:00")
+    ap.add_argument("--fees-bp", type=float, default=5.0)
+    args = ap.parse_args()
+
+    settings = load_settings()
+    repo = Repository(settings.ga_db_path)
+    repo.init_schema()
+    all_evals = repo.list_evaluations(args.run_id)
+    parseable = [
+        e for e in all_evals
+        if e.get("raw_text") and not e.get("parse_error") and e["fitness"] > 0
+    ]
+    parseable.sort(key=lambda e: e["fitness"], reverse=True)
+    seen: set[str] = set()
+    top: list[dict] = []
+    for e in parseable:
+        if e["genome_id"] in seen:
+            continue
+        seen.add(e["genome_id"])
+        top.append(e)
+        if len(top) >= args.top_k:
+            break
+
+    token = (
+        settings.cerbero_mainnet_token.get_secret_value()
+        if settings.cerbero_mainnet_token
+        else settings.cerbero_testnet_token.get_secret_value()
+    )
+    cerbero = CerberoClient(
+        base_url=settings.cerbero_base_url,
+        token=token,
+        bot_tag=settings.cerbero_bot_tag,
+    )
+    loader = CerberoOHLCVLoader(client=cerbero, cache_dir=settings.series_dir)
+    ohlcv = loader.load(OHLCVRequest(
+        symbol=args.symbol,
+        timeframe=args.timeframe,
+        start=datetime.fromisoformat(args.start),
+        end=datetime.fromisoformat(args.end),
+    ))
+    splits = expanding_walk_forward(ohlcv.index, train_ratio=args.train_ratio, n_folds=args.n_folds)
+    engine = BacktestEngine(fees_bp=args.fees_bp)
+
+    print(f"\n{'=' * 110}")
+    print(f"PER-TRADE ANALYSIS — top-{len(top)} genomes × {len(splits)} folds")
+    print(f"{'=' * 110}")
+
+    for ev in top:
+        strat, err = _try_parse(ev["raw_text"] or "")
+        if strat is None:
+            print(f"\n>>> {ev['genome_id'][:16]} — parse error: {err}")
+            continue
+
+        print(f"\n>>> {ev['genome_id']} (fit_IS={ev['fitness']:.4f}, sharpe_IS={ev['sharpe']:.3f})")
+        print(f"{'fold':<5} {'period':<26} {'trades':>7} {'wins':>5} {'losses':>7} {'win%':>6} {'avg_w':>9} {'avg_l':>9} {'ret':>7} {'maxDD':>7} {'sharpe':>7}")
+        for s in splits:
+            test_df = ohlcv.loc[s.test_idx]
+            try:
+                signal_fn = compile_strategy(strat)
+                signals = signal_fn(test_df)
+                bt = engine.run(test_df, signals)
+            except Exception as e:
+                print(f"  fold {s.fold}: error {e}")
+                continue
+            trades = bt.trades
+            n_trades = len(trades)
+            wins = [t.net_pnl for t in trades if t.net_pnl > 0]
+            losses = [t.net_pnl for t in trades if t.net_pnl <= 0]
+            n_wins = len(wins)
+            n_losses = len(losses)
+            win_rate = (n_wins / n_trades * 100) if n_trades else 0.0
+            avg_w = (sum(wins) / n_wins) if n_wins else 0.0
+            avg_l = (sum(losses) / n_losses) if n_losses else 0.0
+            # Normalize equity for DD/return
+            if n_trades > 0:
+                notional = float(test_df["close"].iloc[0])
+                equity_pos = (bt.equity_curve / notional) + 1.0
+                ret_pct = total_return(equity_pos)
+                dd = max_drawdown(equity_pos)
+                sr = sharpe_ratio(bt.returns, periods_per_year=8760)
+            else:
+                ret_pct = dd = sr = 0.0
+            period = f"{str(s.test_idx[0])[:10]}..{str(s.test_idx[-1])[:10]}"
+            print(f"{s.fold:<5} {period:<26} {n_trades:>7} {n_wins:>5} {n_losses:>7} {win_rate:>5.1f}% {avg_w:>9.1f} {avg_l:>9.1f} {ret_pct:>6.2%} {dd:>6.2%} {sr:>7.3f}")
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,139 @@
+"""Confronto per-trade dei 4 winner cross-run (BTC/ETH × 1h/5m).
+
+Per ogni winner: ri-esegue il backtest su 4 fold WFA expanding-window e raccoglie
+trade buoni/non buoni, win-rate, avg PnL, return, max DD, Sharpe.
+"""
+from __future__ import annotations
+
+import argparse
+from datetime import datetime
+
+import pandas as pd  # type: ignore[import-untyped]
+
+from multi_swarm_core.agents.hypothesis import _try_parse
+from multi_swarm_core.backtest.engine import BacktestEngine
+from multi_swarm_core.cerbero.client import CerberoClient
+from multi_swarm_core.config import load_settings
+from multi_swarm_core.data.cerbero_ohlcv import CerberoOHLCVLoader, OHLCVRequest
+from multi_swarm_core.data.splits import expanding_walk_forward
+from multi_swarm_core.metrics.basic import max_drawdown, sharpe_ratio, total_return
+from multi_swarm_core.persistence.repository import Repository
+from multi_swarm_core.protocol.compiler import compile_strategy
+
+
+# (run_name, genome_id, symbol, timeframe, label)
+WINNERS = [
+    ("phase1-btc-100-001", "238e481262c1594c", "BTC-PERPETUAL", "1h", "BTC 1h sharpshooter (Gen 7)"),
+    ("phase1-btc-100-001", "23a24989e2ed0f84", "BTC-PERPETUAL", "1h", "BTC 1h robust (Gen 0 elite)"),
+    ("phase1-eth-100-001", "4b45a72c13acf1d5", "ETH-PERPETUAL", "1h", "ETH 1h best-by-sharpe (killed)"),
+    ("phase1-btc-100-5m-001", "f8ca6642adf7e0cd", "BTC-PERPETUAL", "5m", "BTC 5m robust winner"),
+    ("phase1-eth-100-5m-001", "c04dff7086bb9588", "ETH-PERPETUAL", "5m", "ETH 5m OOS winner"),
+]
+
+
+def analyze_genome(run_id: str, genome_id: str, symbol: str, timeframe: str, label: str,
+                   settings, cerbero, loader) -> None:
+    repo = Repository(settings.ga_db_path)
+    repo.init_schema()
+    evs = [e for e in repo.list_evaluations(run_id) if e["genome_id"] == genome_id]
+    if not evs:
+        print(f"  no eval for {genome_id} in {run_id}")
+        return
+    ev = evs[0]
+    strat, err = _try_parse(ev.get("raw_text") or "")
+    if strat is None:
+        print(f"  parse error: {err}")
+        return
+
+    req = OHLCVRequest(
+        symbol=symbol, timeframe=timeframe,
+        start=datetime.fromisoformat("2018-09-01T00:00:00+00:00"),
+        end=datetime.fromisoformat("2026-01-01T00:00:00+00:00"),
+    )
+    ohlcv = loader.load(req)
+    splits = expanding_walk_forward(ohlcv.index, train_ratio=0.5, n_folds=4)
+    engine = BacktestEngine(fees_bp=5.0)
+
+    print(f"\n>>> {label}")
+    print(f"    {genome_id} | fit_IS={ev['fitness']:.4f} sharpe_IS={ev['sharpe']:.3f} trades_IS={ev['n_trades']}")
+    print(f"    {'fold':<5} {'period':<26} {'trades':>7} {'wins':>5} {'losses':>7} {'win%':>6} {'avg_w':>10} {'avg_l':>10} {'ret':>8} {'maxDD':>7} {'sharpe':>7}")
+
+    sum_ret = 0.0
+    sum_trades = 0
+    sum_wins = 0
+    for s in splits:
+        test_df = ohlcv.loc[s.test_idx]
+        try:
+            signal_fn = compile_strategy(strat)
+            signals = signal_fn(test_df)
+            bt = engine.run(test_df, signals)
+        except Exception as e:
+            print(f"  fold {s.fold}: error {e}")
+            continue
+
+        trades = bt.trades
+        n = len(trades)
+        wins = [t.net_pnl for t in trades if t.net_pnl > 0]
+        losses = [t.net_pnl for t in trades if t.net_pnl <= 0]
+        nw, nl = len(wins), len(losses)
+        wr = (nw / n * 100) if n else 0.0
+        aw = (sum(wins) / nw) if nw else 0.0
+        al = (sum(losses) / nl) if nl else 0.0
+        if n > 0:
+            notional = float(test_df["close"].iloc[0])
+            eq = (bt.equity_curve / notional) + 1.0
+            ret = total_return(eq)
+            dd = max_drawdown(eq)
+            sr = sharpe_ratio(bt.returns, periods_per_year=8760)
+        else:
+            ret = dd = sr = 0.0
+        period = f"{str(s.test_idx[0])[:10]}..{str(s.test_idx[-1])[:10]}"
+        print(f"    {s.fold:<5} {period:<26} {n:>7} {nw:>5} {nl:>7} {wr:>5.1f}% {aw:>10.1f} {al:>10.1f} {ret:>7.2%} {dd:>6.2%} {sr:>7.3f}")
+        sum_ret += ret
+        sum_trades += n
+        sum_wins += nw
+    overall_wr = (sum_wins / sum_trades * 100) if sum_trades else 0.0
+    print(f"    {'='*5} TOTALS:                {sum_trades:>7} {sum_wins:>5} {sum_trades-sum_wins:>7} {overall_wr:>5.1f}%  cum_ret={sum_ret:+.2%}")
+
+
+def main() -> None:
+    settings = load_settings()
+    repo = Repository(settings.ga_db_path)
+    repo.init_schema()
+    name_to_id: dict[str, str] = {}
+    for w in WINNERS:
+        run_name = w[0]
+        if run_name in name_to_id:
+            continue
+        runs = repo.list_runs()
+        for r in runs:
+            if r["name"] == run_name:
+                name_to_id[run_name] = r["id"]
+                break
+
+    token = (
+        settings.cerbero_mainnet_token.get_secret_value()
+        if settings.cerbero_mainnet_token
+        else settings.cerbero_testnet_token.get_secret_value()
+    )
+    cerbero = CerberoClient(
+        base_url=settings.cerbero_base_url,
+        token=token,
+        bot_tag=settings.cerbero_bot_tag,
+    )
+    loader = CerberoOHLCVLoader(client=cerbero, cache_dir=settings.series_dir)
+
+    print(f"{'='*120}")
+    print(f"PER-TRADE COMPARISON — {len(WINNERS)} winner candidates × 4 folds WFA")
+    print(f"{'='*120}")
+
+    for run_name, genome_id, symbol, timeframe, label in WINNERS:
+        run_id = name_to_id.get(run_name)
+        if not run_id:
+            print(f"!!! run not found: {run_name}")
+            continue
+        analyze_genome(run_id, genome_id, symbol, timeframe, label, settings, cerbero, loader)
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,112 @@
+"""2 winner cross-tick: BTC 238e4812 + ETH c04dff7086 su 5m / 15m / 1h.
+
+Per ogni combinazione strategy × timeframe: backtest year-by-year (2019-2025)
+con metriche per-anno e totale 7y.
+"""
+from __future__ import annotations
+
+from datetime import datetime
+from pathlib import Path
+
+from multi_swarm_core.backtest.engine import BacktestEngine
+from multi_swarm_core.cerbero.client import CerberoClient
+from multi_swarm_core.config import load_settings
+from multi_swarm_core.data.cerbero_ohlcv import CerberoOHLCVLoader, OHLCVRequest
+from multi_swarm_core.metrics.basic import max_drawdown, sharpe_ratio, total_return
+from multi_swarm_core.protocol.compiler import compile_strategy
+from multi_swarm_core.protocol.parser import parse_strategy
+
+
+WINNERS = [
+    # (label, path, symbol)
+    ("BTC NEW (238e4812, native=1h)", "btc_238e4812.json", "BTC-PERPETUAL"),
+    ("ETH NEW (c04dff7086, native=5m)", "eth_c04dff7086.json", "ETH-PERPETUAL"),
+]
+TIMEFRAMES = ["5m", "15m", "1h"]
+
+YEARS = [
+    ("2019", "2019-01-01T00:00:00+00:00", "2020-01-01T00:00:00+00:00"),
+    ("2020", "2020-01-01T00:00:00+00:00", "2021-01-01T00:00:00+00:00"),
+    ("2021", "2021-01-01T00:00:00+00:00", "2022-01-01T00:00:00+00:00"),
+    ("2022", "2022-01-01T00:00:00+00:00", "2023-01-01T00:00:00+00:00"),
+    ("2023", "2023-01-01T00:00:00+00:00", "2024-01-01T00:00:00+00:00"),
+    ("2024", "2024-01-01T00:00:00+00:00", "2025-01-01T00:00:00+00:00"),
+    ("2025", "2025-01-01T00:00:00+00:00", "2026-01-01T00:00:00+00:00"),
+]
+
+
+def evaluate(strat, ohlcv, engine, label, tf) -> None:
+    print(f"\n  >>> tick={tf} | {len(ohlcv)} bars")
+    print(f"  {'year':<6} {'trades':>7} {'wins':>5} {'losses':>7} {'win%':>6} {'ret':>8} {'maxDD':>7} {'sharpe':>7}")
+    sum_ret = 0.0
+    sum_trades = 0
+    sum_wins = 0
+    for year_label, start, end in YEARS:
+        mask = (ohlcv.index >= datetime.fromisoformat(start)) & (ohlcv.index < datetime.fromisoformat(end))
+        slice_df = ohlcv[mask]
+        if len(slice_df) == 0:
+            continue
+        try:
+            signal_fn = compile_strategy(strat)
+            signals = signal_fn(slice_df)
+            bt = engine.run(slice_df, signals)
+        except Exception as e:
+            print(f"  {year_label:<6} ERROR: {e}")
+            continue
+        trades = bt.trades
+        n = len(trades)
+        wins = [t.net_pnl for t in trades if t.net_pnl > 0]
+        losses = [t.net_pnl for t in trades if t.net_pnl <= 0]
+        nw, nl = len(wins), len(losses)
+        wr = (nw / n * 100) if n else 0.0
+        if n > 0:
+            notional = float(slice_df["close"].iloc[0])
+            eq = (bt.equity_curve / notional) + 1.0
+            ret = total_return(eq)
+            dd = max_drawdown(eq)
+            sr = sharpe_ratio(bt.returns, periods_per_year=8760)
+        else:
+            ret = dd = sr = 0.0
+        print(f"  {year_label:<6} {n:>7} {nw:>5} {nl:>7} {wr:>5.1f}% {ret:>7.2%} {dd:>6.2%} {sr:>7.3f}")
+        sum_ret += ret
+        sum_trades += n
+        sum_wins += nw
+    overall_wr = (sum_wins / sum_trades * 100) if sum_trades else 0.0
+    print(f"  ===== 7y TOT: {sum_trades:>7} {sum_wins:>5} {sum_trades-sum_wins:>7} {overall_wr:>5.1f}%  cum_ret={sum_ret:+.2%}")
+
+
+def main() -> None:
+    settings = load_settings()
+    token = (
+        settings.cerbero_mainnet_token.get_secret_value()
+        if settings.cerbero_mainnet_token
+        else settings.cerbero_testnet_token.get_secret_value()
+    )
+    cerbero = CerberoClient(
+        base_url=settings.cerbero_base_url,
+        token=token,
+        bot_tag=settings.cerbero_bot_tag,
+    )
+    loader = CerberoOHLCVLoader(client=cerbero, cache_dir=settings.series_dir)
+    engine = BacktestEngine(fees_bp=5.0)
+    strategies_dir = Path("/app/strategies")
+
+    for label, fname, symbol in WINNERS:
+        path = strategies_dir / fname
+        strat = parse_strategy(path.read_text())
+        print(f"\n{'='*100}")
+        print(f">>> {label} — symbol={symbol}")
+        for tf in TIMEFRAMES:
+            try:
+                ohlcv = loader.load(OHLCVRequest(
+                    symbol=symbol, timeframe=tf,
+                    start=datetime.fromisoformat("2018-09-01T00:00:00+00:00"),
+                    end=datetime.fromisoformat("2026-01-01T00:00:00+00:00"),
+                ))
+                evaluate(strat, ohlcv, engine, label, tf)
+            except Exception as e:
+                print(f"\n  >>> tick={tf} FAILED TO LOAD: {e}")
+
+
+if __name__ == "__main__":
+    main()
@@ -70,9 +70,11 @@ def load_assets(strategies_dir: Path) -> list[AssetConfig]:
        raise FileNotFoundError(
            f"Expected btc_*.json and eth_*.json in {strategies_dir}"
        )
+    # ETH winner c04dff7086 e' tunato su 5m: a 1h la strategia perde (cum_ret -33% 7y).
+    # BTC winner 238e4812 e' tunato su 1h: tick native = paper tick.
    return [
-        AssetConfig(symbol="BTC-PERPETUAL", strategy_file=btc_files[0]),
-        AssetConfig(symbol="ETH-PERPETUAL", strategy_file=eth_files[0]),
+        AssetConfig(symbol="BTC-PERPETUAL", strategy_file=btc_files[0], timeframe="1h"),
+        AssetConfig(symbol="ETH-PERPETUAL", strategy_file=eth_files[0], timeframe="5m"),
    ]


@@ -1,16 +1,48 @@
 from __future__ import annotations

 import argparse
+import importlib.resources
 from datetime import datetime
+from pathlib import Path

 from multi_swarm_core.cerbero.client import CerberoClient
 from multi_swarm_core.config import load_settings
 from multi_swarm_core.data.cerbero_ohlcv import CerberoOHLCVLoader, OHLCVRequest
 from multi_swarm_core.genome.hypothesis import ModelTier
+from multi_swarm_core.genome.prompt_library import PromptLibrary
 from multi_swarm_core.llm.client import LLMClient
 from multi_swarm_core.orchestrator.run import RunConfig, run_phase1


+def _default_prompt_library_path() -> Path:
+    """Default: prompts.json shippato col package strategy_crypto."""
+    return Path(str(importlib.resources.files("strategy_crypto") / "prompts.json"))
+
+
+# Default v2 hard-kill list: oltre ai degenerate originali, fees_eat_alpha e
+# negative_net_pnl sono deal-breaker non recuperabili via soft penalty (vedi
+# 7y-overfit incident 2026-05-16: elite IS Sharpe 1.93 -> net -5% su 7y per fees).
+_DEFAULT_V2_HARD_KILL: tuple[str, ...] = (
+    "no_trades",
+    "degenerate",
+    "undertrading",
+    "fees_eat_alpha",
+    "negative_net_pnl",
+)
+
+
+def _resolve_hard_kill(args) -> tuple[str, ...] | None:
+    """Resolve la lista hard-kill da CLI args.
+
+    Priority: ``--fitness-hard-kill`` esplicito > default v2 > ``None`` (v1).
+    """
+    if args.fitness_hard_kill:
+        return tuple(s.strip() for s in args.fitness_hard_kill.split(",") if s.strip())
+    if args.fitness_v2:
+        return _DEFAULT_V2_HARD_KILL
+    return None
+
+
 def parse_args() -> argparse.Namespace:
    p = argparse.ArgumentParser(description="Multi-Swarm Phase 1 runner")
    p.add_argument("--name", default="phase1-spike-001")
@@ -27,7 +59,10 @@ def parse_args() -> argparse.Namespace:
    )
    p.add_argument("--symbol", default="BTC-PERPETUAL")
    p.add_argument("--timeframe", default="1h")
-    p.add_argument("--start", default="2024-01-01T00:00:00+00:00")
+    # Default esteso a 7.3 anni: copre bear 2018-19, halving 2020, COVID,
+    # ATH 2021, winter 2022, ETF rally 2024, regime corrente. Una finestra
+    # corta lascia il GA libero di overfit a un singolo regime.
+    p.add_argument("--start", default="2018-09-01T00:00:00+00:00")
    p.add_argument("--end", default="2026-01-01T00:00:00+00:00")
    p.add_argument("--fees-bp", type=float, default=5.0)
    p.add_argument("--n-trials-dsr", type=int, default=50)
@@ -59,8 +94,10 @@ def parse_args() -> argparse.Namespace:
        "--fitness-v2",
        action="store_true",
        help=(
-            "Attiva fitness v2: solo {no_trades, degenerate, undertrading} azzerano; "
-            "gli altri HIGH applicano soft penalty multiplicativa"
+            "Attiva fitness v2: hard-kill su {no_trades, degenerate, undertrading, "
+            "fees_eat_alpha, negative_net_pnl}; gli altri HIGH applicano soft penalty "
+            "multiplicativa. Versione hardened post 7y-overfit incident: fees + "
+            "net negativo non sono recuperabili."
        ),
    )
    p.add_argument(
@@ -69,6 +106,16 @@ def parse_args() -> argparse.Namespace:
        default=0.4,
        help="v2: fattore soft penalty per HIGH non-hard (default 0.4 → 1 HIGH → 0.71x)",
    )
+    p.add_argument(
+        "--fitness-hard-kill",
+        type=str,
+        default=None,
+        help=(
+            "Override comma-separated della lista di finding name che azzerano la "
+            "fitness in modalita' v2. Es: 'no_trades,degenerate'. Default v2: "
+            "no_trades,degenerate,undertrading,fees_eat_alpha,negative_net_pnl."
+        ),
+    )
    p.add_argument(
        "--wfa-train-split",
        type=float,
@@ -96,6 +143,26 @@ def parse_args() -> argparse.Namespace:
        default=0.5,
        help="Multi-objective: peso IS (1-alpha = OOS). 1.0=solo IS, 0.5=bilanciato, 0.0=solo OOS",
    )
+    p.add_argument(
+        "--prompt-library",
+        type=Path,
+        default=None,
+        help=(
+            "Path al file JSON con stili cognitivi + direttive system_prompt. "
+            "Default: strategy_crypto/prompts.json shippato col package. "
+            "Schema: {styles: {<name>: {directive: <testo>}}}"
+        ),
+    )
+    p.add_argument(
+        "--llm-concurrency",
+        type=int,
+        default=1,
+        help=(
+            "Numero di propose() LLM concorrenti per generazione (default 1 = "
+            "serial). 6-10 tipicamente accettati da OpenRouter qwen-2.5 senza "
+            "rate-limit; riduce wall time GA loop di 5-8x."
+        ),
+    )
    return p.parse_args()


@@ -103,6 +170,13 @@ def main() -> None:
    args = parse_args()
    settings = load_settings()

+    prompt_lib_path = args.prompt_library or _default_prompt_library_path()
+    prompt_library = PromptLibrary.from_json(prompt_lib_path)
+    print(
+        f"PromptLibrary loaded from {prompt_lib_path}: "
+        f"{len(prompt_library.styles)} stili ({', '.join(prompt_library.cognitive_styles)})"
+    )
+
    token = (
        settings.cerbero_mainnet_token.get_secret_value()
        if settings.cerbero_mainnet_token
@@ -153,14 +227,14 @@ def main() -> None:
        fees_eat_alpha_threshold=args.fees_eat_alpha_threshold,
        flat_too_long_threshold=args.flat_too_long_threshold,
        undertrading_threshold=args.undertrading_threshold,
-        fitness_hard_kill_findings=(
-            ("no_trades", "degenerate", "undertrading") if args.fitness_v2 else None
-        ),
+        fitness_hard_kill_findings=_resolve_hard_kill(args),
        fitness_adversarial_soft_penalty=args.fitness_soft_penalty,
        wfa_train_split=args.wfa_train_split,
        wfa_top_k=args.wfa_top_k,
        eval_oos_during_loop=args.eval_oos_during_loop,
        fitness_combined_alpha=args.fitness_combined_alpha,
+        prompt_library=prompt_library,
+        llm_concurrency=args.llm_concurrency,
    )

    run_id = run_phase1(cfg, ohlcv=ohlcv, llm=llm)
@@ -0,0 +1,271 @@
+"""Multi-fold validation di un run esistente.
+
+Prende un ``run_id`` da ``state/runs.db``, seleziona i top-K genomi per fitness IS,
+e li rivaluta su N fold expanding-window di un dataset OHLCV (tipicamente piu'
+lungo del train del GA). Output: per-fold + aggregati (mean / min / std) della
+fitness OOS.
+
+Use case: il GA puo' selezionare un "lucky-shot" overfit a uno specifico regime.
+Validare i top-K su finestre temporali diverse rivela quali strategie sono
+robuste vs overfitter.
+
+Esempio::
+
+  python scripts/validate_run.py \\
+    --run-id e263651598894da688d95fda90a34a96 \\
+    --top-k 10 --n-folds 4 \\
+    --symbol BTC-PERPETUAL --timeframe 1h \\
+    --start 2018-09-01 --end 2026-01-01
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import statistics
+from datetime import datetime
+from pathlib import Path
+
+import pandas as pd  # type: ignore[import-untyped]
+
+from multi_swarm_core.agents.adversarial import AdversarialAgent
+from multi_swarm_core.agents.falsification import FalsificationAgent
+from multi_swarm_core.agents.hypothesis import _try_parse
+from multi_swarm_core.cerbero.client import CerberoClient
+from multi_swarm_core.config import load_settings
+from multi_swarm_core.data.cerbero_ohlcv import CerberoOHLCVLoader, OHLCVRequest
+from multi_swarm_core.data.splits import expanding_walk_forward
+from multi_swarm_core.ga.fitness import compute_fitness
+from multi_swarm_core.persistence.repository import Repository
+
+
+def parse_args() -> argparse.Namespace:
+    p = argparse.ArgumentParser(description="Multi-fold validation di top-K genomi")
+    p.add_argument("--run-id", required=True, help="run_id da validare")
+    p.add_argument("--top-k", type=int, default=10, help="quanti genomi top valutare")
+    p.add_argument("--n-folds", type=int, default=4, help="numero fold expanding-window")
+    p.add_argument(
+        "--train-ratio",
+        type=float,
+        default=0.5,
+        help="frazione iniziale per il train iniziale (folds testano la coda)",
+    )
+    p.add_argument("--symbol", default="BTC-PERPETUAL")
+    p.add_argument("--timeframe", default="1h")
+    p.add_argument("--exchange", default="deribit", choices=["deribit", "bybit", "hyperliquid"])
+    p.add_argument("--start", default="2018-09-01T00:00:00+00:00")
+    p.add_argument("--end", default="2026-01-01T00:00:00+00:00")
+    p.add_argument("--fees-bp", type=float, default=5.0)
+    p.add_argument("--n-trials-dsr", type=int, default=50)
+    p.add_argument(
+        "--fees-eat-alpha-threshold", type=float, default=0.5,
+    )
+    p.add_argument(
+        "--flat-too-long-threshold", type=float, default=0.95,
+    )
+    p.add_argument(
+        "--undertrading-threshold", type=int, default=10,
+    )
+    p.add_argument(
+        "--fitness-v2", action="store_true",
+        help="Coerente con --fitness-v2 del run originale",
+    )
+    p.add_argument(
+        "--fitness-soft-penalty", type=float, default=0.4,
+    )
+    p.add_argument(
+        "--output-json",
+        type=Path,
+        default=None,
+        help="Path JSON dove salvare i risultati (default: stdout solo)",
+    )
+    return p.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+    settings = load_settings()
+
+    # Repository: top-K genomi per fitness IS, con raw_text parsable.
+    repo = Repository(settings.ga_db_path)
+    repo.init_schema()
+    run = repo.get_run(args.run_id)
+    if run is None:
+        raise SystemExit(f"run_id non trovato: {args.run_id}")
+    print(f"Validating run: {run['name']} ({args.run_id})")
+    print(f"  status: {run['status']}, cost: ${run['total_cost_usd']:.4f}")
+
+    all_evals = repo.list_evaluations(args.run_id)
+    parseable = [
+        e for e in all_evals
+        if e.get("raw_text") and not e.get("parse_error") and e["fitness"] > 0
+    ]
+    parseable.sort(key=lambda e: e["fitness"], reverse=True)
+
+    # Dedup by genome_id (gli elite vengono salvati una sola volta ma possono apparire
+    # in evaluations multiple se rivalutati con eval_oos_during_loop).
+    seen_ids: set[str] = set()
+    top_genomes: list[dict] = []
+    for e in parseable:
+        if e["genome_id"] in seen_ids:
+            continue
+        seen_ids.add(e["genome_id"])
+        top_genomes.append(e)
+        if len(top_genomes) >= args.top_k:
+            break
+    print(f"  selected top-{len(top_genomes)} genomes for validation")
+
+    # OHLCV: carica il dataset esteso.
+    token = (
+        settings.cerbero_mainnet_token.get_secret_value()
+        if settings.cerbero_mainnet_token
+        else settings.cerbero_testnet_token.get_secret_value()
+    )
+    cerbero = CerberoClient(
+        base_url=settings.cerbero_base_url,
+        token=token,
+        bot_tag=settings.cerbero_bot_tag,
+    )
+    loader = CerberoOHLCVLoader(client=cerbero, cache_dir=settings.series_dir)
+    req = OHLCVRequest(
+        symbol=args.symbol,
+        timeframe=args.timeframe,
+        start=datetime.fromisoformat(args.start),
+        end=datetime.fromisoformat(args.end),
+        exchange=args.exchange,
+    )
+    ohlcv = loader.load(req)
+    print(f"  OHLCV: {len(ohlcv)} bars from {ohlcv.index[0]} to {ohlcv.index[-1]}")
+
+    splits = expanding_walk_forward(
+        ohlcv.index, train_ratio=args.train_ratio, n_folds=args.n_folds,
+    )
+    print(f"  generated {len(splits)} folds")
+    for s in splits:
+        print(f"    fold {s.fold}: test [{s.test_idx[0]} -> {s.test_idx[-1]}] ({len(s.test_idx)} bars)")
+
+    fals_agent = FalsificationAgent(fees_bp=args.fees_bp, n_trials_dsr=args.n_trials_dsr)
+    adv_agent = AdversarialAgent(
+        fees_bp=args.fees_bp,
+        fees_eat_alpha_threshold=args.fees_eat_alpha_threshold,
+        flat_too_long_threshold=args.flat_too_long_threshold,
+        undertrading_threshold=args.undertrading_threshold,
+    )
+    hard_kill = (
+        ("no_trades", "degenerate", "undertrading") if args.fitness_v2 else None
+    )
+
+    # Itera per ogni genome + fold.
+    results: list[dict] = []
+    for gi, ev in enumerate(top_genomes):
+        strategy, parse_err = _try_parse(ev["raw_text"] or "")
+        if strategy is None:
+            print(f"  [{gi}] {ev['genome_id'][:12]} skip (parse error: {parse_err})")
+            continue
+
+        per_fold: list[dict] = []
+        for s in splits:
+            test_df = ohlcv.loc[s.test_idx]
+            try:
+                fals = fals_agent.evaluate(strategy, test_df)
+                adv = adv_agent.review(strategy, test_df)
+                fit = compute_fitness(
+                    fals, adv,
+                    hard_kill_findings=hard_kill,
+                    adversarial_soft_penalty=args.fitness_soft_penalty,
+                )
+            except Exception as e:
+                print(f"    fold {s.fold} eval failed: {e}")
+                continue
+            per_fold.append({
+                "fold": s.fold,
+                "fitness": float(fit),
+                "sharpe": float(fals.sharpe),
+                "dsr": float(fals.dsr),
+                "dsr_pvalue": float(fals.dsr_pvalue),
+                "return": float(fals.total_return),
+                "max_dd": float(fals.max_drawdown),
+                "n_trades": int(fals.n_trades),
+                "test_start": str(s.test_idx[0]),
+                "test_end": str(s.test_idx[-1]),
+            })
+
+        if not per_fold:
+            continue
+
+        fits = [pf["fitness"] for pf in per_fold]
+        sharps = [pf["sharpe"] for pf in per_fold]
+        results.append({
+            "genome_id": ev["genome_id"],
+            "fitness_is": float(ev["fitness"]),
+            "sharpe_is": float(ev["sharpe"]),
+            "folds": per_fold,
+            "fitness_oos_mean": statistics.mean(fits),
+            "fitness_oos_min": min(fits),
+            "fitness_oos_max": max(fits),
+            "fitness_oos_std": statistics.pstdev(fits) if len(fits) > 1 else 0.0,
+            "sharpe_oos_mean": statistics.mean(sharps),
+            "sharpe_oos_min": min(sharps),
+            "robust_score": min(fits),  # min across folds = pessimismo
+        })
+
+    # Ranking finale: per robust_score (min fitness) decrescente.
+    results.sort(key=lambda r: r["robust_score"], reverse=True)
+
+    print()
+    print(f"{'='*120}")
+    print(f"VALIDATION RESULTS ({len(results)} genomes, {len(splits)} folds)")
+    print(f"{'='*120}")
+    print(
+        f"{'rank':>4} {'genome':12} {'fit_is':>8} {'sh_is':>7} "
+        f"{'fit_mean':>9} {'fit_min':>8} {'fit_max':>8} {'fit_std':>8} "
+        f"{'sh_mean':>8} {'sh_min':>8} {'robust':>7}"
+    )
+    print("-" * 120)
+    for rank, r in enumerate(results, 1):
+        print(
+            f"{rank:>4} {r['genome_id'][:12]:12} "
+            f"{r['fitness_is']:>8.4f} {r['sharpe_is']:>7.3f} "
+            f"{r['fitness_oos_mean']:>9.4f} {r['fitness_oos_min']:>8.4f} "
+            f"{r['fitness_oos_max']:>8.4f} {r['fitness_oos_std']:>8.4f} "
+            f"{r['sharpe_oos_mean']:>8.3f} {r['sharpe_oos_min']:>8.3f} "
+            f"{r['robust_score']:>7.4f}"
+        )
+
+    if results:
+        winner = results[0]
+        print()
+        print(f"ROBUST WINNER: {winner['genome_id']}")
+        print(f"  fitness_is={winner['fitness_is']:.4f}, "
+              f"fitness_oos_min={winner['fitness_oos_min']:.4f}, "
+              f"fitness_oos_mean={winner['fitness_oos_mean']:.4f}")
+        print(f"  sharpe_is={winner['sharpe_is']:.3f}, "
+              f"sharpe_oos_min={winner['sharpe_oos_min']:.3f}")
+        print(f"  per-fold breakdown:")
+        for pf in winner["folds"]:
+            print(
+                f"    fold {pf['fold']} [{pf['test_start'][:10]} .. {pf['test_end'][:10]}]: "
+                f"fit={pf['fitness']:.4f} sharpe={pf['sharpe']:.3f} "
+                f"ret={pf['return']:.3f} n_trades={pf['n_trades']}"
+            )
+
+    if args.output_json:
+        payload = {
+            "run_id": args.run_id,
+            "run_name": run["name"],
+            "n_folds": len(splits),
+            "top_k_requested": args.top_k,
+            "top_k_evaluated": len(results),
+            "symbol": args.symbol,
+            "timeframe": args.timeframe,
+            "start": args.start,
+            "end": args.end,
+            "ohlcv_bars": len(ohlcv),
+            "results": results,
+        }
+        args.output_json.write_text(json.dumps(payload, indent=2, default=str))
+        print(f"\nResults saved to: {args.output_json}")
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,112 @@
+"""Per-year breakdown delle 4 strategie: 2 NEW (BTC 238e4812 + ETH c04dff7086)
+ 2 OLD freezate (btc_9cf506b8 hardened-001 + eth_facd6af85d5d).
+
+Backtest anno-per-anno (2019-2025) sul tick di discovery di ciascuna strategia.
+Output: trade, wins/losses, win%, return%, max DD%, Sharpe per ogni anno.
+"""
+from __future__ import annotations
+
+from datetime import datetime
+from pathlib import Path
+
+from multi_swarm_core.backtest.engine import BacktestEngine
+from multi_swarm_core.cerbero.client import CerberoClient
+from multi_swarm_core.config import load_settings
+from multi_swarm_core.data.cerbero_ohlcv import CerberoOHLCVLoader, OHLCVRequest
+from multi_swarm_core.metrics.basic import max_drawdown, sharpe_ratio, total_return
+from multi_swarm_core.protocol.compiler import compile_strategy
+from multi_swarm_core.protocol.parser import parse_strategy
+
+
+STRATEGIES = [
+    # (label, path, symbol, timeframe)
+    ("BTC NEW (238e4812, paper attuale)", "btc_238e4812.json", "BTC-PERPETUAL", "1h"),
+    ("BTC OLD (9cf506b8, hardened-001 prev paper)", "archive/btc_9cf506b8.json", "BTC-PERPETUAL", "1h"),
+    ("ETH NEW (c04dff7086, paper attuale)", "eth_c04dff7086.json", "ETH-PERPETUAL", "5m"),
+    ("ETH OLD (facd6af85d5d, prev paper)", "archive/eth_facd6af85d5d.json", "ETH-PERPETUAL", "1h"),
+]
+
+YEARS = [
+    ("2019", "2019-01-01T00:00:00+00:00", "2020-01-01T00:00:00+00:00"),
+    ("2020", "2020-01-01T00:00:00+00:00", "2021-01-01T00:00:00+00:00"),
+    ("2021", "2021-01-01T00:00:00+00:00", "2022-01-01T00:00:00+00:00"),
+    ("2022", "2022-01-01T00:00:00+00:00", "2023-01-01T00:00:00+00:00"),
+    ("2023", "2023-01-01T00:00:00+00:00", "2024-01-01T00:00:00+00:00"),
+    ("2024", "2024-01-01T00:00:00+00:00", "2025-01-01T00:00:00+00:00"),
+    ("2025", "2025-01-01T00:00:00+00:00", "2026-01-01T00:00:00+00:00"),
+]
+
+
+def main() -> None:
+    settings = load_settings()
+    token = (
+        settings.cerbero_mainnet_token.get_secret_value()
+        if settings.cerbero_mainnet_token
+        else settings.cerbero_testnet_token.get_secret_value()
+    )
+    cerbero = CerberoClient(
+        base_url=settings.cerbero_base_url,
+        token=token,
+        bot_tag=settings.cerbero_bot_tag,
+    )
+    loader = CerberoOHLCVLoader(client=cerbero, cache_dir=settings.series_dir)
+    engine = BacktestEngine(fees_bp=5.0)
+    strategies_dir = Path("/app/strategies")
+
+    for label, fname, symbol, timeframe in STRATEGIES:
+        path = strategies_dir / fname
+        strat = parse_strategy(path.read_text())
+
+        # Carica intero range una volta sola
+        ohlcv = loader.load(OHLCVRequest(
+            symbol=symbol, timeframe=timeframe,
+            start=datetime.fromisoformat("2018-09-01T00:00:00+00:00"),
+            end=datetime.fromisoformat("2026-01-01T00:00:00+00:00"),
+        ))
+
+        print(f"\n{'=' * 110}")
+        print(f">>> {label}")
+        print(f"    symbol={symbol} timeframe={timeframe} | {len(ohlcv)} bars total")
+        print(f"    {'year':<6} {'bars':>6} {'trades':>7} {'wins':>5} {'losses':>7} {'win%':>6} {'avg_w':>10} {'avg_l':>10} {'ret':>8} {'maxDD':>7} {'sharpe':>7}")
+
+        sum_ret = 0.0
+        sum_trades = 0
+        sum_wins = 0
+        for year_label, start, end in YEARS:
+            mask = (ohlcv.index >= datetime.fromisoformat(start)) & (ohlcv.index < datetime.fromisoformat(end))
+            slice_df = ohlcv[mask]
+            if len(slice_df) == 0:
+                continue
+            try:
+                signal_fn = compile_strategy(strat)
+                signals = signal_fn(slice_df)
+                bt = engine.run(slice_df, signals)
+            except Exception as e:
+                print(f"    {year_label:<6} ERROR: {e}")
+                continue
+            trades = bt.trades
+            n = len(trades)
+            wins = [t.net_pnl for t in trades if t.net_pnl > 0]
+            losses = [t.net_pnl for t in trades if t.net_pnl <= 0]
+            nw, nl = len(wins), len(losses)
+            wr = (nw / n * 100) if n else 0.0
+            aw = (sum(wins) / nw) if nw else 0.0
+            al = (sum(losses) / nl) if nl else 0.0
+            if n > 0:
+                notional = float(slice_df["close"].iloc[0])
+                eq = (bt.equity_curve / notional) + 1.0
+                ret = total_return(eq)
+                dd = max_drawdown(eq)
+                sr = sharpe_ratio(bt.returns, periods_per_year=8760)
+            else:
+                ret = dd = sr = 0.0
+            print(f"    {year_label:<6} {len(slice_df):>6} {n:>7} {nw:>5} {nl:>7} {wr:>5.1f}% {aw:>10.1f} {al:>10.1f} {ret:>7.2%} {dd:>6.2%} {sr:>7.3f}")
+            sum_ret += ret
+            sum_trades += n
+            sum_wins += nw
+        overall_wr = (sum_wins / sum_trades * 100) if sum_trades else 0.0
+        print(f"    {'='*5} TOTALS 7y:                 {sum_trades:>7} {sum_wins:>5} {sum_trades-sum_wins:>7} {overall_wr:>5.1f}%                       cum_ret={sum_ret:+.2%}")
+
+
+if __name__ == "__main__":
+    main()
@@ -172,10 +172,11 @@ class AdversarialAgent:
                )
            )

-        # Fees-eat-alpha: gross_pnl > 0 ma fees > 50% del lordo.
+        # Fees-eat-alpha: gross_pnl > 0 ma fees > soglia del lordo.
        # La strategia ha edge teorico ma il margine viene mangiato dai
        # costi di transazione: non sostenibile in produzione.
-        # Se gross_pnl <= 0 il check non si applica (gia' perdente).
+        # Se gross_pnl <= 0 il check non si applica (la condizione e' coperta
+        # da ``negative_net_pnl`` sotto).
        gross_pnl = sum(t.gross_pnl for t in result.trades)
        total_fees = sum(t.fees for t in result.trades)
        if gross_pnl > 0 and total_fees / gross_pnl > self._fees_eat_alpha_threshold:
@@ -190,4 +191,22 @@ class AdversarialAgent:
                )
            )

+        # Negative-net-pnl: somma di ``trade.net_pnl`` < 0 sul training.
+        # Cattura sia il caso "gross negativo" (no edge direzionale) sia il
+        # caso "gross positivo ma fees superiori a gross" (edge insufficiente).
+        # Sintesi del net-after-fees su finestra continua: deal-breaker, non
+        # negoziabile via soft penalty.
+        net_pnl = gross_pnl - total_fees
+        if net_pnl < 0:
+            report.findings.append(
+                Finding(
+                    name="negative_net_pnl",
+                    severity=Severity.HIGH,
+                    detail=(
+                        f"Net PnL ${net_pnl:.2f} < 0 after fees over {n_bars} bars; "
+                        f"gross ${gross_pnl:.2f}, fees ${total_fees:.2f}"
+                    ),
+                )
+            )
+
        return report
@@ -2,10 +2,12 @@ from __future__ import annotations

 import re
 from dataclasses import dataclass, field
+from typing import Any

 import openai

 from ..genome.hypothesis import HypothesisAgentGenome
+from ..genome.prompt_library import PromptLibrary
 from ..llm.client import CompletionResult, EmptyCompletionError, LLMClient
 from ..protocol.parser import ParseError, Strategy, parse_strategy
 from ..protocol.validator import ValidationError, validate_strategy
@@ -27,6 +29,13 @@ class MarketSummary:
    vol_percentile: float = 50.0         # 0-100 percentile della vol corrente
    seasonality_hour: float = 0.0        # 0-1 varianza spiegata da hour
    seasonality_dow: float = 0.0         # 0-1 varianza spiegata da dow
+    efficiency_ratio: float = 0.0        # Kaufman, 0-1
+    tail_index_left: float = 5.0         # Hill left tail
+    tail_index_right: float = 5.0        # Hill right tail
+    structural_uptrend: float = 0.5      # HH/HL score 0-1
+    compression: float = 1.0             # range recent / range ref
+    spectral_entropy: float = 1.0        # 0-1, Shannon FFT normalizzata
+    dominant_cycle: float | None = None  # periodo barre, None se spectrum piatto


@dataclass(frozen=True)
@@ -47,12 +56,9 @@ class HypothesisProposal:
    n_attempts: int = 1


-SYSTEM_TEMPLATE = """\
-Sei un agente generatore di ipotesi di trading quantitativo per un sistema swarm.
-
-Il tuo stile cognitivo: {cognitive_style}
-Direttiva personale: {system_prompt}
+# === CORE SCAFFOLD constants (universal, legato al protocol/compiler) ===

+_SYSTEM_GRAMMAR_SPEC = """\
 Devi proporre una strategia di trading espressa in JSON STRETTO.
 La risposta deve essere un singolo oggetto JSON dentro fence ```json...```
 con questa shape:
@@ -127,35 +133,18 @@ Esempi di gating temporale:
  {{"op": "eq", "args": [{{"kind": "feature", "name": "is_weekend"}}, {{"kind": "literal", "value": 1}}]}}

 Leaf - letterale numerico:
-  {{"kind": "literal", "value": 70.0}}
-
-PATTERN GUIDANCE (oltre agli indicatori, considera forma delle curve e ripetibilità):
-
-  Forme di curva:
-  - Trend ascendente: SMA(short) > SMA(long) E close > SMA(short)
-  - Trend discendente: SMA(short) < SMA(long) E close < SMA(short)
-  - Compressione di volatilità (pre-breakout): atr_pct(N) < 0.01 (sotto 1% del prezzo)
-  - Espansione di volatilità: atr_pct(N) > 0.03 (sopra 3%) OPPURE ATR(N) > ATR(N*2) confronto relativo
-  - Mean reversion strutturale: sma_pct(long) > 0.05 (close 5% sopra media) OR sma_pct(long) < -0.05
-  - Momentum positivo conferma: macd_pct(12,26,9) > 0.005 (> 0.5% del prezzo)
-
-  Ripetibilità dell'andamento:
-  - Eventi crossover/crossunder ricorrenti (golden/death cross, RSI cross zone)
-  - Pattern intra-day: usa 'hour' per sfruttare orari di forte volatilità ricorrente
-  - Pattern settimanali: usa 'dow' o 'is_weekend' per cicli mercato
-  - Doppio top approx: RSI > 70 + crossunder RSI 70 (1° picco), poi entro N bar
-    nuovo crossover RSI 70 a livello close simile → entry short
-  - Range breakout: close > SMA(N) con SMA(short) > SMA(long) (compressione + spinta)
-
-  Cerca pattern che si REPLICANO nei dati storici, non singoli eventi rari.
+  {{"kind": "literal", "value": 70.0}}"""

+_SYSTEM_CONSTRAINTS = """\
 VINCOLI
 - Gli indicator NON sono annidabili: 'params' accetta solo numeri, mai altri nodi.
 - Le regole sono valutate in ordine; la prima che matcha vince per ogni timestamp.
 - Default action se nessuna regola matcha = flat.
 - 'op' e 'kind' sono mutuamente esclusivi sullo stesso nodo.

-Rispondi SOLO con il fence ```json...``` contenente l'oggetto strategy.
+Rispondi SOLO con il fence ```json...``` contenente l'oggetto strategy."""
+
+_SYSTEM_EXAMPLE = """\
 Esempio:

 ```json
@@ -177,8 +166,49 @@ Esempio:
    }}
  ]
 }}
-```
-"""
+```"""
+
+
+def _build_system_prompt(lib: PromptLibrary, genome: HypothesisAgentGenome) -> str:
+    """Compone il SYSTEM prompt da scaffold core + contenuto strategy-specific."""
+    parts: list[str] = []
+    # 1. Header strategy-specific (da prompts.json)
+    parts.append(lib.agent_role)
+    parts.append("")
+    # 2. Cognitive style + directive (sempre, dal genome)
+    parts.append(f"Il tuo stile cognitivo: {genome.cognitive_style}")
+    parts.append(f"Direttiva personale: {genome.system_prompt}")
+    parts.append("")
+    # 3. Grammar spec (core scaffold)
+    parts.append(_SYSTEM_GRAMMAR_SPEC)
+    # 4. Pattern guidance (da prompts.json, opzionale)
+    if lib.pattern_guidance:
+        parts.append(
+            "\nPATTERN GUIDANCE (oltre agli indicatori, considera forma delle curve e ripetibilità):\n"
+        )
+        parts.append(lib.pattern_guidance)
+        parts.append(
+            "\n  Cerca pattern che si REPLICANO nei dati storici, non singoli eventi rari.\n"
+        )
+    # 5. Domain warnings (da prompts.json, opzionale)
+    if lib.domain_warnings:
+        parts.append("\nWARNING DI DOMINIO:\n")
+        parts.append(lib.domain_warnings)
+        parts.append("")
+    # 6. Vincoli (core scaffold)
+    parts.append(_SYSTEM_CONSTRAINTS)
+    # 7. NEW v3.1: anti-pattern e output priorities (da prompts.json, opzionali)
+    if lib.anti_patterns:
+        parts.append("\nANTI-PATTERN DA EVITARE:\n")
+        parts.append(lib.anti_patterns)
+        parts.append("")
+    if lib.output_priorities:
+        parts.append("\nPRIORITA' DI OUTPUT (trade-off):\n")
+        parts.append(lib.output_priorities)
+        parts.append("")
+    # 8. Esempio (core scaffold)
+    parts.append(_SYSTEM_EXAMPLE)
+    return "\n".join(parts)


 USER_TEMPLATE = """\
@@ -193,13 +223,66 @@ Regime recente (ultime 500 barre):
  vol_pct: {vol_percentile:.0f}° percentile storico
  stagionalita: hour={seasonality_hour:.2f}, dow={seasonality_dow:.2f}  (0-1, varianza spiegata)

+Geometria & frattali:
+  efficiency_ratio: {efficiency_ratio:.3f}  (Kaufman, 0=noise, 1=trend efficiente)
+  tail_index: left={tail_index_left:.2f}, right={tail_index_right:.2f}  (Hill; <2 fat tail, >5 light)
+  structural_uptrend: {structural_uptrend:.2f}  (HH/HL score, 0.5=range)
+  compression: {compression:.2f}  (range recent / ref; <1 compressione, >1 espansione)
+  spectral_entropy: {spectral_entropy:.2f}  (0=struttura, 1=rumore bianco)
+  dominant_cycle: {dominant_cycle_str}  (None se spettro piatto)
+
 Feature accessibili dal tuo genoma: {feature_access}.
 Lookback massimo che puoi usare nel ragionamento: {lookback_window} barre.

-Genera una strategia che cerchi anomalie sfruttabili in questo regime.
+{instruction}
 """


+def _render_focus_block(keys: list[str], market: MarketSummary) -> str:
+    """Renderizza 'Focus per la tua lente' come riga del USER_TEMPLATE.
+
+    Mappa nomi simbolici a valori della MarketSummary. Skippa silenziosamente
+    chiavi sconosciute (fault-tolerant per evoluzione futura).
+    """
+    field_map: dict[str, Any] = {
+        # Statistiche base
+        "mean": market.return_mean,
+        "std": market.return_std,
+        "skew": market.skew,
+        "kurt": market.kurtosis,
+        "vol_regime": market.volatility_regime,
+        # Regime recente
+        "autocorr_recent": market.autocorr_lag1_recent,
+        "autocorr_baseline": market.autocorr_lag1_baseline,
+        "hurst": market.hurst_recent,
+        "vol_pct": market.vol_percentile,
+        "seasonality_hour": market.seasonality_hour,
+        "seasonality_dow": market.seasonality_dow,
+        # Geometria/frattali
+        "efficiency_ratio": market.efficiency_ratio,
+        "tail_left": market.tail_index_left,
+        "tail_right": market.tail_index_right,
+        "structural_uptrend": market.structural_uptrend,
+        "compression": market.compression,
+        "spectral_entropy": market.spectral_entropy,
+        "dominant_cycle": market.dominant_cycle,
+    }
+    parts: list[str] = []
+    for k in keys:
+        if k not in field_map:
+            continue
+        v = field_map[k]
+        if v is None:
+            parts.append(f"{k}=N/A")
+        elif isinstance(v, float):
+            parts.append(f"{k}={v:.3f}")
+        else:
+            parts.append(f"{k}={v}")
+    if not parts:
+        return ""
+    return "\nFocus per la tua lente: " + ", ".join(parts) + "\n"
+
+
 _RETRY_TEMPLATE = """\
 {original_user}

@@ -282,20 +365,34 @@ def _try_parse(text: str) -> tuple[Strategy | None, str | None]:


 class HypothesisAgent:
-    def __init__(self, llm: LLMClient, max_retries: int = 1):
+    def __init__(
+        self,
+        llm: LLMClient,
+        max_retries: int = 1,
+        prompt_library: PromptLibrary | None = None,
+    ):
        if max_retries < 0:
            raise ValueError("max_retries must be >= 0")
        self._llm = llm
        self._max_retries = max_retries
+        self._prompt_library = prompt_library or PromptLibrary.default()

    def propose(
        self,
        genome: HypothesisAgentGenome,
        market: MarketSummary,
    ) -> HypothesisProposal:
-        system = SYSTEM_TEMPLATE.format(
-            cognitive_style=genome.cognitive_style,
-            system_prompt=genome.system_prompt,
+        system = _build_system_prompt(self._prompt_library, genome)
+        dominant_cycle_str = (
+            f"{market.dominant_cycle:.0f} barre"
+            if market.dominant_cycle is not None
+            else "N/A (spettro piatto)"
+        )
+        focus_keys = self._prompt_library.focus_metrics_for(genome.cognitive_style)
+        focus_block = _render_focus_block(focus_keys, market) if focus_keys else ""
+        instruction = (
+            self._prompt_library.instruction
+            or "Genera una strategia che cerchi anomalie sfruttabili in questo regime."
        )
        original_user = USER_TEMPLATE.format(
            symbol=market.symbol,
@@ -312,9 +409,17 @@ class HypothesisAgent:
            vol_percentile=market.vol_percentile,
            seasonality_hour=market.seasonality_hour,
            seasonality_dow=market.seasonality_dow,
+            efficiency_ratio=market.efficiency_ratio,
+            tail_index_left=market.tail_index_left,
+            tail_index_right=market.tail_index_right,
+            structural_uptrend=market.structural_uptrend,
+            compression=market.compression,
+            spectral_entropy=market.spectral_entropy,
+            dominant_cycle_str=dominant_cycle_str,
            feature_access=", ".join(genome.feature_access),
            lookback_window=genome.lookback_window,
-        )
+            instruction=instruction,
+        ) + focus_block

        completions: list[CompletionResult] = []
        errors: list[str] = []
@@ -5,8 +5,13 @@ from scipy import stats  # type: ignore[import-untyped]

 from ..metrics.basic import (
    autocorr_lag1,
+    compression_ratio,
+    efficiency_ratio_kaufman,
    hurst_exponent,
    seasonality_strength,
+    spectral_entropy_and_cycle,
+    structural_uptrend_score,
+    tail_index_hill,
    vol_percentile_historical,
 )
 from .hypothesis import MarketSummary
@@ -40,6 +45,13 @@ def build_market_summary(
    season_h = seasonality_strength(returns, by="hour")
    season_d = seasonality_strength(returns, by="dow")

+    eff_ratio = efficiency_ratio_kaufman(ohlcv["close"], length=100)
+    tail_l = tail_index_hill(returns, side="left")
+    tail_r = tail_index_hill(returns, side="right")
+    hh_hl = structural_uptrend_score(ohlcv["close"], window=5)
+    compress = compression_ratio(ohlcv["close"], recent_window=50, ref_window=200)
+    spec_entropy, dom_cycle = spectral_entropy_and_cycle(returns, length=256)
+
    return MarketSummary(
        symbol=symbol,
        timeframe=timeframe,
@@ -55,4 +67,11 @@ def build_market_summary(
        vol_percentile=vol_pct,
        seasonality_hour=season_h,
        seasonality_dow=season_d,
+        efficiency_ratio=eff_ratio,
+        tail_index_left=tail_l,
+        tail_index_right=tail_r,
+        structural_uptrend=hh_hl,
+        compression=compress,
+        spectral_entropy=spec_entropy,
+        dominant_cycle=dom_cycle,
    )
@@ -2,6 +2,7 @@ from __future__ import annotations

 from dataclasses import dataclass

+import numpy as np
 import pandas as pd  # type: ignore[import-untyped]

 from .orders import Position, Side, Trade
@@ -28,74 +29,110 @@ class BacktestEngine:
        self.fees_bp = fees_bp

    def run(self, ohlcv: pd.DataFrame, signals: pd.Series) -> BacktestResult:
+        n = len(ohlcv)
+        if n == 0:
+            empty = pd.Series([], dtype=float)
+            return BacktestResult(equity_curve=empty, returns=empty, trades=[])
+
        signals = signals.reindex(ohlcv.index).ffill().fillna(Side.FLAT)
-
        # Esecuzione con delay 1: segnale a t-1 esegue a open di t.
-        shifted = [Side.FLAT, *list(signals.iloc[:-1])]
-        executed_side = pd.Series(shifted, index=ohlcv.index, dtype=object)
+        executed = pd.Series(
+            [Side.FLAT, *list(signals.iloc[:-1])],
+            index=ohlcv.index,
+            dtype=object,
+        )
+
+        # Codifica side in int per vectorizzazione: 0=FLAT, +1=LONG, -1=SHORT.
+        side_code = np.where(
+            executed.values == Side.LONG, 1,
+            np.where(executed.values == Side.SHORT, -1, 0),
+        ).astype(np.int8)
+        opens = ohlcv["open"].to_numpy(dtype=np.float64)
+        closes = ohlcv["close"].to_numpy(dtype=np.float64)
+        ts_index = ohlcv.index
+
+        # Identifica transizioni: punto in cui side[i] != side[i-1] (con side[-1]=0).
+        prev = np.concatenate(([0], side_code[:-1]))
+        change = side_code != prev
+
+        # Indici di entry (cambio verso side != 0).
+        entry_idxs = np.flatnonzero(change & (side_code != 0))
+        # Indici di chiusura: per ogni entry, il prossimo indice dove side[i] != side_entry.
+        # Vectorized: per ogni entry_idx, cerca change & side != side_entry oltre l'entry.

-        position: Position | None = None
-        position_entry_ts: pd.Timestamp | None = None
        trades: list[Trade] = []
-        equity = 0.0
-        equity_history: list[float] = []
-        returns_history: list[float] = []
-        prev_equity = 0.0
+        # realized_pnl[t]: PnL netto cumulato dopo le chiusure avvenute a OPEN di t.
+        realized_pnl = np.zeros(n, dtype=np.float64)
+        fees_rate = self.fees_bp / 10000.0
+        size = 1.0

-        for ts, row in ohlcv.iterrows():
-            target_side = executed_side.loc[ts]
-            current_side = position.side if position else Side.FLAT
+        # Posizione corrente all'inizio di ogni indice t (prima di applicare il transitorio):
+        # used per MtM computation. open_side_at_t / open_entry_at_t.
+        open_side = np.zeros(n, dtype=np.int8)
+        open_entry = np.zeros(n, dtype=np.float64)

-            if target_side != current_side:
-                if position is not None:
-                    assert position_entry_ts is not None
-                    trade = Trade(
-                        entry_ts=position_entry_ts,
-                        exit_ts=ts,
-                        side=position.side,
-                        size=position.size,
-                        entry_price=position.entry_price,
-                        exit_price=float(row["open"]),
-                        fees_bp=self.fees_bp,
-                    )
-                    trades.append(trade)
-                    equity += trade.net_pnl
-                    position = None
-                    position_entry_ts = None
-                if target_side in (Side.LONG, Side.SHORT):
-                    position = Position(
-                        side=target_side, entry_price=float(row["open"]), size=1.0
-                    )
-                    position_entry_ts = ts
+        for entry_idx in entry_idxs:
+            entry_side = int(side_code[entry_idx])
+            entry_price = opens[entry_idx]
+            # Cerca exit: primo indice > entry_idx dove side differisce.
+            after = side_code[entry_idx + 1:]
+            rel = np.flatnonzero(after != entry_side)
+            if rel.size > 0:
+                exit_idx = entry_idx + 1 + int(rel[0])
+                exit_price = opens[exit_idx]
+                exit_ts = ts_index[exit_idx]
+                gross = entry_side * (exit_price - entry_price) * size
+                fees = fees_rate * size * (entry_price + exit_price)
+                net = gross - fees
+                # La chiusura avviene a open[exit_idx]: dal bar exit_idx in poi il
+                # PnL e' realizzato (non piu' MtM).
+                realized_pnl[exit_idx:] += net
+                # Posizione aperta in [entry_idx, exit_idx-1].
+                open_side[entry_idx:exit_idx] = entry_side
+                open_entry[entry_idx:exit_idx] = entry_price
+                trades.append(Trade(
+                    entry_ts=ts_index[entry_idx],
+                    exit_ts=exit_ts,
+                    side=Side.LONG if entry_side == 1 else Side.SHORT,
+                    size=size,
+                    entry_price=entry_price,
+                    exit_price=exit_price,
+                    fees_bp=self.fees_bp,
+                ))
+            else:
+                # Ultima posizione ancora aperta: chiusura forced a close[-1].
+                # Parita' col loop legacy: MtM su [entry_idx, n-1), realized totale
+                # SOLO al bar n-1 (legacy fa equity_history[-1] = equity).
+                last_close = closes[-1]
+                gross = entry_side * (last_close - entry_price) * size
+                fees = fees_rate * size * (entry_price + last_close)
+                net = gross - fees
+                if entry_idx < n - 1:
+                    open_side[entry_idx:n - 1] = entry_side
+                    open_entry[entry_idx:n - 1] = entry_price
+                realized_pnl[-1] += net
+                trades.append(Trade(
+                    entry_ts=ts_index[entry_idx],
+                    exit_ts=ts_index[-1],
+                    side=Side.LONG if entry_side == 1 else Side.SHORT,
+                    size=size,
+                    entry_price=entry_price,
+                    exit_price=last_close,
+                    fees_bp=self.fees_bp,
+                ))

-            mark = float(row["close"])
-            mtm = position.unrealized_pnl(mark) if position else 0.0
-            current_equity = equity + mtm
-            equity_history.append(current_equity)
-            returns_history.append(current_equity - prev_equity)
-            prev_equity = current_equity
-
-        if position is not None:
-            assert position_entry_ts is not None
-            last_ts = ohlcv.index[-1]
-            last_close = float(ohlcv["close"].iloc[-1])
-            trade = Trade(
-                entry_ts=position_entry_ts,
-                exit_ts=last_ts,
-                side=position.side,
-                size=position.size,
-                entry_price=position.entry_price,
-                exit_price=last_close,
-                fees_bp=self.fees_bp,
-            )
-            trades.append(trade)
-            equity += trade.net_pnl
-            equity_history[-1] = equity
-            if len(returns_history) >= 2:
-                returns_history[-1] = equity - equity_history[-2]
+        # MtM unrealized per ogni bar in cui c'e' una posizione aperta.
+        mtm = open_side.astype(np.float64) * (closes - open_entry) * size
+        equity_arr = realized_pnl + mtm
+        # Returns = first diff dell'equity (col loop legacy il primo bar e' equity[0]-0).
+        returns_arr = np.concatenate(([equity_arr[0]], np.diff(equity_arr)))

        return BacktestResult(
-            equity_curve=pd.Series(equity_history, index=ohlcv.index, name="equity"),
-            returns=pd.Series(returns_history, index=ohlcv.index, name="returns"),
+            equity_curve=pd.Series(equity_arr, index=ts_index, name="equity"),
+            returns=pd.Series(returns_arr, index=ts_index, name="returns"),
            trades=trades,
        )
+
+
+# Lo facade Position re-export e' tenuto per backward-compat con import legacy.
+__all__ = ["BacktestEngine", "BacktestResult", "Position", "Side", "Signal", "Trade"]
@@ -3,34 +3,12 @@ from __future__ import annotations
 import random

 from ..genome.hypothesis import HypothesisAgentGenome, ModelTier
-from ..genome.mutation import COGNITIVE_STYLES
+from ..genome.prompt_library import PromptLibrary

-STYLE_PROMPTS: dict[str, str] = {
-    "physicist": (
-        "Cerca leggi conservative, simmetrie, regimi di scala. "
-        "Pensa in termini di flussi e potenziali."
-    ),
-    "biologist": (
-        "Cerca pattern adattivi, nicchie ecologiche, "
-        "predator-prey dynamics tra partecipanti del mercato."
-    ),
-    "historian": (
-        "Cerca pattern ricorrenti su scale temporali multiple, "
-        "analogie con regimi storici, mean reversion strutturali."
-    ),
-    "meteorologist": (
-        "Cerca regimi di volatilità che si autoalimentano, "
-        "transizioni di stato come fronti, persistenza locale."
-    ),
-    "ecologist": (
-        "Cerca interazioni multi-asset, correlazioni cluster, "
-        "segnali di stress sistemico nelle dinamiche di flusso."
-    ),
-    "engineer": (
-        "Cerca segnali con rapporto S/N favorevole, filtri causali, "
-        "robustezza a perturbazioni di calibrazione."
-    ),
-}
+# Mantenuto come alias backcompat: equivalente a PromptLibrary.default().styles.
+# Nuovi caller dovrebbero usare PromptLibrary direttamente per supportare
+# l'override via prompts.json di una strategia.
+STYLE_PROMPTS: dict[str, str] = PromptLibrary.default().styles


 def build_initial_population(
@@ -38,15 +16,22 @@ def build_initial_population(
    model_tier: ModelTier,
    rng: random.Random,
    feature_pool: tuple[str, ...] = ("close", "high", "low", "volume"),
+    prompt_library: PromptLibrary | None = None,
 ) -> list[HypothesisAgentGenome]:
-    """Costruisce una popolazione iniziale K varia per stile cognitivo + parametri."""
+    """Costruisce una popolazione iniziale K varia per stile cognitivo + parametri.
+
+    ``prompt_library`` controlla quali stili sono disponibili e quale system_prompt
+    iniziale viene assegnato. Default = builtin 6 stili (physicist, biologist, ...).
+    Override tipico: ``PromptLibrary.from_json(strategy_crypto/prompts.json)``.
+    """
+    lib = prompt_library or PromptLibrary.default()
    population: list[HypothesisAgentGenome] = []
    for i in range(k):
-        style = COGNITIVE_STYLES[i % len(COGNITIVE_STYLES)]
+        style = lib.style_at(i)
        n_features = rng.randint(1, len(feature_pool))
        feats = sorted(rng.sample(feature_pool, k=n_features))
        g = HypothesisAgentGenome(
-            system_prompt=STYLE_PROMPTS[style],
+            system_prompt=lib.directive(style),
            feature_access=feats,
            temperature=round(rng.uniform(0.7, 1.2), 2),
            top_p=0.95,
@@ -7,6 +7,10 @@ from .hypothesis import HypothesisAgentGenome

 FEATURE_POOL: tuple[str, ...] = ("open", "high", "low", "close", "volume")

+# Lista di default builtin (allineata con PromptLibrary.default()).
+# Il dispatcher run_phase1 sovrascrive `COGNITIVE_STYLES` con la lista letta
+# da prompts.json prima del loop GA, cosi' `mutate_cognitive_style` pesca
+# dai candidati corretti per la strategia in corso.
 COGNITIVE_STYLES: tuple[str, ...] = (
    "physicist",
    "biologist",
@@ -17,6 +21,18 @@ COGNITIVE_STYLES: tuple[str, ...] = (
 )


+def set_cognitive_styles(styles: tuple[str, ...]) -> None:
+    """Sovrascrive la lista globale di stili candidati per la mutation.
+
+    Da chiamare PRIMA del GA loop (es. in run_phase1 dopo aver caricato la
+    PromptLibrary). Non thread-safe: pensata per uno script CLI.
+    """
+    global COGNITIVE_STYLES
+    if not styles:
+        raise ValueError("set_cognitive_styles: lista vuota")
+    COGNITIVE_STYLES = tuple(styles)
+
+
 def _clone_with(g: HypothesisAgentGenome, **overrides: Any) -> HypothesisAgentGenome:
    payload: dict[str, Any] = g.to_dict()
    payload.update(overrides)
@@ -0,0 +1,234 @@
+"""Libreria di stili cognitivi + direttive system_prompt per il GA.
+
+Carica da un file JSON esterno (tipicamente shippato dal singolo strategy
+member, es. ``strategy_crypto/prompts.json``) le coppie ``style -> directive``
+usate da:
+  - :func:`multi_swarm_core.ga.initial.build_initial_population` per il
+    bootstrap della popolazione (style assegnato round-robin, directive
+    come system_prompt iniziale).
+  - :func:`multi_swarm_core.genome.mutation.mutate_cognitive_style` per
+    pescare i candidati di mutazione (range = key del JSON).
+
+Schema JSON atteso::
+
+  {
+    "styles": {
+      "<name>": {"directive": "<testo system_prompt>"},
+      ...
+    },
+    "agent_role": "<descrizione agente, opzionale>",
+    "pattern_guidance": "<sezione PATTERN GUIDANCE, opzionale>",
+    "instruction": "<frase finale USER, opzionale>",
+    "domain_warnings": "<warning di dominio, opzionale>"
+  }
+
+I 6 stili default (physicist, biologist, historian, meteorologist,
+ecologist, engineer) sono comunque disponibili via :meth:`PromptLibrary.default`
+per backcompat con test/script senza file esterno.
+"""
+
+from __future__ import annotations
+
+import json
+from dataclasses import dataclass, field
+from pathlib import Path
+
+_DEFAULT_STYLES: dict[str, str] = {
+    "physicist": (
+        "Cerca leggi conservative, simmetrie, regimi di scala. "
+        "Pensa in termini di flussi e potenziali."
+    ),
+    "biologist": (
+        "Cerca pattern adattivi, nicchie ecologiche, "
+        "predator-prey dynamics tra partecipanti del mercato."
+    ),
+    "historian": (
+        "Cerca pattern ricorrenti su scale temporali multiple, "
+        "analogie con regimi storici, mean reversion strutturali."
+    ),
+    "meteorologist": (
+        "Cerca regimi di volatilità che si autoalimentano, "
+        "transizioni di stato come fronti, persistenza locale."
+    ),
+    "ecologist": (
+        "Cerca interazioni multi-asset, correlazioni cluster, "
+        "segnali di stress sistemico nelle dinamiche di flusso."
+    ),
+    "engineer": (
+        "Cerca segnali con rapporto S/N favorevole, filtri causali, "
+        "robustezza a perturbazioni di calibrazione."
+    ),
+}
+
+
+class PromptLibraryError(ValueError):
+    """Sollevata su JSON malformato o stili invalid."""
+
+
+_DEFAULT_PATTERN_GUIDANCE = (
+    "Forme di curva: trend (SMA cross), compressione volatilita (atr_pct basso), "
+    "espansione volatilita (atr_pct alto), mean reversion (sma_pct estremo), "
+    "momentum (macd_pct, rsi zone).\n\n"
+    "  Ripetibilita dell'andamento: crossover/crossunder ricorrenti, pattern intra-day "
+    "(hour gate), pattern settimanali (dow/is_weekend), range breakout."
+)
+
+_DEFAULT_AGENT_ROLE = (
+    "Sei un agente generatore di ipotesi di trading quantitativo per un sistema swarm."
+)
+
+_DEFAULT_INSTRUCTION = (
+    "Genera una strategia che cerchi anomalie sfruttabili in questo regime."
+)
+
+
+@dataclass(frozen=True)
+class PromptLibrary:
+    """Set immutabile di stili cognitivi + direttive system_prompt.
+
+    v3.0: aggiunge 4 campi top-level strategy-specific iniettabili nel prompt
+    dal compositor ``_build_system_prompt``. Tutti opzionali con default sensati
+    per backcompat.
+    v3.1: aggiunge anti_patterns e output_priorities iniettati dopo i VINCOLI core.
+    """
+
+    styles: dict[str, str]
+    focus: dict[str, list[str]]  # style -> lista metriche prioritarie
+    # NEW v3.0: contenuto strategy-specific iniettabile nel prompt
+    agent_role: str = field(default="")       # header SYSTEM, descrive chi e' l'agente
+    pattern_guidance: str = field(default="") # sezione PATTERN GUIDANCE del SYSTEM
+    instruction: str = field(default="")      # frase finale USER ("Genera una strategia...")
+    domain_warnings: str = field(default="")  # opzionale: warning di dominio (es. crypto 24/7)
+    anti_patterns: str = field(default="")       # NEW v3.1: lista esplicita di pattern da evitare
+    output_priorities: str = field(default="")   # NEW v3.1: trade-off espliciti di output
+
+    def __post_init__(self) -> None:
+        if not self.styles:
+            raise PromptLibraryError("PromptLibrary deve avere almeno uno stile")
+        for name, directive in self.styles.items():
+            if not isinstance(name, str) or not name.strip():
+                raise PromptLibraryError(f"nome stile invalido: {name!r}")
+            if not isinstance(directive, str) or not directive.strip():
+                raise PromptLibraryError(
+                    f"directive vuota o invalida per stile {name!r}"
+                )
+        # Validate new optional string fields: if provided must be non-whitespace
+        for field_name, value in (
+            ("agent_role", self.agent_role),
+            ("pattern_guidance", self.pattern_guidance),
+            ("instruction", self.instruction),
+            ("domain_warnings", self.domain_warnings),
+            ("anti_patterns", self.anti_patterns),
+            ("output_priorities", self.output_priorities),
+        ):
+            if not isinstance(value, str):
+                raise PromptLibraryError(
+                    f"campo '{field_name}' deve essere stringa, non {type(value)}"
+                )
+            if value and not value.strip():
+                raise PromptLibraryError(
+                    f"campo '{field_name}' fornito ma contiene solo whitespace"
+                )
+
+    @classmethod
+    def default(cls) -> PromptLibrary:
+        """Libreria builtin con i 6 stili originali (fallback senza file)."""
+        return cls(
+            styles=dict(_DEFAULT_STYLES),
+            focus={},
+            agent_role=_DEFAULT_AGENT_ROLE,
+            pattern_guidance=_DEFAULT_PATTERN_GUIDANCE,
+            instruction=_DEFAULT_INSTRUCTION,
+            domain_warnings="",
+            anti_patterns="",
+            output_priorities="",
+        )
+
+    @classmethod
+    def from_json(cls, path: Path | str) -> PromptLibrary:
+        """Carica da file JSON.
+
+        Schema::
+
+          {
+            "styles": {<name>: {"directive": "...", "focus_metrics": [...]}},
+            "agent_role": "...",          // opzionale
+            "pattern_guidance": "...",    // opzionale
+            "instruction": "...",         // opzionale
+            "domain_warnings": "..."      // opzionale
+          }
+        """
+        p = Path(path)
+        try:
+            data = json.loads(p.read_text(encoding="utf-8"))
+        except FileNotFoundError as e:
+            raise PromptLibraryError(f"file non trovato: {p}") from e
+        except json.JSONDecodeError as e:
+            raise PromptLibraryError(f"JSON malformato in {p}: {e}") from e
+
+        if not isinstance(data, dict):
+            raise PromptLibraryError(f"root JSON deve essere un object, non {type(data)}")
+        styles_raw = data.get("styles")
+        if not isinstance(styles_raw, dict):
+            raise PromptLibraryError("manca chiave 'styles' (object) nel JSON")
+
+        styles: dict[str, str] = {}
+        focus: dict[str, list[str]] = {}
+        for name, entry in styles_raw.items():
+            if not isinstance(entry, dict):
+                raise PromptLibraryError(
+                    f"entry per stile {name!r} deve essere object, non {type(entry)}"
+                )
+            directive = entry.get("directive")
+            if not isinstance(directive, str):
+                raise PromptLibraryError(
+                    f"manca 'directive' (string) per stile {name!r}"
+                )
+            styles[name] = directive
+            fm = entry.get("focus_metrics", [])
+            if not isinstance(fm, list):
+                raise PromptLibraryError(
+                    f"focus_metrics di {name!r} deve essere lista, non {type(fm)}"
+                )
+            focus[name] = [str(x) for x in fm]
+
+        # Parse new optional top-level fields (v3.0)
+        agent_role = data.get("agent_role", "")
+        pattern_guidance = data.get("pattern_guidance", "")
+        instruction = data.get("instruction", "")
+        domain_warnings = data.get("domain_warnings", "")
+        # Parse new optional top-level fields (v3.1)
+        anti_patterns_raw = data.get("anti_patterns", "")
+        output_priorities_raw = data.get("output_priorities", "")
+        if not isinstance(anti_patterns_raw, str):
+            raise PromptLibraryError(f"anti_patterns deve essere stringa, non {type(anti_patterns_raw)}")
+        if not isinstance(output_priorities_raw, str):
+            raise PromptLibraryError(f"output_priorities deve essere stringa, non {type(output_priorities_raw)}")
+
+        return cls(
+            styles=styles,
+            focus=focus,
+            agent_role=agent_role,
+            pattern_guidance=pattern_guidance,
+            instruction=instruction,
+            domain_warnings=domain_warnings,
+            anti_patterns=anti_patterns_raw,
+            output_priorities=output_priorities_raw,
+        )
+
+    @property
+    def cognitive_styles(self) -> tuple[str, ...]:
+        """Tupla immutabile dei nomi degli stili, nell'ordine di insertion del JSON."""
+        return tuple(self.styles)
+
+    def directive(self, style: str) -> str:
+        """Ritorna la directive di ``style`` o solleva KeyError."""
+        return self.styles[style]
+
+    def style_at(self, index: int) -> str:
+        """Round-robin: ``cognitive_styles[index % len()]``."""
+        return self.cognitive_styles[index % len(self.cognitive_styles)]
+
+    def focus_metrics_for(self, style: str) -> list[str]:
+        """Ritorna lista delle metriche prioritarie per ``style``. Empty list se non specificate."""
+        return self.focus.get(style, [])
@@ -141,3 +141,135 @@ def total_return(equity: pd.Series) -> float:
    if equity.iloc[0] == 0:
        return float(equity.iloc[-1])
    return float(equity.iloc[-1] / equity.iloc[0] - 1.0)
+
+
+def efficiency_ratio_kaufman(prices: pd.Series, length: int = 100) -> float:
+    """Kaufman Efficiency Ratio: |net move| / sum(|step|) su rolling length.
+
+    Output 0-1: 0 = puro noise (movimento dissipativo), 1 = puro trend efficiente.
+    Discriminatore trending/ranging robusto.
+    """
+    if len(prices) < length + 1:
+        return 0.0
+    recent = prices.iloc[-length:]
+    net_move = abs(recent.iloc[-1] - recent.iloc[0])
+    total_move = recent.diff().abs().sum()
+    if total_move == 0 or pd.isna(total_move):
+        return 0.0
+    return float(net_move / total_move)
+
+
+def tail_index_hill(returns: pd.Series, side: str, top_frac: float = 0.05) -> float:
+    """Hill estimator del tail index (pendenza coda) per side in {'left', 'right'}.
+
+    Output: indice di pesantezza coda. Valori piu bassi = coda piu pesante.
+    <2 = varianza infinita (Cauchy-like), 3-4 = tipico crypto, >5 = quasi-Gaussiana.
+    Robusto al singolo outlier (vs kurtosis).
+    """
+    r = returns.dropna()
+    n = len(r)
+    if n < 50:
+        return 5.0  # default quasi-Gaussiano
+    k = max(int(n * top_frac), 5)
+    if side == "left":
+        tail = (-r).nlargest(k)
+    elif side == "right":
+        tail = r.nlargest(k)
+    else:
+        raise ValueError(f"side deve essere 'left' o 'right', non {side!r}")
+    # Filtra valori non-positivi (Hill richiede tail positiva)
+    tail = tail[tail > 0]
+    if len(tail) < 5:
+        return 5.0
+    log_tail = np.log(tail.values)
+    threshold = log_tail[-1]
+    excess = log_tail[:-1] - threshold
+    if excess.sum() <= 0:
+        return 5.0
+    hill = (len(excess)) / excess.sum()
+    return float(np.clip(hill, 1.0, 10.0))
+
+
+def structural_uptrend_score(prices: pd.Series, window: int = 5) -> float:
+    """Frazione di periodi in struttura HH/HL (Dow-style uptrend).
+
+    Identifica swing high/low usando rolling max/min con finestra ``window``.
+    Conta sequenze higher-high + higher-low (uptrend) vs lower-high + lower-low (downtrend).
+    Output: 0 (puro downtrend) - 0.5 (range/mixed) - 1 (puro uptrend).
+    """
+    if len(prices) < 4 * window:
+        return 0.5
+    swing_high = prices.rolling(2 * window + 1, center=True).max() == prices
+    swing_low = prices.rolling(2 * window + 1, center=True).min() == prices
+
+    highs = prices[swing_high].dropna()
+    lows = prices[swing_low].dropna()
+
+    if len(highs) < 3 or len(lows) < 3:
+        return 0.5
+
+    hh = (highs.diff() > 0).sum()
+    lh = (highs.diff() < 0).sum()
+    hl = (lows.diff() > 0).sum()
+    ll = (lows.diff() < 0).sum()
+
+    up_signals = hh + hl
+    down_signals = lh + ll
+    total = up_signals + down_signals
+    if total == 0:
+        return 0.5
+    return float(up_signals / total)
+
+
+def compression_ratio(prices: pd.Series, recent_window: int = 50, ref_window: int = 200) -> float:
+    """Range(recent) / Range(ref). <1 = compressione vol, >1 = espansione.
+
+    Range = high - low della finestra. Cattura il "coil" pre-breakout.
+    """
+    if len(prices) < ref_window:
+        return 1.0
+    recent = prices.iloc[-recent_window:]
+    ref = prices.iloc[-ref_window:]
+    recent_range = float(recent.max() - recent.min())
+    ref_range = float(ref.max() - ref.min())
+    if ref_range <= 0:
+        return 1.0
+    return recent_range / ref_range
+
+
+def spectral_entropy_and_cycle(
+    returns: pd.Series, length: int = 256
+) -> tuple[float, float | None]:
+    """FFT su returns -> (entropy normalizzata, dominant_cycle gated).
+
+    entropy: 0-1, Shannon entropy normalizzata dello spettro di potenza.
+      0 = una sola frequenza domina, 1 = spettro piatto (rumore bianco).
+    dominant_cycle: periodo (barre) della frequenza dominante,
+      MA solo se entropy < 0.6 (struttura presente). Altrimenti None.
+    """
+    r = returns.dropna()
+    if len(r) < length:
+        return 1.0, None
+    series = r.iloc[-length:].values
+    # Detrend
+    series = series - series.mean()
+    fft = np.fft.rfft(series)
+    power = np.abs(fft) ** 2
+    if power.sum() <= 0:
+        return 1.0, None
+    p = power / power.sum()
+    # Entropy normalizzata (Shannon / log(N))
+    nonzero = p[p > 0]
+    entropy = float(-(nonzero * np.log(nonzero)).sum() / np.log(len(nonzero)))
+    entropy = max(0.0, min(1.0, entropy))
+
+    if entropy >= 0.6:
+        return entropy, None
+    # Skip DC component (index 0)
+    if len(power) <= 1:
+        return entropy, None
+    peak_idx = int(np.argmax(power[1:])) + 1
+    if peak_idx == 0:
+        return entropy, None
+    cycle = float(length / peak_idx)
+    return entropy, cycle
@@ -13,6 +13,7 @@ possa leggere lo stato a run terminato (o in corso).
 from __future__ import annotations

 import random
+from concurrent.futures import ThreadPoolExecutor
 from dataclasses import dataclass, field
 from pathlib import Path

@@ -20,13 +21,15 @@ import pandas as pd  # type: ignore[import-untyped]

 from ..agents.adversarial import AdversarialAgent
 from ..agents.falsification import FalsificationAgent
-from ..agents.hypothesis import HypothesisAgent
+from ..agents.hypothesis import HypothesisAgent, HypothesisProposal, MarketSummary
 from ..agents.market_summary import build_market_summary
 from ..ga.fitness import compute_fitness
 from ..ga.initial import build_initial_population
 from ..ga.loop import GAConfig, next_generation
 from ..ga.summary import generation_summary
-from ..genome.hypothesis import ModelTier
+from ..genome.hypothesis import HypothesisAgentGenome, ModelTier
+from ..genome.mutation import set_cognitive_styles
+from ..genome.prompt_library import PromptLibrary
 from ..llm.client import LLMClient
 from ..llm.cost_tracker import CostTracker
 from ..persistence.repository import Repository
@@ -67,6 +70,33 @@ class RunConfig:
    # 2x costo backtest engine.
    eval_oos_during_loop: bool = False
    fitness_combined_alpha: float = 0.5  # peso IS (1-alpha = OOS)
+    # Libreria di stili cognitivi + system_prompt iniziali. Se None usa
+    # i 6 builtin (PromptLibrary.default()). Tipicamente caricata da
+    # strategy_crypto/prompts.json via PromptLibrary.from_json().
+    prompt_library: PromptLibrary | None = None
+    # Numero di propose() LLM concorrenti per generazione. 1 = sequenziale (default,
+    # backward compat). 6-10 tipicamente accettati da OpenRouter qwen-2.5 senza
+    # rate-limit. Riduce wall time GA loop di 5-8x su tier C.
+    llm_concurrency: int = 1
+
+
+def _parallel_propose(
+    agent: HypothesisAgent,
+    genomes: list[HypothesisAgentGenome],
+    market: MarketSummary,
+    n_workers: int,
+) -> list[HypothesisProposal]:
+    """Esegue ``agent.propose()`` su una lista di genomi, opzionalmente in parallelo.
+
+    ``n_workers <= 1`` mantiene il comportamento serial originale (ordine fisso,
+    determinismo data un seed). ``n_workers > 1`` usa un thread pool: l'order
+    dei risultati e' preservato (1:1 con ``genomes``). OpenAI/openrouter client
+    e' thread-safe; ``PromptLibrary`` e ``HypothesisAgent`` non hanno stato mutabile.
+    """
+    if n_workers <= 1 or len(genomes) <= 1:
+        return [agent.propose(g, market) for g in genomes]
+    with ThreadPoolExecutor(max_workers=n_workers) as pool:
+        return list(pool.map(lambda g: agent.propose(g, market), genomes))


 def run_phase1(
@@ -82,10 +112,16 @@ def run_phase1(

    repo = Repository(cfg.db_path)
    repo.init_schema()
+    # Escludi prompt_library (PromptLibrary dataclass non e' JSON-serializable);
+    # salva solo i nomi degli stili per reproducibility.
    config_dict = {
-        **cfg.__dict__,
+        **{k: v for k, v in cfg.__dict__.items() if k != "prompt_library"},
        "db_path": str(cfg.db_path),
        "model_tier": cfg.model_tier.value,
+        "prompt_library_styles": (
+            list(cfg.prompt_library.cognitive_styles)
+            if cfg.prompt_library is not None else None
+        ),
    }
    run_id = repo.create_run(name=cfg.run_name, config=config_dict)

@@ -100,7 +136,13 @@ def run_phase1(

    market = build_market_summary(train_ohlcv, symbol=cfg.symbol, timeframe=cfg.timeframe)

-    hypothesis_agent = HypothesisAgent(llm=llm)
+    # Propaga la libreria di stili al modulo mutation (cosi' mutate_cognitive_style
+    # pesca dai candidati coerenti col JSON della strategia in corso). Va FATTO
+    # PRIMA di istanziare HypothesisAgent (che la riceve in costruttore).
+    prompt_library = cfg.prompt_library or PromptLibrary.default()
+    set_cognitive_styles(prompt_library.cognitive_styles)
+
+    hypothesis_agent = HypothesisAgent(llm=llm, prompt_library=prompt_library)
    falsification_agent = FalsificationAgent(
        fees_bp=cfg.fees_bp, n_trials_dsr=cfg.n_trials_dsr
    )
@@ -113,7 +155,10 @@ def run_phase1(
    cost_tracker = CostTracker()

    population = build_initial_population(
-        k=cfg.population_size, model_tier=cfg.model_tier, rng=rng
+        k=cfg.population_size,
+        model_tier=cfg.model_tier,
+        rng=rng,
+        prompt_library=prompt_library,
    )
    fitnesses: dict[str, float] = {}

@@ -127,11 +172,20 @@ def run_phase1(

    try:
        for gen in range(cfg.n_generations):
+            # Step 1: raccogli i genomi da valutare in questa generazione (esclude
+            # elite gia' presenti nella cache fitnesses) e lancia propose() in
+            # parallelo. La sezione DB-write resta serial sotto.
+            uncached = [g for g in population if g.id not in fitnesses]
+            proposals = _parallel_propose(
+                hypothesis_agent, uncached, market, cfg.llm_concurrency
+            )
+            proposal_by_id = {g.id: p for g, p in zip(uncached, proposals, strict=True)}
+
            for genome in population:
                if genome.id in fitnesses:
                    continue  # elite gia' valutata in generazione precedente
                repo.save_genome(run_id=run_id, generation_idx=gen, genome=genome)
-                proposal = hypothesis_agent.propose(genome, market)
+                proposal = proposal_by_id[genome.id]
                # Registra costo per OGNI completion (incluse retry).
                for completion in proposal.completions:
                    cost_record = cost_tracker.record(
@@ -205,7 +259,7 @@ def run_phase1(
                            cfg.fitness_combined_alpha * fit
                            + (1.0 - cfg.fitness_combined_alpha) * fit_oos_inloop
                        )
-                    except Exception:  # noqa: BLE001
+                    except Exception:
                        pass  # fallback: usa solo IS
                repo.save_evaluation(
                    run_id=run_id,
@@ -246,7 +300,7 @@ def run_phase1(
        # WFA re-eval: i top_k genomi (by fitness in-sample > 0) vengono rivalutati
        # sul test_ohlcv. Le metriche OOS finiscono in evaluations.fitness_oos etc.
        if test_ohlcv is not None and len(test_ohlcv) >= 100:
-            from ..agents.hypothesis import _try_parse  # noqa: PLC0415
+            from ..agents.hypothesis import _try_parse

            all_evals = repo.list_evaluations(run_id)
            top_evals = sorted(
@@ -261,7 +315,7 @@ def run_phase1(
                try:
                    fals_oos = falsification_agent.evaluate(strategy, test_ohlcv)
                    adv_oos = adversarial_agent.review(strategy, test_ohlcv)
-                except Exception:  # noqa: BLE001
+                except Exception:
                    continue
                fit_oos = compute_fitness(
                    fals_oos, adv_oos,
@@ -108,7 +108,13 @@ def test_e2e_wfa_populates_fitness_oos(
    fake_llm,
    mocker,
 ):
-    """WFA: train_split=0.7 → top genomi devono avere fitness_oos popolato."""
+    """WFA: train_split=0.7 → top genomi devono avere fitness_oos popolato.
+
+    Usa fitness v2 con hard-kill minimale (solo no_trades): il fixture sintetico
+    non produce strategie profittevoli, quindi i check aggressivi
+    fees_eat_alpha/negative_net_pnl azzererebbero tutti i genomi rendendo
+    inverificabile il wiring WFA.
+    """
    cfg = RunConfig(
        run_name="e2e-wfa-test",
        population_size=5,
@@ -125,6 +131,7 @@ def test_e2e_wfa_populates_fitness_oos(
        db_path=tmp_path / "runs.db",
        wfa_train_split=0.7,
        wfa_top_k=3,
+        fitness_hard_kill_findings=("no_trades",),
    )
    run_id = run_phase1(cfg, ohlcv=synthetic_ohlcv, llm=fake_llm)
    repo = Repository(db_path=tmp_path / "runs.db")
@@ -3,7 +3,6 @@ import json
 import numpy as np
 import pandas as pd
 import pytest
-
 from multi_swarm_core.agents.adversarial import (
    AdversarialAgent,
    AdversarialReport,
@@ -54,7 +53,10 @@ def test_degenerate_always_long_flagged(ohlcv: pd.DataFrame) -> None:
    assert any(f.name == "degenerate" and f.severity == Severity.HIGH for f in report.findings)


-def test_no_findings_on_reasonable_strategy(ohlcv: pd.DataFrame) -> None:
+def test_rsi_mean_reversion_loses_money_on_synthetic_data(ohlcv: pd.DataFrame) -> None:
+    """RSI mean-reversion sul fixture sintetico ha net negativo: deve firare
+    negative_net_pnl (deal-breaker). Conferma che il check cattura strategie
+    che perdono sul training, indipendentemente dal motivo (no edge / fees)."""
    src = json.dumps(
        {
            "rules": [
@@ -84,8 +86,59 @@ def test_no_findings_on_reasonable_strategy(ohlcv: pd.DataFrame) -> None:
    ast = parse_strategy(src)
    agent = AdversarialAgent()
    report = agent.review(ast, ohlcv)
+    assert any(
+        f.name == "negative_net_pnl" and f.severity == Severity.HIGH
+        for f in report.findings
+    )
+
+
+def test_profitable_strategy_no_high_findings(
+    monkeypatch: pytest.MonkeyPatch, ohlcv: pd.DataFrame
+) -> None:
+    """Sanity test: una strategia con gross > 0 e fees << gross + n_trades ragionevole
+    + signal misto non deve produrre nessun finding HIGH."""
+    n = 15
+    # entry=100 exit=110 gross=10 per trade, fees a 5bp -> 0.105 per trade
+    # totali: gross=150, fees=1.575 -> net=+148.4
+    fake_trades = [
+        _make_trade(
+            ohlcv.index[i * 30],
+            ohlcv.index[i * 30 + 1],
+            entry_price=100.0,
+            exit_price=110.0,
+        )
+        for i in range(n)
+    ]
+    # 50/50 LONG/FLAT per evitare degenerate/flat_too_long/time_in_market.
+    fake_signals = pd.Series(
+        [Side.LONG if i % 2 == 0 else Side.FLAT for i in range(len(ohlcv))],
+        index=ohlcv.index,
+        dtype=object,
+    )
+
+    def fake_run(self, ohlcv: pd.DataFrame, signals: pd.Series) -> BacktestResult:  # type: ignore[no-untyped-def]
+        return BacktestResult(
+            equity_curve=pd.Series([0.0] * len(ohlcv), index=ohlcv.index, name="equity"),
+            returns=pd.Series([0.0] * len(ohlcv), index=ohlcv.index, name="returns"),
+            trades=fake_trades,
+        )
+
+    def fake_compile(strategy):  # type: ignore[no-untyped-def]
+        return lambda df: fake_signals
+
+    monkeypatch.setattr(
+        "multi_swarm_core.agents.adversarial.BacktestEngine.run", fake_run
+    )
+    monkeypatch.setattr(
+        "multi_swarm_core.agents.adversarial.compile_strategy", fake_compile
+    )
+
+    ast = parse_strategy(_MINIMAL_STRATEGY_SRC)
+    report = AdversarialAgent().review(ast, ohlcv)
    high_findings = [f for f in report.findings if f.severity == Severity.HIGH]
-    assert len(high_findings) == 0
+    assert high_findings == [], (
+        f"expected no HIGH findings, got: {[f.name for f in high_findings]}"
+    )


 def test_zero_trade_strategy_flagged(ohlcv: pd.DataFrame) -> None:
@@ -383,6 +436,55 @@ def test_fees_eat_alpha_flagged(monkeypatch: pytest.MonkeyPatch,
    )


+def test_negative_net_pnl_fires_on_negative_gross(
+    monkeypatch: pytest.MonkeyPatch, ohlcv: pd.DataFrame
+) -> None:
+    """gross_pnl < 0 (perdente direzionale) -> HIGH negative_net_pnl.
+    fees_eat_alpha NON deve firare perche' la sua condizione richiede gross > 0.
+    """
+    n = 15
+    # entry=100 exit=95 gross=-5 per trade (LONG perdente)
+    fake_trades = [
+        _make_trade(
+            ohlcv.index[i * 30],
+            ohlcv.index[i * 30 + 1],
+            entry_price=100.0,
+            exit_price=95.0,
+        )
+        for i in range(n)
+    ]
+    fake_signals = pd.Series(
+        [Side.LONG if i % 2 == 0 else Side.FLAT for i in range(len(ohlcv))],
+        index=ohlcv.index,
+        dtype=object,
+    )
+
+    def fake_run(self, ohlcv, signals):  # type: ignore[no-untyped-def]
+        return BacktestResult(
+            equity_curve=pd.Series([0.0] * len(ohlcv), index=ohlcv.index, name="equity"),
+            returns=pd.Series([0.0] * len(ohlcv), index=ohlcv.index, name="returns"),
+            trades=fake_trades,
+        )
+
+    def fake_compile(strategy):  # type: ignore[no-untyped-def]
+        return lambda df: fake_signals
+
+    monkeypatch.setattr(
+        "multi_swarm_core.agents.adversarial.BacktestEngine.run", fake_run
+    )
+    monkeypatch.setattr(
+        "multi_swarm_core.agents.adversarial.compile_strategy", fake_compile
+    )
+
+    ast = parse_strategy(_MINIMAL_STRATEGY_SRC)
+    report = AdversarialAgent().review(ast, ohlcv)
+    assert any(
+        f.name == "negative_net_pnl" and f.severity == Severity.HIGH
+        for f in report.findings
+    )
+    assert not any(f.name == "fees_eat_alpha" for f in report.findings)
+
+
 def test_time_in_market_too_high_flagged(monkeypatch: pytest.MonkeyPatch,
                                          ohlcv: pd.DataFrame) -> None:
    """Signal LONG per >80% delle bar -> HIGH time_in_market_too_high."""
@@ -0,0 +1,160 @@
+"""Parity check: engine vettorializzato vs reference iterrows implementation.
+
+Mantiene una copia inline del loop ``iterrows`` come reference per garantire
+che la vettorizzazione produca esattamente gli stessi trades, equity_curve e
+returns su input pseudocasuali, indipendentemente dal regime di prezzo.
+"""
+
+from __future__ import annotations
+
+import numpy as np
+import pandas as pd
+import pytest
+from multi_swarm_core.backtest.engine import BacktestEngine, BacktestResult
+from multi_swarm_core.backtest.orders import Position, Side, Trade
+
+
+def _legacy_run(
+    ohlcv: pd.DataFrame, signals: pd.Series, fees_bp: float = 5.0
+) -> BacktestResult:
+    """Reference implementation: il loop iterrows originale (pre-vectorize).
+
+    Mantenuto qui esclusivamente come oracolo per i test di parità.
+    """
+    signals = signals.reindex(ohlcv.index).ffill().fillna(Side.FLAT)
+    shifted = [Side.FLAT, *list(signals.iloc[:-1])]
+    executed_side = pd.Series(shifted, index=ohlcv.index, dtype=object)
+
+    position: Position | None = None
+    position_entry_ts: pd.Timestamp | None = None
+    trades: list[Trade] = []
+    equity = 0.0
+    equity_history: list[float] = []
+    returns_history: list[float] = []
+    prev_equity = 0.0
+
+    for ts, row in ohlcv.iterrows():
+        target_side = executed_side.loc[ts]
+        current_side = position.side if position else Side.FLAT
+        if target_side != current_side:
+            if position is not None:
+                assert position_entry_ts is not None
+                trade = Trade(
+                    entry_ts=position_entry_ts,
+                    exit_ts=ts,
+                    side=position.side,
+                    size=position.size,
+                    entry_price=position.entry_price,
+                    exit_price=float(row["open"]),
+                    fees_bp=fees_bp,
+                )
+                trades.append(trade)
+                equity += trade.net_pnl
+                position = None
+                position_entry_ts = None
+            if target_side in (Side.LONG, Side.SHORT):
+                position = Position(
+                    side=target_side, entry_price=float(row["open"]), size=1.0
+                )
+                position_entry_ts = ts
+        mark = float(row["close"])
+        mtm = position.unrealized_pnl(mark) if position else 0.0
+        current_equity = equity + mtm
+        equity_history.append(current_equity)
+        returns_history.append(current_equity - prev_equity)
+        prev_equity = current_equity
+    if position is not None:
+        assert position_entry_ts is not None
+        last_ts = ohlcv.index[-1]
+        last_close = float(ohlcv["close"].iloc[-1])
+        trade = Trade(
+            entry_ts=position_entry_ts,
+            exit_ts=last_ts,
+            side=position.side,
+            size=position.size,
+            entry_price=position.entry_price,
+            exit_price=last_close,
+            fees_bp=fees_bp,
+        )
+        trades.append(trade)
+        equity += trade.net_pnl
+        equity_history[-1] = equity
+        if len(returns_history) >= 2:
+            returns_history[-1] = equity - equity_history[-2]
+
+    return BacktestResult(
+        equity_curve=pd.Series(equity_history, index=ohlcv.index, name="equity"),
+        returns=pd.Series(returns_history, index=ohlcv.index, name="returns"),
+        trades=trades,
+    )
+
+
+def _random_ohlcv(n: int, seed: int) -> pd.DataFrame:
+    rng = np.random.default_rng(seed)
+    rets = rng.normal(0.0, 0.01, size=n)
+    close = 100.0 * np.exp(np.cumsum(rets))
+    idx = pd.date_range("2024-01-01", periods=n, freq="1h", tz="UTC")
+    return pd.DataFrame(
+        {
+            "open": close * (1 + rng.normal(0, 0.001, n)),
+            "high": close * 1.005,
+            "low": close * 0.995,
+            "close": close,
+            "volume": rng.uniform(1.0, 100.0, n),
+        },
+        index=idx,
+    )
+
+
+def _random_signals(n: int, seed: int, p_change: float = 0.1) -> pd.Series:
+    """Segnali con persistenza: ad ogni bar con prob p_change cambia stato."""
+    rng = np.random.default_rng(seed + 999)
+    states = [Side.LONG, Side.SHORT, Side.FLAT]
+    out: list[Side] = [rng.choice(states)]
+    for _ in range(1, n):
+        out.append(rng.choice(states) if rng.random() < p_change else out[-1])
+    idx = pd.date_range("2024-01-01", periods=n, freq="1h", tz="UTC")
+    return pd.Series(out, index=idx, dtype=object)
+
+
+@pytest.mark.parametrize("seed", [0, 1, 42, 123, 999])
+def test_vectorized_equals_legacy(seed: int) -> None:
+    df = _random_ohlcv(500, seed)
+    signals = _random_signals(500, seed)
+    engine = BacktestEngine(fees_bp=5.0)
+    new = engine.run(df, signals)
+    ref = _legacy_run(df, signals, fees_bp=5.0)
+
+    pd.testing.assert_series_equal(
+        new.equity_curve, ref.equity_curve, rtol=1e-9, atol=1e-9
+    )
+    pd.testing.assert_series_equal(
+        new.returns, ref.returns, rtol=1e-9, atol=1e-9
+    )
+    assert len(new.trades) == len(ref.trades)
+    for nt, rt in zip(new.trades, ref.trades, strict=True):
+        assert nt.entry_ts == rt.entry_ts
+        assert nt.exit_ts == rt.exit_ts
+        assert nt.side == rt.side
+        assert nt.entry_price == pytest.approx(rt.entry_price, abs=1e-12)
+        assert nt.exit_price == pytest.approx(rt.exit_price, abs=1e-12)
+        assert nt.net_pnl == pytest.approx(rt.net_pnl, abs=1e-12)
+
+
+def test_vectorized_handles_position_still_open_at_end() -> None:
+    """Edge case: signal LONG fino all'ultimo bar (exit a close[-1] forced)."""
+    df = _random_ohlcv(100, seed=7)
+    signals = pd.Series([Side.LONG] * 100, index=df.index)
+    new = BacktestEngine(fees_bp=10.0).run(df, signals)
+    ref = _legacy_run(df, signals, fees_bp=10.0)
+    pd.testing.assert_series_equal(new.equity_curve, ref.equity_curve, atol=1e-9)
+    assert len(new.trades) == 1
+    assert new.trades[0].side == Side.LONG
+
+
+def test_vectorized_zero_signals_no_trades() -> None:
+    df = _random_ohlcv(50, seed=3)
+    signals = pd.Series([Side.FLAT] * 50, index=df.index)
+    new = BacktestEngine().run(df, signals)
+    assert len(new.trades) == 0
+    assert (new.equity_curve == 0).all()
@@ -2,6 +2,7 @@ import json

 from multi_swarm_core.agents.hypothesis import HypothesisAgent, MarketSummary
 from multi_swarm_core.genome.hypothesis import HypothesisAgentGenome, ModelTier
+from multi_swarm_core.genome.prompt_library import PromptLibrary
 from multi_swarm_core.llm.client import CompletionResult, EmptyCompletionError


@@ -248,3 +249,174 @@ def test_hypothesis_agent_returns_failed_proposal_on_only_empty_completions(mock
    assert "empty_completion" in proposal.parse_error
    # 3 tentativi tutti falliti.
    assert fake_llm.complete.call_count == 3
+
+
+def test_propose_renders_focus_block_for_style(mocker):  # type: ignore[no-untyped-def]
+    """Con PromptLibrary che ha focus_metrics, il LLM mock riceve 'Focus per la tua lente' nel user message."""
+    fake_llm = mocker.MagicMock()
+    fake_llm.complete.return_value = CompletionResult(
+        text=VALID_STRATEGY_JSON,
+        input_tokens=200,
+        output_tokens=80,
+        tier=ModelTier.C,
+        model="qwen",
+    )
+    # PromptLibrary con focus_metrics per physicist
+    lib = PromptLibrary(
+        styles={"physicist": "Pensa come un fisico."},
+        focus={"physicist": ["efficiency_ratio", "spectral_entropy", "hurst"]},
+    )
+    agent = HypothesisAgent(llm=fake_llm, prompt_library=lib)
+    proposal = agent.propose(make_genome(), make_summary())
+    assert proposal.strategy is not None
+    # Verifica che il user message contenga il focus block
+    call_kwargs = fake_llm.complete.call_args.kwargs
+    user_msg = call_kwargs["user"]
+    assert "Focus per la tua lente:" in user_msg
+    assert "efficiency_ratio=" in user_msg
+    assert "spectral_entropy=" in user_msg
+    assert "hurst=" in user_msg
+
+
+def test_propose_no_focus_block_when_style_not_in_library(mocker):  # type: ignore[no-untyped-def]
+    """Stile senza focus_metrics → nessun focus block nel user message."""
+    fake_llm = mocker.MagicMock()
+    fake_llm.complete.return_value = CompletionResult(
+        text=VALID_STRATEGY_JSON,
+        input_tokens=200,
+        output_tokens=80,
+        tier=ModelTier.C,
+        model="qwen",
+    )
+    # PromptLibrary senza focus_metrics per physicist
+    lib = PromptLibrary(
+        styles={"physicist": "Pensa come un fisico."},
+        focus={},
+    )
+    agent = HypothesisAgent(llm=fake_llm, prompt_library=lib)
+    proposal = agent.propose(make_genome(), make_summary())
+    assert proposal.strategy is not None
+    call_kwargs = fake_llm.complete.call_args.kwargs
+    user_msg = call_kwargs["user"]
+    assert "Focus per la tua lente:" not in user_msg
+
+
+def test_build_system_prompt_includes_role_and_guidance_from_library(mocker):  # type: ignore[no-untyped-def]
+    """agent_role e pattern_guidance da PromptLibrary appaiono nel SYSTEM prompt."""
+    fake_llm = mocker.MagicMock()
+    fake_llm.complete.return_value = CompletionResult(
+        text=VALID_STRATEGY_JSON,
+        input_tokens=200,
+        output_tokens=80,
+        tier=ModelTier.C,
+        model="qwen",
+    )
+    lib = PromptLibrary(
+        styles={"physicist": "Pensa come un fisico."},
+        focus={},
+        agent_role="ROLE_X",
+        pattern_guidance="GUIDE_X",
+    )
+    agent = HypothesisAgent(llm=fake_llm, prompt_library=lib)
+    agent.propose(make_genome(), make_summary())
+    call_kwargs = fake_llm.complete.call_args.kwargs
+    system_msg = call_kwargs["system"]
+    assert "ROLE_X" in system_msg
+    assert "GUIDE_X" in system_msg
+
+
+def test_build_system_prompt_skips_empty_optional_sections(mocker):  # type: ignore[no-untyped-def]
+    """domain_warnings='' e pattern_guidance='' → sezioni opzionali assenti nel SYSTEM."""
+    fake_llm = mocker.MagicMock()
+    fake_llm.complete.return_value = CompletionResult(
+        text=VALID_STRATEGY_JSON,
+        input_tokens=200,
+        output_tokens=80,
+        tier=ModelTier.C,
+        model="qwen",
+    )
+    lib = PromptLibrary(
+        styles={"physicist": "Pensa come un fisico."},
+        focus={},
+        domain_warnings="",
+        pattern_guidance="",
+    )
+    agent = HypothesisAgent(llm=fake_llm, prompt_library=lib)
+    agent.propose(make_genome(), make_summary())
+    call_kwargs = fake_llm.complete.call_args.kwargs
+    system_msg = call_kwargs["system"]
+    assert "WARNING DI DOMINIO" not in system_msg
+    assert "PATTERN GUIDANCE" not in system_msg
+
+
+def test_user_template_uses_instruction_from_library(mocker):  # type: ignore[no-untyped-def]
+    """instruction da PromptLibrary appare nel USER message; default non viene usato."""
+    fake_llm = mocker.MagicMock()
+    fake_llm.complete.return_value = CompletionResult(
+        text=VALID_STRATEGY_JSON,
+        input_tokens=200,
+        output_tokens=80,
+        tier=ModelTier.C,
+        model="qwen",
+    )
+    lib = PromptLibrary(
+        styles={"physicist": "Pensa come un fisico."},
+        focus={},
+        instruction="INSTR_X",
+    )
+    agent = HypothesisAgent(llm=fake_llm, prompt_library=lib)
+    agent.propose(make_genome(), make_summary())
+    call_kwargs = fake_llm.complete.call_args.kwargs
+    user_msg = call_kwargs["user"]
+    assert "INSTR_X" in user_msg
+    assert "Genera una strategia che cerchi anomalie sfruttabili in questo regime." not in user_msg
+
+
+def test_build_system_prompt_includes_anti_patterns_and_priorities(mocker):  # type: ignore[no-untyped-def]
+    """anti_patterns e output_priorities da PromptLibrary appaiono nel SYSTEM prompt."""
+    fake_llm = mocker.MagicMock()
+    fake_llm.complete.return_value = CompletionResult(
+        text=VALID_STRATEGY_JSON,
+        input_tokens=200,
+        output_tokens=80,
+        tier=ModelTier.C,
+        model="qwen",
+    )
+    lib = PromptLibrary(
+        styles={"physicist": "Pensa come un fisico."},
+        focus={},
+        anti_patterns="EVITA_X",
+        output_priorities="PRIORI_X",
+    )
+    agent = HypothesisAgent(llm=fake_llm, prompt_library=lib)
+    agent.propose(make_genome(), make_summary())
+    call_kwargs = fake_llm.complete.call_args.kwargs
+    system_msg = call_kwargs["system"]
+    assert "ANTI-PATTERN DA EVITARE" in system_msg
+    assert "EVITA_X" in system_msg
+    assert "PRIORITA' DI OUTPUT" in system_msg
+    assert "PRIORI_X" in system_msg
+
+
+def test_build_system_prompt_skips_anti_patterns_and_priorities_when_empty(mocker):  # type: ignore[no-untyped-def]
+    """anti_patterns='' e output_priorities='' -> sezioni opzionali assenti nel SYSTEM."""
+    fake_llm = mocker.MagicMock()
+    fake_llm.complete.return_value = CompletionResult(
+        text=VALID_STRATEGY_JSON,
+        input_tokens=200,
+        output_tokens=80,
+        tier=ModelTier.C,
+        model="qwen",
+    )
+    lib = PromptLibrary(
+        styles={"physicist": "Pensa come un fisico."},
+        focus={},
+        anti_patterns="",
+        output_priorities="",
+    )
+    agent = HypothesisAgent(llm=fake_llm, prompt_library=lib)
+    agent.propose(make_genome(), make_summary())
+    call_kwargs = fake_llm.complete.call_args.kwargs
+    system_msg = call_kwargs["system"]
+    assert "ANTI-PATTERN" not in system_msg
+    assert "PRIORITA' DI OUTPUT" not in system_msg
@@ -63,3 +63,51 @@ def test_build_summary_new_fields_populated() -> None:
    assert isinstance(s.seasonality_dow, float)
    assert 0.0 <= s.seasonality_hour <= 1.0
    assert 0.0 <= s.seasonality_dow <= 1.0
+
+
+def test_build_summary_geometric_fields_populated() -> None:
+    """I 7 nuovi campi geometrico-frattali devono essere non-default e nei range attesi."""
+    idx = pd.date_range("2024-01-01", periods=600, freq="1h", tz="UTC")
+    np.random.seed(5)
+    close = 100 + np.cumsum(np.random.normal(0, 1, 600))
+    df = pd.DataFrame(
+        {"open": close, "high": close + 0.5, "low": close - 0.5, "close": close, "volume": 1.0},
+        index=idx,
+    )
+    s = build_market_summary(df, symbol="BTC/USDT", timeframe="1h")
+
+    # efficiency_ratio: float in [0, 1]
+    assert isinstance(s.efficiency_ratio, float)
+    assert 0.0 <= s.efficiency_ratio <= 1.0
+
+    # tail_index_left/right: float in [1, 10]
+    assert isinstance(s.tail_index_left, float)
+    assert isinstance(s.tail_index_right, float)
+    assert 1.0 <= s.tail_index_left <= 10.0
+    assert 1.0 <= s.tail_index_right <= 10.0
+
+    # structural_uptrend: float in [0, 1]
+    assert isinstance(s.structural_uptrend, float)
+    assert 0.0 <= s.structural_uptrend <= 1.0
+
+    # compression: float > 0
+    assert isinstance(s.compression, float)
+    assert s.compression > 0.0
+
+    # spectral_entropy: float in [0, 1]
+    assert isinstance(s.spectral_entropy, float)
+    assert 0.0 <= s.spectral_entropy <= 1.0
+
+    # dominant_cycle: None or positive float
+    assert s.dominant_cycle is None or (isinstance(s.dominant_cycle, float) and s.dominant_cycle > 0)
+
+    # Verifica che i campi siano stati popolati (non tutti ai valori default)
+    defaults_only = (
+        s.efficiency_ratio == 0.0
+        and s.tail_index_left == 5.0
+        and s.tail_index_right == 5.0
+        and s.structural_uptrend == 0.5
+        and s.compression == 1.0
+        and s.spectral_entropy == 1.0
+    )
+    assert not defaults_only, "Tutti i campi geometrici sono rimasti ai valori default: calcolo non avvenuto"
@@ -4,10 +4,15 @@ import pytest

 from multi_swarm_core.metrics.basic import (
    autocorr_lag1,
+    compression_ratio,
+    efficiency_ratio_kaufman,
    hurst_exponent,
    max_drawdown,
    seasonality_strength,
    sharpe_ratio,
+    spectral_entropy_and_cycle,
+    structural_uptrend_score,
+    tail_index_hill,
    total_return,
    vol_percentile_historical,
 )
@@ -166,3 +171,120 @@ def test_seasonality_strength_invalid_by():
    r = pd.Series(np.random.normal(0, 0.01, 100), index=idx)
    with pytest.raises(ValueError, match="'minute'"):
        seasonality_strength(r, by="minute")
+
+
+# --- Geometric/fractal metrics tests ---
+
+
+def test_efficiency_ratio_pure_trend():
+    """Slope perfetto (prezzi linearmente crescenti) -> efficiency_ratio vicino a 1.0."""
+    prices = pd.Series(np.linspace(100.0, 200.0, 200))
+    er = efficiency_ratio_kaufman(prices, length=100)
+    assert er == pytest.approx(1.0, abs=1e-6)
+
+
+def test_efficiency_ratio_random_walk():
+    """Random walk iid -> efficiency_ratio basso (attorno a 0.1-0.4)."""
+    np.random.seed(42)
+    prices = pd.Series(100.0 + np.cumsum(np.random.normal(0, 1, 300)))
+    er = efficiency_ratio_kaufman(prices, length=100)
+    assert 0.0 <= er <= 0.6
+
+
+def test_efficiency_ratio_short_series():
+    """Serie troppo corta ritorna 0.0."""
+    prices = pd.Series([100.0, 101.0, 102.0])
+    assert efficiency_ratio_kaufman(prices, length=100) == 0.0
+
+
+def test_tail_index_hill_left_right_separate():
+    """tail_index_hill left e right ritornano float in [1, 10]."""
+    np.random.seed(7)
+    r = pd.Series(np.random.normal(0, 0.01, 500))
+    left = tail_index_hill(r, side="left")
+    right = tail_index_hill(r, side="right")
+    assert 1.0 <= left <= 10.0
+    assert 1.0 <= right <= 10.0
+
+
+def test_tail_index_hill_invalid_side():
+    """side non valido solleva ValueError."""
+    r = pd.Series(np.random.normal(0, 0.01, 200))
+    with pytest.raises(ValueError, match="'both'"):
+        tail_index_hill(r, side="both")
+
+
+def test_tail_index_hill_short_series():
+    """Serie < 50 ritorna default 5.0."""
+    r = pd.Series([0.01, -0.01, 0.02])
+    assert tail_index_hill(r, side="left") == pytest.approx(5.0)
+    assert tail_index_hill(r, side="right") == pytest.approx(5.0)
+
+
+def test_structural_uptrend_pure_uptrend():
+    """Prezzi costantemente crescenti -> structural_uptrend vicino a 1.0."""
+    prices = pd.Series(np.linspace(100.0, 200.0, 100))
+    score = structural_uptrend_score(prices, window=5)
+    assert score >= 0.5
+
+
+def test_structural_uptrend_pure_downtrend():
+    """Prezzi costantemente decrescenti -> structural_uptrend vicino a 0.0."""
+    prices = pd.Series(np.linspace(200.0, 100.0, 100))
+    score = structural_uptrend_score(prices, window=5)
+    assert score <= 0.5
+
+
+def test_structural_uptrend_short_series():
+    """Serie troppo corta ritorna 0.5 (neutro)."""
+    prices = pd.Series([100.0, 101.0, 99.0])
+    assert structural_uptrend_score(prices, window=5) == pytest.approx(0.5)
+
+
+def test_compression_ratio_compress_then_expand():
+    """Finestra recente compressa rispetto a ref -> ratio < 1."""
+    np.random.seed(1)
+    # Ref window: alta volatilita
+    ref_part = np.random.normal(0, 5.0, 200)
+    # Recent window: bassa volatilita
+    recent_part = np.random.normal(0, 0.2, 50)
+    prices = pd.Series(100.0 + np.cumsum(np.concatenate([ref_part, recent_part])))
+    ratio = compression_ratio(prices, recent_window=50, ref_window=200)
+    assert ratio < 1.0
+
+
+def test_compression_ratio_short_series():
+    """Serie troppo corta ritorna 1.0 (neutro)."""
+    prices = pd.Series([100.0, 101.0, 99.0])
+    assert compression_ratio(prices, recent_window=50, ref_window=200) == pytest.approx(1.0)
+
+
+def test_spectral_entropy_pure_sine():
+    """Seno puro -> entropy bassa e dominant_cycle ben definito."""
+    n = 256
+    t = np.arange(n)
+    cycle_period = 32  # barre
+    series = np.sin(2 * np.pi * t / cycle_period)
+    r = pd.Series(series)
+    entropy, cycle = spectral_entropy_and_cycle(r, length=256)
+    assert entropy < 0.6
+    assert cycle is not None
+    # Il ciclo dominante deve essere vicino a cycle_period
+    assert abs(cycle - cycle_period) <= 5.0
+
+
+def test_spectral_entropy_white_noise():
+    """Rumore bianco -> entropy alta e dominant_cycle = None."""
+    np.random.seed(99)
+    r = pd.Series(np.random.normal(0, 1, 512))
+    entropy, cycle = spectral_entropy_and_cycle(r, length=256)
+    assert entropy >= 0.6
+    assert cycle is None
+
+
+def test_spectral_entropy_short_series():
+    """Serie troppo corta ritorna (1.0, None)."""
+    r = pd.Series([0.01, -0.01, 0.02])
+    entropy, cycle = spectral_entropy_and_cycle(r, length=256)
+    assert entropy == pytest.approx(1.0)
+    assert cycle is None
@@ -0,0 +1,78 @@
+"""Test che `_parallel_propose` preservi l'ordine dei risultati e funzioni
+sia in modalita' sequenziale (workers=1) che concorrente (workers>1).
+
+Non vogliamo testare il vero `HypothesisAgent.propose()` (che fa chiamate
+LLM); usiamo un dummy con una latenza simulata per validare ordine e parallelismo.
+"""
+
+from __future__ import annotations
+
+import time
+from dataclasses import dataclass
+from typing import Any
+
+from multi_swarm_core.orchestrator.run import _parallel_propose
+
+
+@dataclass
+class _FakeGenome:
+    id: str
+
+
+@dataclass
+class _FakeProposal:
+    genome_id: str
+
+
+class _FakeAgent:
+    """Agent dummy: propose() dorme 50ms e ritorna un proposal con l'id del genome."""
+
+    def __init__(self, delay_s: float = 0.05) -> None:
+        self._delay = delay_s
+        self.call_count = 0
+
+    def propose(self, genome: _FakeGenome, market: Any) -> _FakeProposal:
+        time.sleep(self._delay)
+        self.call_count += 1
+        return _FakeProposal(genome_id=genome.id)
+
+
+def test_parallel_propose_preserves_order_serial() -> None:
+    agent = _FakeAgent(delay_s=0.01)
+    genomes = [_FakeGenome(id=f"g{i}") for i in range(5)]
+    results = _parallel_propose(agent, genomes, market=None, n_workers=1)
+    assert [r.genome_id for r in results] == ["g0", "g1", "g2", "g3", "g4"]
+    assert agent.call_count == 5
+
+
+def test_parallel_propose_preserves_order_concurrent() -> None:
+    agent = _FakeAgent(delay_s=0.05)
+    genomes = [_FakeGenome(id=f"g{i}") for i in range(8)]
+    results = _parallel_propose(agent, genomes, market=None, n_workers=4)
+    assert [r.genome_id for r in results] == [f"g{i}" for i in range(8)]
+    assert agent.call_count == 8
+
+
+def test_parallel_propose_actually_parallelizes() -> None:
+    """Wall time con 4 worker su 4 task da 100ms deve essere ~100ms, non ~400ms."""
+    agent = _FakeAgent(delay_s=0.1)
+    genomes = [_FakeGenome(id=f"g{i}") for i in range(4)]
+    t0 = time.time()
+    _parallel_propose(agent, genomes, market=None, n_workers=4)
+    elapsed = time.time() - t0
+    # serial sarebbe 0.4s; con 4 worker scendiamo a ~0.1s (max 0.2 per overhead).
+    assert elapsed < 0.2, f"expected <200ms with 4 workers, got {elapsed * 1000:.0f}ms"
+
+
+def test_parallel_propose_handles_single_genome() -> None:
+    agent = _FakeAgent()
+    results = _parallel_propose(agent, [_FakeGenome(id="solo")], market=None, n_workers=8)
+    assert len(results) == 1
+    assert results[0].genome_id == "solo"
+
+
+def test_parallel_propose_empty_input() -> None:
+    agent = _FakeAgent()
+    results = _parallel_propose(agent, [], market=None, n_workers=4)
+    assert results == []
+    assert agent.call_count == 0
@@ -0,0 +1,146 @@
+"""Unit tests per PromptLibrary v3.0 — nuovi campi top-level."""
+
+from __future__ import annotations
+
+import json
+import tempfile
+from pathlib import Path
+
+import pytest
+
+from multi_swarm_core.genome.prompt_library import PromptLibrary, PromptLibraryError
+
+
+def _write_json(data: dict, tmp_path: Path) -> Path:
+    p = tmp_path / "prompts.json"
+    p.write_text(json.dumps(data), encoding="utf-8")
+    return p
+
+
+def test_from_json_loads_new_top_level_fields(tmp_path: Path) -> None:
+    """from_json() legge agent_role, pattern_guidance, instruction, domain_warnings."""
+    data = {
+        "styles": {"physicist": {"directive": "Cerca leggi conservative."}},
+        "agent_role": "Sei un agente test.",
+        "pattern_guidance": "Forme di curva: trend, compressione.",
+        "instruction": "Genera qualcosa di utile.",
+        "domain_warnings": "Attenzione: mercato 24/7.",
+    }
+    lib = PromptLibrary.from_json(_write_json(data, tmp_path))
+    assert lib.agent_role == "Sei un agente test."
+    assert lib.pattern_guidance == "Forme di curva: trend, compressione."
+    assert lib.instruction == "Genera qualcosa di utile."
+    assert lib.domain_warnings == "Attenzione: mercato 24/7."
+
+
+def test_from_json_defaults_new_fields_when_absent(tmp_path: Path) -> None:
+    """from_json() usa stringa vuota come default per i 4 campi opzionali."""
+    data = {
+        "styles": {"physicist": {"directive": "Cerca leggi conservative."}},
+    }
+    lib = PromptLibrary.from_json(_write_json(data, tmp_path))
+    assert lib.agent_role == ""
+    assert lib.pattern_guidance == ""
+    assert lib.instruction == ""
+    assert lib.domain_warnings == ""
+
+
+def test_default_provides_universal_fallbacks() -> None:
+    """PromptLibrary.default() ha agent_role, pattern_guidance, instruction non vuoti; domain_warnings vuoto."""
+    lib = PromptLibrary.default()
+    assert lib.agent_role != ""
+    assert lib.pattern_guidance != ""
+    assert lib.instruction != ""
+    assert lib.domain_warnings == ""
+
+
+def test_prompt_library_rejects_whitespace_only_fields() -> None:
+    """Campi opzionali forniti ma solo whitespace sollevano PromptLibraryError."""
+    with pytest.raises(PromptLibraryError, match="agent_role"):
+        PromptLibrary(
+            styles={"physicist": "Cerca leggi."},
+            focus={},
+            agent_role="   ",
+        )
+
+
+def test_prompt_library_accepts_empty_string_for_optional_fields() -> None:
+    """Stringa vuota '' e' accettata per tutti i campi opzionali."""
+    lib = PromptLibrary(
+        styles={"physicist": "Cerca leggi."},
+        focus={},
+        agent_role="",
+        pattern_guidance="",
+        instruction="",
+        domain_warnings="",
+    )
+    assert lib.agent_role == ""
+    assert lib.domain_warnings == ""
+
+
+def test_from_json_loads_anti_patterns_and_output_priorities(tmp_path: Path) -> None:
+    """from_json() legge anti_patterns e output_priorities (v3.1)."""
+    data = {
+        "styles": {"physicist": {"directive": "Cerca leggi conservative."}},
+        "anti_patterns": "Evita overfitting.",
+        "output_priorities": "Robustezza > ottimalita.",
+    }
+    lib = PromptLibrary.from_json(_write_json(data, tmp_path))
+    assert lib.anti_patterns == "Evita overfitting."
+    assert lib.output_priorities == "Robustezza > ottimalita."
+
+
+def test_strategy_crypto_directives_ascii_safe() -> None:
+    """REGRESSION GUARD: nessuna directive contiene caratteri > U+007F.
+
+    v3.1 aveva regredito introducendo il carattere circa-uguale (U+2248) in 3 stili.
+    v3.2 ripristina ASCII-strict come invariante permanente.
+    """
+    import importlib.resources
+
+    path = importlib.resources.files("strategy_crypto") / "prompts.json"
+    lib = PromptLibrary.from_json(str(path))
+
+    for style, directive in lib.styles.items():
+        non_ascii = [c for c in directive if ord(c) > 127]
+        assert not non_ascii, (
+            f"directive di {style!r} contiene caratteri non-ASCII: "
+            f"{non_ascii} (codepoints: {[hex(ord(c)) for c in non_ascii]})"
+        )
+
+
+def test_strategy_crypto_directives_have_archetype_marker() -> None:
+    """REGRESSION GUARD: ogni directive chiude con 'Archetipo dominante: ...'.
+
+    L'archetipo e' l'ancora semantica identitaria della lente; deve essere
+    presente per resistere alle riscritture di mutate_prompt_llm.
+    """
+    import importlib.resources
+
+    path = importlib.resources.files("strategy_crypto") / "prompts.json"
+    lib = PromptLibrary.from_json(str(path))
+
+    for style, directive in lib.styles.items():
+        assert "Archetipo dominante:" in directive, (
+            f"directive di {style!r} manca del marker 'Archetipo dominante:'"
+        )
+
+
+def test_strategy_crypto_directives_have_lookback_hint() -> None:
+    """REGRESSION GUARD: ogni directive contiene un hint 'Lookback consigliato: X-Y barre'.
+
+    Il range numerico orienta il parametro evoluto lookback_window del genoma;
+    differenziato per stile per favorire diversita di scala temporale nella
+    popolazione iniziale.
+    """
+    import re
+    import importlib.resources
+
+    path = importlib.resources.files("strategy_crypto") / "prompts.json"
+    lib = PromptLibrary.from_json(str(path))
+
+    pattern = re.compile(r"[Ll]ookback consigliato:\s*\d+\s*-\s*\d+", re.IGNORECASE)
+    for style, directive in lib.styles.items():
+        assert pattern.search(directive), (
+            f"directive di {style!r} manca dell'hint 'Lookback consigliato: X-Y'"
+        )
@@ -18,3 +18,4 @@ build-backend = "hatchling.build"

 [tool.hatch.build.targets.wheel.force-include]
 "strategy_crypto/strategies" = "strategy_crypto/strategies"
+"strategy_crypto/prompts.json" = "strategy_crypto/prompts.json"
@@ -1,29 +1,50 @@
 {
  "_comment": "Stili cognitivi e direttive del system_prompt per il GA di strategy_crypto. Modifica liberamente: cambia 'directive' di uno stile esistente o aggiungi nuove voci a 'styles'. Il nome dello stile (key) viene usato come 'cognitive_style' del genoma; la 'directive' diventa il system_prompt iniziale.",
-  "_schema": "2.1",
-  "_changelog": "v2.1 - directive estese con interpretazione style-specific dei 4 nuovi input statistici (autocorr_lag1, hurst, vol_pct, seasonality). v2.0 - Riprogettato per blind-generator GA. Directive medio-compatte (~700 char) che orientano l'esplorazione cognitiva senza prescrivere indicatori specifici (lascia evolvere il GA). Mappate sulle 4 statistiche disponibili: mean, std, skew, kurtosis + volatility_regime. Rimosse ecologist (richiede multi-asset), game_theorist/epidemiologist (richiedono info esterne non visibili all'agente). Tenute 7 lenti che mappano bene sulle statistiche aggregate.",
-  "_design_notes": "Le directive sono BIAS DI ESPLORAZIONE, non template. Suggeriscono cosa cercare nei 4 momenti e quali archetipi di strategia preferire, lasciando al GA la scoperta della combinazione esatta di indicatori e soglie. Sono pensate per essere riscritte dall'operatore mutate_prompt_llm mantenendo coerenza con la lente.",
+  "_schema": "3.2",
+  "_changelog": "v3.2 - Patch consolidamento: ripristinati 3 invarianti regrediti in v3.1 (ASCII-safe, archetipo dominante, hint lookback); voce attiva rinforzata; anti_patterns +2 (chattering, isteresi); output_priorities +1 (#1 coerenza con lente cognitiva); domain_warnings +1 frase (soglia seasonality 0.05); NEW _design_invariants metadata. Lunghezza directive 800-950 char (era 545-614 in v3.1, troppo snellite). v3.1 - Refactor contenuto post-diagnosi. v3.0 - Refactor compositore. v2.2 - Metriche geometrico-frattali. v2.1 - directive estese. v2.0 - Riprogettato per blind-generator GA.",
+  "_focus_metrics_design": "Le focus_metrics sono ENFASI per la lente, non filtri. Standardizzate a 4 per stile (cognitive budget). Evitano ridondanze con la sezione 'Regime recente' del USER_TEMPLATE.",
+  "_design_invariants": "Caratteristiche che future versioni DEVONO preservare: (1) ASCII-safe: nessun carattere > U+007F nelle directive (es. il carattere circa-uguale U+2248 era una regressione di v3.1, ripristinato in v3.2); (2) Archetipo dominante: ogni directive chiude con 'Archetipo dominante: <metafora>.' come ancora identitaria resistente a mutate_prompt_llm; (3) Lookback consigliato: ogni directive include un range numerico diversificato per stile per orientare il parametro evoluto lookback_window del genoma; (4) Metafora ancorante: la lente cognitiva e' descritta in forma 'Il mercato e ...' come prima frase; (5) Lunghezza directive: tra 800 e 950 char (sweet spot per robustezza a mutate_prompt_llm).",
+
+  "agent_role": "Sei un agente generatore di ipotesi di trading quantitativo per un sistema swarm coevolutivo specializzato in mercati crypto. Sei parte di una popolazione che esplora collettivamente lo spazio delle strategie: la diversita delle ipotesi e un asset critico per il sistema. Preferisci esplorare territori meno ovvi rispetto a quelli che la tua lente cognitiva renderebbe predicibili.",
+
+  "pattern_guidance": "Forme di curva da cercare:\n  - Trend strutturale (direzione persistente con basso ritracciamento)\n  - Compressione di volatilita (pre-breakout, energia accumulata)\n  - Espansione di volatilita (regime di shock: momentum o cattura difensiva)\n  - Mean reversion strutturale (deviazione eccessiva dalla media verso ritorno atteso)\n  - Esaurimento direzionale (estremi di oscillatori, divergenza prezzo-momento)\n\nRipetibilita da sfruttare:\n  - Eventi crossover ricorrenti su serie correlate\n  - Cicli intra-day se la seasonality oraria e significativa\n  - Cicli settimanali se la seasonality settimanale e significativa\n  - Pattern doppio (top/bottom) con conferma su livello simile\n  - Range breakout dopo periodo di compressione\n\nCerca pattern che si REPLICANO nei dati storici, non singoli eventi rari.",
+
+  "instruction": "Genera una strategia che cerchi anomalie sfruttabili in questo regime crypto.",
+
+  "domain_warnings": "Crypto trada 24/7 senza CME gap: non assumere chiusure settimanali. Tail pesanti e vol clustering esteso caratterizzano BTC/ETH: evita ipotesi gaussiane. NON tentare di inferire funding rate, news o eventi macro: non sono accessibili. Le statistiche fornite sono l'unica informazione su cui basarsi. Una seasonality_hour o seasonality_dow > 0 NON significa che la stagionalita sia significativa: usa la soglia 0.05 come gate minimo (sotto e' rumore statistico).",
+
+  "anti_patterns": "Evita: (1) basare strategia su singolo evento estremo (kurt outlier senza ricorrenza nelle 500 barre recenti); (2) usare piu di 4 condizioni in AND (sovra-fitting combinatorio, brittle a piccoli shift di regime); (3) confondere correlazione storica con causalita (autocorr o pattern temporali sono BIAS, non leggi); (4) assumere stazionarieta perfetta del regime (i delta 'recente vs baseline' indicano lo scostamento); (5) usare feature temporali (hour, dow, is_weekend) quando la seasonality corrispondente e' < 0.05 (rumore, no signal); (6) crossover tra indicatori dello stesso tipo con lookback vicini (chattering: oscilla tra entry/exit senza segnale netto); (7) soglie hard senza isteresi tra entry ed exit (genera chattering al confine; usa soglie diverse per entry vs exit, almeno 10-20% di gap).",
+
+  "output_priorities": "Quando emerge trade-off: (1) coerenza con la lente cognitiva: la strategia deve essere riconoscibile come emanata dal tuo stile (es. engineer = pochi gate robusti, psychologist = contrarian su estremi). E' il fondamento del design swarm: ipotesi omogeneizzate riducono la diversita della popolazione; (2) robustezza cross-regime > ottimalita su singolo regime; (3) semplicita interpretabile (2-3 condizioni con razionale chiaro) > complessita raffinata (5+ condizioni accordate); (4) selettivita (poche entry forti, alto SNR) > attivita (molte entry deboli); (5) condizioni con razionale meccanico > pattern statistici fortuiti.",
+
  "styles": {
    "physicist": {
-      "directive": "Pensa come un fisico: il mercato e un sistema con leggi di conservazione e regimi di scala. Leggi std come dispersione energetica e kurtosis come densita di eventi estremi (kurt alta = fat tails, sistema fuori equilibrio). Cerca simmetrie nei ritorni (skew ≈0 = sistema simmetrico) e rotture (skew marcato = forzante asimmetrica). AR(1) positivo significativamente sopra baseline = sistema con memoria fuori equilibrio, momentum legittimo; Hurst > 0.55 conferma persistenza di scala; vol percentile alto + kurt bassa = energia immagazzinata non ancora rilasciata. Preferisci ritorno all'equilibrio in regime simmetrico/basso vol, propagazione (momentum/breakout) in regime asimmetrico/alto vol. Pattern coerenti su piu lookback sono robusti, pattern singoli sono rumore."
+      "directive": "Il mercato e un sistema fisico con energia (std), simmetrie (skew) e memoria (autocorr). Leggi kurt come densita di eventi estremi (fat tails = fuori equilibrio), skew come forzante asimmetrica. AR(1) positivo molto sopra baseline = memoria coerente, costruisci ipotesi di momentum legittimo; Hurst > 0.55 conferma persistenza di scala su orizzonti multipli; vol_pct alto con kurt bassa = energia immagazzinata non ancora rilasciata, cattura il rilascio; efficiency_ratio elevato = movimento efficiente, sfrutta direzionalita; spectral_entropy bassa con dominant_cycle definito = modi armonici sfruttabili, combina con conferma sulla fase. Preferisci pattern coerenti su piu lookback rispetto a singoli eventi rumorosi. Diagnostica regimi simmetrici e rotture di simmetria. Lookback consigliato: 150-300 barre. Archetipo dominante: sistema fisico in equilibrio (o pre-rottura di simmetria).",
+      "focus_metrics": ["hurst", "dominant_cycle", "efficiency_ratio", "spectral_entropy"]
    },
    "biologist": {
-      "directive": "Pensa come un biologo evoluzionista: il mercato e un ecosistema di strategie in competizione. Skew negativo = predazione asimmetrica (vol-selling crowded che subisce shock). Skew positivo = predatori che cacciano breakout. Kurt alta = eventi di estinzione/fioritura. AR(1) positivo persistente = una specie sta colonizzando la nicchia (overcrowding imminente, fade); Hurst > 0.55 con vol percentile basso = nicchia stabile (occupa); seasonality forte = ritmo biologico ricorrente, sfruttabile. Preferisci contrarian in regime di skew estremo (fade la specie dominante) e coordinamento in regime simmetrico. Pattern asimmetrici: cattura la coda opposta al consensus."
+      "directive": "Il mercato e un ecosistema dove strategie competono per alpha finito. Skew negativo segnala predazione asimmetrica (vol-selling crowded subisce shock), positivo predatori che cacciano breakout. Kurt alta = eventi di estinzione o fioritura. AR(1) positivo persistente = una specie sta colonizzando la nicchia (overcrowding imminente, preferisci fade); Hurst > 0.55 con vol_pct basso = nicchia stabile (occupa con strategie direzionali); tail asimmetrico (left molto piu pesante di right) = predazione asimmetrica strutturale, costruisci contrarian sulla coda; structural_uptrend persistente = specie dominante stabile. Combina seasonality con uno o due gate di regime per evitare di sovrapporti a fasi gia mature. Cattura la coda opposta al consensus. Lookback consigliato: 80-200 barre. Archetipo dominante: ecosistema con dinamiche predator-prey e nicchie evolutive.",
+      "focus_metrics": ["tail_left", "tail_right", "structural_uptrend", "seasonality_hour"]
    },
    "historian": {
-      "directive": "Pensa come uno storico: i regimi si ripetono ma non identici. Usa mean come drift strutturale e std come ampiezza ciclo. Kurt alta + vol regime medium/high = fase tardiva (eventi estremi addensati, pre-transizione); kurt bassa + skew ≈0 = accumulazione/stabilita. AR(1) recente >> baseline storica = regime sta accelerando rispetto al normale; Hurst > 0.55 + vol percentile alto = fase markup matura, mean reversion attesa; seasonality forte = abitudini collettive consolidate, replicabili. Preferisci mean reversion strutturali: deviazioni significative tendono a ritornare su orizzonti multipli. Identifica analogie tra regime corrente e fasi tipiche (compressione vol, espansione, esaurimento trend)."
+      "directive": "Il mercato attraversa fasi cicliche che si ripetono in forma simile. Mean = drift strutturale, std = ampiezza ciclo, kurt alta + vol regime medium/high = fase tardiva (pre-transizione); kurt bassa + skew vicino a zero = fase di accumulazione o stabilita. AR(1) recente molto sopra baseline storica = regime accelera rispetto al normale, diagnostica se markup o distribuzione; Hurst > 0.55 con vol_pct alto = fase markup matura, costruisci ipotesi di mean reversion strutturale attesa; structural_uptrend sostenuto = fase di accumulo o markup attiva, sfrutta la direzionalita; compression < 1 = consolidamento pre-fase nuova, preferisci breakout direzionali a conferma di rottura. Identifica analogie tra il regime corrente e fasi tipiche (accumulazione, markup, distribuzione, markdown). Lookback consigliato: 200-500 barre. Archetipo dominante: ciclo storico ricorrente in fasi tipiche.",
+      "focus_metrics": ["autocorr_recent", "structural_uptrend", "compression", "hurst"]
    },
    "meteorologist": {
-      "directive": "Pensa come un meteorologo: la volatilita ha regimi persistenti con transizioni brusche. Vol regime e input primario. Std + kurt = microclima: std bassa + kurt bassa = calma stabile (vendi vol), std alta + kurt alta = tempesta (compra convexity), std bassa + kurt alta = calma ingannevole (riduci esposizione). AR(1) recente > baseline = fronte persistente in arrivo; Hurst > 0.55 = sistema su scala lunga (ciclone), Hurst < 0.45 = turbolenza locale (no trend persistente); vol percentile estremo = posizione nel ciclo seasonal; seasonality alta = pattern cyclonico ricorrente. Strategie regime-aware: gate espliciti su vol che attivano logiche diverse. Pattern multi-regime sono robusti."
+      "directive": "La volatilita ha climi persistenti e fronti di transizione. Vol_regime + std + kurt definiscono il microclima: std bassa + kurt bassa = calma stabile (preferisci vendere vol con gate sicuri); std alta + kurt alta = tempesta (compra convexity o resta flat); std bassa + kurt alta = calma ingannevole pre-fronte (riduci esposizione). AR(1) recente sopra baseline = fronte persistente in arrivo, cattura la direzione del fronte; Hurst > 0.55 = sistema su scala lunga (ciclone), Hurst < 0.45 = turbolenza locale (no trend persistente, preferisci range-trading); vol_pct estremo = posizione nel ciclo seasonal, modula la size; compression < 1 = compressione vol pre-rilascio, posizionati per il breakout. Costruisci strategie con gate espliciti su vol che attivano logiche diverse. Lookback consigliato: 50-150 barre. Archetipo dominante: clima atmosferico con fronti e regimi persistenti.",
+      "focus_metrics": ["vol_pct", "compression", "tail_right", "dominant_cycle"]
    },
    "engineer": {
-      "directive": "Pensa come un ingegnere di controllo: ogni segnale deve avere SNR favorevole e robustezza. Std e il rumore di fondo: il segnale deve essere significativamente piu grande della varianza intraday. Kurt alta riduce affidabilita dei segnali medi. AR(1) > 0.05 con std contenuta = SNR favorevole, segnale azionabile; AR(1) ≈0 = random walk, non costruirci sopra; Hurst < 0.45 = filtro mean-reversion causale efficace; vol percentile > 80 = saturazione (sensori instabili, riduci leverage); seasonality < 0.05 = feature temporali sono rumore, NON usarle. Pattern semplici e tarabili: poche condizioni in AND, soglie con margine, isteresi entry/exit. Robustezza > ottimalita su singolo regime."
+      "directive": "Tratta ogni segnale come un sistema di controllo: serve SNR favorevole, causalita, robustezza. Std e il rumore di fondo. Kurt alta riduce affidabilita dei segnali medi (gli estremi dominano le statistiche). AR(1) > 0.05 con std contenuta = SNR favorevole, costruisci ipotesi di momentum filtrato; AR(1) vicino a zero = random walk, evita di costruire signal su questo; Hurst < 0.45 = filtro mean-reversion causale efficace, sfrutta con isteresi; efficiency_ratio < 0.2 = no signal, non costruire; spectral_entropy > 0.8 = white noise, regime non modellabile; tail_index < 2.5 = saturazione dei sensori, riduci leverage; seasonality < 0.05 = feature temporali sono rumore, NON usarle. Preferisci pattern semplici e tarabili: poche condizioni in AND, soglie con margine, isteresi entry/exit. Lookback consigliato: 60-120 barre. Archetipo dominante: sistema di controllo ingegneristico con SNR e robustezza.",
+      "focus_metrics": ["efficiency_ratio", "spectral_entropy", "tail_left", "autocorr_recent"]
    },
    "military_strategist": {
-      "directive": "Pensa come uno stratega: distingui regime offensivo da difensivo. Vol regime medium/low + skew positivo + kurt moderata = terreno favorevole all'attacco (entry direzionali su breakout/momentum). Vol regime high + kurt elevata = terreno ostile (difesa: posizioni limitate, exit rapide). AR(1) > 0 = vento alle spalle, carica con momentum; AR(1) < 0 = imboscata possibile, contrarian; Hurst > 0.55 = posizione difendibile (hold trade); vol percentile alta = artiglieria nemica attiva (ritirata); seasonality forte = via predicibile da percorrere, sfruttala. Concentrazione: poche condizioni forti. Economia: flat quando il segnale non e dominante. Sorpresa: contrarian su skew estremo (consensus). Sicurezza: ogni entry con exit chiara."
+      "directive": "Distingui campagna offensiva da campagna difensiva e adatta dottrina. Vol regime medium/low + skew positivo + kurt moderata = terreno favorevole all'attacco (costruisci entry direzionali su breakout o momentum). Vol regime high + kurt elevata = terreno ostile (difesa: posizioni limitate, exit rapide, gate restrittivi). AR(1) > 0 = vento alle spalle, carica con momentum; AR(1) negativo = imboscata possibile, preferisci contrarian; Hurst > 0.55 = posizione difendibile, hold trade; structural_uptrend > 0.7 = terreno occupato dall'avversario (decidi se attaccare o ritirarti); compression < 0.5 = preparazione attacco silenziosa, posizionati per breakout; vol_pct alta = artiglieria nemica attiva, ritirata. Concentrazione: poche condizioni forti. Sorpresa: contrarian su consensus estremo. Lookback consigliato: 100-200 barre. Archetipo dominante: stratega militare che bilancia offesa e difesa.",
+      "focus_metrics": ["structural_uptrend", "compression", "vol_pct", "tail_left"]
    },
    "psychologist": {
-      "directive": "Pensa come uno psicologo del comportamento collettivo: skew e kurt catturano emozioni. Skew neg + kurt alta = paura ricorrente (capitulation spikes, fade gli estremi al ribasso). Skew pos + kurt alta = euforia (FOMO spikes, fade gli estremi al rialzo). Skew ≈0 + kurt bassa = apatia/range (gioca i bordi). AR(1) recente >> baseline = euforia coordinata in corso, posizionati contro l'ultimo arrivato; Hurst > 0.55 = trance collettiva (trend trance, dura piu del razionale); vol percentile estremo + kurt alta = momentum emozionale puro; seasonality intra-day forte = bias circadiani sfruttabili (FOMO apertura US, panico chiusura asiatica). Estremi di oscillatori in regimi emotivi (kurt alta), crossover in regimi razionali (kurt ≈3). Contrarian sugli estremi, continuazione sulle medie."
+      "directive": "Il mercato e folla con emozioni misurabili. Skew e kurt sono il termometro emotivo: skew neg + kurt alta = paura ricorrente (capitulation spikes, cattura il rimbalzo); skew pos + kurt alta = euforia (FOMO spikes, preferisci fade gli estremi al rialzo); skew vicino a zero + kurt bassa = apatia o range (gioca i bordi del range). AR(1) recente molto sopra baseline = euforia coordinata in corso, posizionati contro l'ultimo arrivato; Hurst > 0.55 = trance collettiva (trend trance, dura piu del razionale); tail_left pesante (Hill < 2.5) = paura sistemica strutturale, contrarian sulla capitulation; spectral_entropy alta = caos comportamentale, riduci dimensionalita del signal; vol_pct estremo = momentum emozionale puro, fade gli estremi. Sfrutta crossover di oscillatori in regimi razionali (kurt vicina a 3). Lookback consigliato: 50-120 barre. Archetipo dominante: psicologo del comportamento collettivo.",
+      "focus_metrics": ["tail_left", "tail_right", "autocorr_recent", "spectral_entropy"]
    }
  }
 }
@@ -0,0 +1,140 @@
+{
+  "rules": [
+    {
+      "condition": {
+        "op": "and",
+        "args": [
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "realized_vol",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": 0.01
+              }
+            ]
+          },
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "atr_pct",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": 0.02
+              }
+            ]
+          },
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "sma_pct",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": 0.05
+              }
+            ]
+          }
+        ]
+      },
+      "action": "entry-long"
+    },
+    {
+      "condition": {
+        "op": "and",
+        "args": [
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "realized_vol",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": 0.005
+              }
+            ]
+          },
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "atr_pct",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": 0.03
+              }
+            ]
+          },
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "sma_pct",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": -0.05
+              }
+            ]
+          }
+        ]
+      },
+      "action": "entry-short"
+    },
+    {
+      "condition": {
+        "op": "or",
+        "args": [
+          {
+            "op": "eq",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "sma_pct",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": 0.0
+              }
+            ]
+          },
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "realized_vol",
+                "params": [150]
+              },
+              {
+                "kind": "literal",
+                "value": 0.001
+              }
+            ]
+          }
+        ]
+      },
+      "action": "exit"
+    }
+  ]
+}
@@ -0,0 +1,84 @@
+{
+  "rules": [
+    {
+      "condition": {
+        "op": "and",
+        "args": [
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "realized_vol",
+                "params": [
+                  150
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 0.007
+              }
+            ]
+          },
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "atr_pct",
+                "params": [
+                  150
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 0.0042
+              }
+            ]
+          }
+        ]
+      },
+      "action": "entry-long"
+    },
+    {
+      "condition": {
+        "op": "or",
+        "args": [
+          {
+            "op": "crossunder",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "rsi",
+                "params": [
+                  14
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 70.0
+              }
+            ]
+          },
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "atr_pct",
+                "params": [
+                  150
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 0.007
+              }
+            ]
+          }
+        ]
+      },
+      "action": "exit"
+    }
+  ]
+}
@@ -0,0 +1,120 @@
+{
+  "rules": [
+    {
+      "condition": {
+        "op": "and",
+        "args": [
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "rsi",
+                "params": [
+                  14
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 30.0
+              }
+            ]
+          },
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "atr_pct",
+                "params": [
+                  14
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 0.01
+              }
+            ]
+          },
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "macd_pct",
+                "params": [
+                  12,
+                  26,
+                  9
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": -0.005
+              }
+            ]
+          }
+        ]
+      },
+      "action": "entry-long"
+    },
+    {
+      "condition": {
+        "op": "or",
+        "args": [
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "rsi",
+                "params": [
+                  14
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 70.0
+              }
+            ]
+          },
+          {
+            "op": "lt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "atr_pct",
+                "params": [
+                  14
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 0.005
+              }
+            ]
+          },
+          {
+            "op": "gt",
+            "args": [
+              {
+                "kind": "indicator",
+                "name": "macd_pct",
+                "params": [
+                  12,
+                  26,
+                  9
+                ]
+              },
+              {
+                "kind": "literal",
+                "value": 0.005
+              }
+            ]
+          }
+        ]
+      },
+      "action": "exit"
+    }
+  ]
+}
@@ -0,0 +1,634 @@
+{
+  "run_id": "0392aa1c2d644459afa5a23f43c38ac6",
+  "run_name": "phase1-btc-100-001",
+  "n_folds": 4,
+  "top_k_requested": 10,
+  "top_k_evaluated": 10,
+  "symbol": "BTC-PERPETUAL",
+  "timeframe": "1h",
+  "start": "2018-09-01T00:00:00+00:00",
+  "end": "2026-01-01T00:00:00+00:00",
+  "ohlcv_bars": 64297,
+  "results": [
+    {
+      "genome_id": "23a24989e2ed0f84",
+      "fitness_is": 0.25047738452013774,
+      "sharpe_is": 0.5152551943136504,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.4454407113532186,
+          "sharpe": 0.940612398713799,
+          "dsr": 0.09856838950479485,
+          "dsr_pvalue": 0.9014316104952051,
+          "return": 0.12691347502077277,
+          "max_dd": 0.08467873586477132,
+          "n_trades": 50,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.33651846595831003,
+          "sharpe": 0.6297236131089199,
+          "dsr": 0.05704792862404472,
+          "dsr_pvalue": 0.9429520713759553,
+          "return": 0.16916039262594973,
+          "max_dd": 0.2420995418754207,
+          "n_trades": 61,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.08496628060413243,
+          "sharpe": -0.291593157960215,
+          "dsr": 0.006828013272159182,
+          "dsr_pvalue": 0.9931719867278408,
+          "return": -0.06496567446731383,
+          "max_dd": 0.1933746053658072,
+          "n_trades": 72,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.10029133262703777,
+          "sharpe": -0.08634860278039096,
+          "dsr": 0.01165220864726802,
+          "dsr_pvalue": 0.988347791352732,
+          "return": -0.007636913661893563,
+          "max_dd": 0.061872083556258554,
+          "n_trades": 29,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.2418041976356747,
+      "fitness_oos_min": 0.08496628060413243,
+      "fitness_oos_max": 0.4454407113532186,
+      "fitness_oos_std": 0.15416115393045135,
+      "sharpe_oos_mean": 0.2980985627705282,
+      "sharpe_oos_min": -0.291593157960215,
+      "robust_score": 0.08496628060413243
+    },
+    {
+      "genome_id": "ddda3a5d7fcf95d8",
+      "fitness_is": 0.24345612215631274,
+      "sharpe_is": 0.4859910845049414,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.38630034957174436,
+          "sharpe": 0.6292230751631145,
+          "dsr": 0.05660411470808308,
+          "dsr_pvalue": 0.9433958852919169,
+          "return": 0.0808908197444953,
+          "max_dd": 0.08123461559976199,
+          "n_trades": 44,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.3444344428619903,
+          "sharpe": 0.670768621640302,
+          "dsr": 0.06172291934756436,
+          "dsr_pvalue": 0.9382770806524356,
+          "return": 0.1769344040247678,
+          "max_dd": 0.24038922925189188,
+          "n_trades": 46,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.08496628060413243,
+          "sharpe": -0.291593157960215,
+          "dsr": 0.006828013272159182,
+          "dsr_pvalue": 0.9931719867278408,
+          "return": -0.06496567446731383,
+          "max_dd": 0.1933746053658072,
+          "n_trades": 72,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.10134367585820397,
+          "sharpe": -0.25028965416710486,
+          "dsr": 0.0070613574740692985,
+          "dsr_pvalue": 0.9929386425259307,
+          "return": -0.01793962898000656,
+          "max_dd": 0.05380115145734951,
+          "n_trades": 18,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.22926118722401778,
+      "fitness_oos_min": 0.08496628060413243,
+      "fitness_oos_max": 0.38630034957174436,
+      "fitness_oos_std": 0.13703109785336132,
+      "sharpe_oos_mean": 0.18952722116902415,
+      "sharpe_oos_min": -0.291593157960215,
+      "robust_score": 0.08496628060413243
+    },
+    {
+      "genome_id": "75fffb926a15ff30",
+      "fitness_is": 0.2317839302261713,
+      "sharpe_is": 0.5074946608465971,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.06821040478798467,
+          "sharpe": -0.42408026342979865,
+          "dsr": 0.004964388516380099,
+          "dsr_pvalue": 0.9950356114836199,
+          "return": -0.004117833402575766,
+          "max_dd": 0.01551842276077859,
+          "n_trades": 12,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.30367484453258525,
+          "sharpe": 0.7648008345130556,
+          "dsr": 0.0596988345287043,
+          "dsr_pvalue": 0.9403011654712957,
+          "return": 0.040989700605122525,
+          "max_dd": 0.036878373324561994,
+          "n_trades": 31,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.014172300674502565,
+          "sharpe": -1.443859476065376,
+          "dsr": 9.400808942425867e-05,
+          "dsr_pvalue": 0.9999059919105757,
+          "return": -0.02894431062955649,
+          "max_dd": 0.036019686142963456,
+          "n_trades": 7,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.15455497647301103,
+          "sharpe": 0.11384980407793575,
+          "dsr": 0.018924861238402986,
+          "dsr_pvalue": 0.981075138761597,
+          "return": 0.004017688385377749,
+          "max_dd": 0.034520559216801125,
+          "n_trades": 21,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.1351531316170209,
+      "fitness_oos_min": 0.014172300674502565,
+      "fitness_oos_max": 0.30367484453258525,
+      "fitness_oos_std": 0.1094231344810833,
+      "sharpe_oos_mean": -0.24732227522604583,
+      "sharpe_oos_min": -1.443859476065376,
+      "robust_score": 0.014172300674502565
+    },
+    {
+      "genome_id": "1cba64abfb67fd63",
+      "fitness_is": 0.24779915639787098,
+      "sharpe_is": 0.686744434641618,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.23623418911693914,
+          "sharpe": 0.3904002210885913,
+          "dsr": 0.03526702699157904,
+          "dsr_pvalue": 0.964732973008421,
+          "return": 0.04525398577066886,
+          "max_dd": 0.09020089585606307,
+          "n_trades": 71,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.018226313030461766,
+          "sharpe": -1.287441608051841,
+          "dsr": 0.0005995136178100196,
+          "dsr_pvalue": 0.99940048638219,
+          "return": -0.08712458661694344,
+          "max_dd": 0.08774369104277313,
+          "n_trades": 23,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.005524367416662587,
+          "sharpe": -1.78701596924334,
+          "dsr": 2.01452172810692e-05,
+          "dsr_pvalue": 0.9999798547827189,
+          "return": -0.338652173037419,
+          "max_dd": 0.3725940753872713,
+          "n_trades": 60,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0633884962634609,
+          "sharpe": -0.551177699295004,
+          "dsr": 0.003506443748025072,
+          "dsr_pvalue": 0.9964935562519749,
+          "return": -0.05484487559184925,
+          "max_dd": 0.10783012035662387,
+          "n_trades": 53,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0808433414568811,
+      "fitness_oos_min": 0.005524367416662587,
+      "fitness_oos_max": 0.23623418911693914,
+      "fitness_oos_std": 0.09225620203631806,
+      "sharpe_oos_mean": -0.8088087638753985,
+      "sharpe_oos_min": -1.78701596924334,
+      "robust_score": 0.005524367416662587
+    },
+    {
+      "genome_id": "238e481262c1594c",
+      "fitness_is": 0.2604059091373716,
+      "sharpe_is": 0.4370965264203703,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.7584616560730608,
+          "sharpe": 2.414722662010953,
+          "dsr": 0.5606058252346415,
+          "dsr_pvalue": 0.43939417476535847,
+          "return": 0.0933177386269215,
+          "max_dd": 0.023570677223421994,
+          "n_trades": 3,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.16218807661575446,
+          "sharpe": 0.012873968372421442,
+          "dsr": 0.015091968754361085,
+          "dsr_pvalue": 0.9849080312456389,
+          "return": 0.0016924869828314204,
+          "max_dd": 0.14842112455963283,
+          "n_trades": 15,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.6516189601780499,
+          "sharpe": 2.0590304959441346,
+          "dsr": 0.41524419503560983,
+          "dsr_pvalue": 0.5847558049643902,
+          "return": 0.22153641670021273,
+          "max_dd": 0.0736556977759442,
+          "n_trades": 11,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.3930671732167163,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.7584616560730608,
+      "fitness_oos_std": 0.3194405713707209,
+      "sharpe_oos_mean": 1.1216567815818772,
+      "sharpe_oos_min": 0.0,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "9d484d1d7154e3d8",
+      "fitness_is": 0.24281082171633084,
+      "sharpe_is": 0.5383459590297983,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": -0.17966776746614493,
+          "dsr": 0.010596058628339637,
+          "dsr_pvalue": 0.9894039413716603,
+          "return": -0.00014822652679680193,
+          "max_dd": 0.00043372790281119014,
+          "n_trades": 1,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.03510752668391289,
+          "sharpe": -0.9516408190993537,
+          "dsr": 2.2816330745239768e-07,
+          "dsr_pvalue": 0.9999997718366925,
+          "return": -0.022264081410075964,
+          "max_dd": 0.026510937718929387,
+          "n_trades": 3,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": -1.2370529718735044,
+          "dsr": 3.550540879152349e-10,
+          "dsr_pvalue": 0.9999999996449459,
+          "return": -0.005011777500302128,
+          "max_dd": 0.005011777500302128,
+          "n_trades": 1,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.009397695173924475,
+          "sharpe": -1.662121842926161,
+          "dsr": 6.564863915519434e-10,
+          "dsr_pvalue": 0.9999999993435136,
+          "return": -0.023018388061418094,
+          "max_dd": 0.02710730869264358,
+          "n_trades": 3,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.011126305464459342,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.03510752668391289,
+      "fitness_oos_std": 0.014367292814669608,
+      "sharpe_oos_mean": -1.0076208503412911,
+      "sharpe_oos_min": -1.662121842926161,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "e1f8b425fec3b95e",
+      "fitness_is": 0.23502719247621787,
+      "sharpe_is": 0.49764640014032885,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": -1.3160615555061568,
+          "dsr": 2.0480110912755528e-14,
+          "dsr_pvalue": 0.9999999999999796,
+          "return": -0.004197667464114874,
+          "max_dd": 0.004197667464114874,
+          "n_trades": 1,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": -0.3290153888765392,
+      "sharpe_oos_min": -1.3160615555061568,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "de4c3cce96bc6233",
+      "fitness_is": 0.22983732104518312,
+      "sharpe_is": 0.3042292766075092,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.7099634427811606,
+          "sharpe": 2.175438924887508,
+          "dsr": 0.4516683289747405,
+          "dsr_pvalue": 0.5483316710252595,
+          "return": 0.05162527264229322,
+          "max_dd": 0.013388106636764294,
+          "n_trades": 5,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.11214839774268798,
+          "sharpe": -0.09717412830030508,
+          "dsr": 0.011407831629063609,
+          "dsr_pvalue": 0.9885921683709364,
+          "return": -0.010810784900084469,
+          "max_dd": 0.146727446687109,
+          "n_trades": 17,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.48948263578942797,
+          "sharpe": 1.2111587578615337,
+          "dsr": 0.15099566068576153,
+          "dsr_pvalue": 0.8490043393142385,
+          "return": 0.09392150449517889,
+          "max_dd": 0.09248924622174504,
+          "n_trades": 17,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.32789861907831913,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.7099634427811606,
+      "fitness_oos_std": 0.28554710047880966,
+      "sharpe_oos_mean": 0.8223558886121842,
+      "sharpe_oos_min": -0.09717412830030508,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "14711c6816c009bb",
+      "fitness_is": 0.2231252597460312,
+      "sharpe_is": 0.5100989455793494,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.041279942400821015,
+          "sharpe": -0.7516739863723977,
+          "dsr": 0.0003109810948998818,
+          "dsr_pvalue": 0.9996890189051001,
+          "return": -0.002913092808893847,
+          "max_dd": 0.003336150788066372,
+          "n_trades": 3,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.010319985600205254,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.041279942400821015,
+      "fitness_oos_std": 0.017874739392934696,
+      "sharpe_oos_mean": -0.18791849659309942,
+      "sharpe_oos_min": -0.7516739863723977,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "c7f6f7c415a8c1df",
+      "fitness_is": 0.2231252597460312,
+      "sharpe_is": 0.5100989455793494,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.041279942400821015,
+          "sharpe": -0.7516739863723977,
+          "dsr": 0.0003109810948998818,
+          "dsr_pvalue": 0.9996890189051001,
+          "return": -0.002913092808893847,
+          "max_dd": 0.003336150788066372,
+          "n_trades": 3,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.010319985600205254,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.041279942400821015,
+      "fitness_oos_std": 0.017874739392934696,
+      "sharpe_oos_mean": -0.18791849659309942,
+      "sharpe_oos_min": -0.7516739863723977,
+      "robust_score": 0.0
+    }
+  ]
+}
@@ -0,0 +1,634 @@
+{
+  "run_id": "0392aa1c2d644459afa5a23f43c38ac6",
+  "run_name": "phase1-btc-100-001",
+  "n_folds": 4,
+  "top_k_requested": 10,
+  "top_k_evaluated": 10,
+  "symbol": "BTC-PERPETUAL",
+  "timeframe": "1h",
+  "start": "2018-09-01T00:00:00+00:00",
+  "end": "2026-01-01T00:00:00+00:00",
+  "ohlcv_bars": 64297,
+  "results": [
+    {
+      "genome_id": "23a24989e2ed0f84",
+      "fitness_is": 0.25047738452013774,
+      "sharpe_is": 0.5152551943136504,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.4454407113532186,
+          "sharpe": 0.940612398713799,
+          "dsr": 0.09856838950479485,
+          "dsr_pvalue": 0.9014316104952051,
+          "return": 0.12691347502077277,
+          "max_dd": 0.08467873586477132,
+          "n_trades": 50,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.33651846595831003,
+          "sharpe": 0.6297236131089199,
+          "dsr": 0.05704792862404472,
+          "dsr_pvalue": 0.9429520713759553,
+          "return": 0.16916039262594973,
+          "max_dd": 0.2420995418754207,
+          "n_trades": 61,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.08496628060413243,
+          "sharpe": -0.291593157960215,
+          "dsr": 0.006828013272159182,
+          "dsr_pvalue": 0.9931719867278408,
+          "return": -0.06496567446731383,
+          "max_dd": 0.1933746053658072,
+          "n_trades": 72,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.10029133262703777,
+          "sharpe": -0.08634860278039096,
+          "dsr": 0.01165220864726802,
+          "dsr_pvalue": 0.988347791352732,
+          "return": -0.007636913661893563,
+          "max_dd": 0.061872083556258554,
+          "n_trades": 29,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.2418041976356747,
+      "fitness_oos_min": 0.08496628060413243,
+      "fitness_oos_max": 0.4454407113532186,
+      "fitness_oos_std": 0.15416115393045135,
+      "sharpe_oos_mean": 0.2980985627705282,
+      "sharpe_oos_min": -0.291593157960215,
+      "robust_score": 0.08496628060413243
+    },
+    {
+      "genome_id": "ddda3a5d7fcf95d8",
+      "fitness_is": 0.24345612215631274,
+      "sharpe_is": 0.4859910845049414,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.38630034957174436,
+          "sharpe": 0.6292230751631145,
+          "dsr": 0.05660411470808308,
+          "dsr_pvalue": 0.9433958852919169,
+          "return": 0.0808908197444953,
+          "max_dd": 0.08123461559976199,
+          "n_trades": 44,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.3444344428619903,
+          "sharpe": 0.670768621640302,
+          "dsr": 0.06172291934756436,
+          "dsr_pvalue": 0.9382770806524356,
+          "return": 0.1769344040247678,
+          "max_dd": 0.24038922925189188,
+          "n_trades": 46,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.08496628060413243,
+          "sharpe": -0.291593157960215,
+          "dsr": 0.006828013272159182,
+          "dsr_pvalue": 0.9931719867278408,
+          "return": -0.06496567446731383,
+          "max_dd": 0.1933746053658072,
+          "n_trades": 72,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.10134367585820397,
+          "sharpe": -0.25028965416710486,
+          "dsr": 0.0070613574740692985,
+          "dsr_pvalue": 0.9929386425259307,
+          "return": -0.01793962898000656,
+          "max_dd": 0.05380115145734951,
+          "n_trades": 18,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.22926118722401778,
+      "fitness_oos_min": 0.08496628060413243,
+      "fitness_oos_max": 0.38630034957174436,
+      "fitness_oos_std": 0.13703109785336132,
+      "sharpe_oos_mean": 0.18952722116902415,
+      "sharpe_oos_min": -0.291593157960215,
+      "robust_score": 0.08496628060413243
+    },
+    {
+      "genome_id": "1cba64abfb67fd63",
+      "fitness_is": 0.24779915639787098,
+      "sharpe_is": 0.686744434641618,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.23623418911693914,
+          "sharpe": 0.3904002210885913,
+          "dsr": 0.03526702699157904,
+          "dsr_pvalue": 0.964732973008421,
+          "return": 0.04525398577066886,
+          "max_dd": 0.09020089585606307,
+          "n_trades": 71,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.018226313030461766,
+          "sharpe": -1.287441608051841,
+          "dsr": 0.0005995136178100196,
+          "dsr_pvalue": 0.99940048638219,
+          "return": -0.08712458661694344,
+          "max_dd": 0.08774369104277313,
+          "n_trades": 23,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.005524367416662587,
+          "sharpe": -1.78701596924334,
+          "dsr": 2.01452172810692e-05,
+          "dsr_pvalue": 0.9999798547827189,
+          "return": -0.338652173037419,
+          "max_dd": 0.3725940753872713,
+          "n_trades": 60,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0633884962634609,
+          "sharpe": -0.551177699295004,
+          "dsr": 0.003506443748025072,
+          "dsr_pvalue": 0.9964935562519749,
+          "return": -0.05484487559184925,
+          "max_dd": 0.10783012035662387,
+          "n_trades": 53,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0808433414568811,
+      "fitness_oos_min": 0.005524367416662587,
+      "fitness_oos_max": 0.23623418911693914,
+      "fitness_oos_std": 0.09225620203631806,
+      "sharpe_oos_mean": -0.8088087638753985,
+      "sharpe_oos_min": -1.78701596924334,
+      "robust_score": 0.005524367416662587
+    },
+    {
+      "genome_id": "238e481262c1594c",
+      "fitness_is": 0.2604059091373716,
+      "sharpe_is": 0.4370965264203703,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 2.414722662010953,
+          "dsr": 0.5606058252346415,
+          "dsr_pvalue": 0.43939417476535847,
+          "return": 0.0933177386269215,
+          "max_dd": 0.023570677223421994,
+          "n_trades": 3,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.16218807661575446,
+          "sharpe": 0.012873968372421442,
+          "dsr": 0.015091968754361085,
+          "dsr_pvalue": 0.9849080312456389,
+          "return": 0.0016924869828314204,
+          "max_dd": 0.14842112455963283,
+          "n_trades": 15,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.6516189601780499,
+          "sharpe": 2.0590304959441346,
+          "dsr": 0.41524419503560983,
+          "dsr_pvalue": 0.5847558049643902,
+          "return": 0.22153641670021273,
+          "max_dd": 0.0736556977759442,
+          "n_trades": 11,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.2034517591984511,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.6516189601780499,
+      "fitness_oos_std": 0.2670869559600687,
+      "sharpe_oos_mean": 1.1216567815818772,
+      "sharpe_oos_min": 0.0,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "9d484d1d7154e3d8",
+      "fitness_is": 0.24281082171633084,
+      "sharpe_is": 0.5383459590297983,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": -0.17966776746614493,
+          "dsr": 0.010596058628339637,
+          "dsr_pvalue": 0.9894039413716603,
+          "return": -0.00014822652679680193,
+          "max_dd": 0.00043372790281119014,
+          "n_trades": 1,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": -0.9516408190993537,
+          "dsr": 2.2816330745239768e-07,
+          "dsr_pvalue": 0.9999997718366925,
+          "return": -0.022264081410075964,
+          "max_dd": 0.026510937718929387,
+          "n_trades": 3,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": -1.2370529718735044,
+          "dsr": 3.550540879152349e-10,
+          "dsr_pvalue": 0.9999999996449459,
+          "return": -0.005011777500302128,
+          "max_dd": 0.005011777500302128,
+          "n_trades": 1,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": -1.662121842926161,
+          "dsr": 6.564863915519434e-10,
+          "dsr_pvalue": 0.9999999993435136,
+          "return": -0.023018388061418094,
+          "max_dd": 0.02710730869264358,
+          "n_trades": 3,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": -1.0076208503412911,
+      "sharpe_oos_min": -1.662121842926161,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "e1f8b425fec3b95e",
+      "fitness_is": 0.23502719247621787,
+      "sharpe_is": 0.49764640014032885,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": -1.3160615555061568,
+          "dsr": 2.0480110912755528e-14,
+          "dsr_pvalue": 0.9999999999999796,
+          "return": -0.004197667464114874,
+          "max_dd": 0.004197667464114874,
+          "n_trades": 1,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": -0.3290153888765392,
+      "sharpe_oos_min": -1.3160615555061568,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "75fffb926a15ff30",
+      "fitness_is": 0.2317839302261713,
+      "sharpe_is": 0.5074946608465971,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.06821040478798467,
+          "sharpe": -0.42408026342979865,
+          "dsr": 0.004964388516380099,
+          "dsr_pvalue": 0.9950356114836199,
+          "return": -0.004117833402575766,
+          "max_dd": 0.01551842276077859,
+          "n_trades": 12,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.30367484453258525,
+          "sharpe": 0.7648008345130556,
+          "dsr": 0.0596988345287043,
+          "dsr_pvalue": 0.9403011654712957,
+          "return": 0.040989700605122525,
+          "max_dd": 0.036878373324561994,
+          "n_trades": 31,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": -1.443859476065376,
+          "dsr": 9.400808942425867e-05,
+          "dsr_pvalue": 0.9999059919105757,
+          "return": -0.02894431062955649,
+          "max_dd": 0.036019686142963456,
+          "n_trades": 7,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.15455497647301103,
+          "sharpe": 0.11384980407793575,
+          "dsr": 0.018924861238402986,
+          "dsr_pvalue": 0.981075138761597,
+          "return": 0.004017688385377749,
+          "max_dd": 0.034520559216801125,
+          "n_trades": 21,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.13161005644839524,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.30367484453258525,
+      "fitness_oos_std": 0.11343884193961566,
+      "sharpe_oos_mean": -0.24732227522604583,
+      "sharpe_oos_min": -1.443859476065376,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "de4c3cce96bc6233",
+      "fitness_is": 0.22983732104518312,
+      "sharpe_is": 0.3042292766075092,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 2.175438924887508,
+          "dsr": 0.4516683289747405,
+          "dsr_pvalue": 0.5483316710252595,
+          "return": 0.05162527264229322,
+          "max_dd": 0.013388106636764294,
+          "n_trades": 5,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.11214839774268798,
+          "sharpe": -0.09717412830030508,
+          "dsr": 0.011407831629063609,
+          "dsr_pvalue": 0.9885921683709364,
+          "return": -0.010810784900084469,
+          "max_dd": 0.146727446687109,
+          "n_trades": 17,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.48948263578942797,
+          "sharpe": 1.2111587578615337,
+          "dsr": 0.15099566068576153,
+          "dsr_pvalue": 0.8490043393142385,
+          "return": 0.09392150449517889,
+          "max_dd": 0.09248924622174504,
+          "n_trades": 17,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.150407758383029,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.48948263578942797,
+      "fitness_oos_std": 0.20104759307710415,
+      "sharpe_oos_mean": 0.8223558886121842,
+      "sharpe_oos_min": -0.09717412830030508,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "14711c6816c009bb",
+      "fitness_is": 0.2231252597460312,
+      "sharpe_is": 0.5100989455793494,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": -0.7516739863723977,
+          "dsr": 0.0003109810948998818,
+          "dsr_pvalue": 0.9996890189051001,
+          "return": -0.002913092808893847,
+          "max_dd": 0.003336150788066372,
+          "n_trades": 3,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": -0.18791849659309942,
+      "sharpe_oos_min": -0.7516739863723977,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "c7f6f7c415a8c1df",
+      "fitness_is": 0.2231252597460312,
+      "sharpe_is": 0.5100989455793494,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": -0.7516739863723977,
+          "dsr": 0.0003109810948998818,
+          "dsr_pvalue": 0.9996890189051001,
+          "return": -0.002913092808893847,
+          "max_dd": 0.003336150788066372,
+          "n_trades": 3,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": -0.18791849659309942,
+      "sharpe_oos_min": -0.7516739863723977,
+      "robust_score": 0.0
+    }
+  ]
+}
@@ -0,0 +1,634 @@
+{
+  "run_id": "1b3ac5a72ade4eeca96f63af1c692529",
+  "run_name": "phase1-btc-100-5m-001",
+  "n_folds": 4,
+  "top_k_requested": 10,
+  "top_k_evaluated": 10,
+  "symbol": "BTC-PERPETUAL",
+  "timeframe": "5m",
+  "start": "2018-09-01T00:00:00+00:00",
+  "end": "2026-01-01T00:00:00+00:00",
+  "ohlcv_bars": 771553,
+  "results": [
+    {
+      "genome_id": "f8ca6642adf7e0cd",
+      "fitness_is": 0.18938156615467444,
+      "sharpe_is": 0.30869023048836236,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.2352965252827188,
+          "sharpe": 0.3693814228315177,
+          "dsr": 1.683506715156222e-10,
+          "dsr_pvalue": 0.9999999998316493,
+          "return": 0.04254736022830463,
+          "max_dd": 0.027161767758352023,
+          "n_trades": 22,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.225344374241631,
+          "sharpe": 0.2966596569312926,
+          "dsr": 1.0070583060973712e-10,
+          "dsr_pvalue": 0.9999999998992942,
+          "return": 0.0323635859324094,
+          "max_dd": 0.020860357367648123,
+          "n_trades": 4,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.22012851949138668,
+          "sharpe": 0.2582559733706436,
+          "dsr": 6.7095779615621785e-12,
+          "dsr_pvalue": 0.9999999999932904,
+          "return": 0.024874816913196707,
+          "max_dd": 0.01617874217143948,
+          "n_trades": 6,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.10824559047403624,
+          "sharpe": -0.19939596313642038,
+          "dsr": 4.446294307701592e-18,
+          "dsr_pvalue": 1.0,
+          "return": -0.012815761172245588,
+          "max_dd": 0.030585031985383666,
+          "n_trades": 4,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.1972537523724432,
+      "fitness_oos_min": 0.10824559047403624,
+      "fitness_oos_max": 0.2352965252827188,
+      "fitness_oos_std": 0.05167698584229592,
+      "sharpe_oos_mean": 0.18122527249925838,
+      "sharpe_oos_min": -0.19939596313642038,
+      "robust_score": 0.10824559047403624
+    },
+    {
+      "genome_id": "44e4031ccb1dc02a",
+      "fitness_is": 0.22571210498548258,
+      "sharpe_is": 0.3821051381739021,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.2750172552747618,
+          "sharpe": 0.6074693455147703,
+          "dsr": 1.1545160457503457e-26,
+          "dsr_pvalue": 1.0,
+          "return": 0.043108172266182354,
+          "max_dd": 0.001458755248422002,
+          "n_trades": 10,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.06258366543576925,
+          "dsr": 2.937538417346168e-13,
+          "dsr_pvalue": 0.9999999999997062,
+          "return": 0.005830222073507807,
+          "max_dd": 0.020860357367648123,
+          "n_trades": 2,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.18648515967821389,
+          "sharpe": 0.06128925460246632,
+          "dsr": 1.8274259811076416e-13,
+          "dsr_pvalue": 0.9999999999998173,
+          "return": 0.004414468283732154,
+          "max_dd": 0.01617874217143948,
+          "n_trades": 3,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.11537560373824392,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.2750172552747618,
+      "fitness_oos_std": 0.11954610588306011,
+      "sharpe_oos_mean": 0.18283556638825146,
+      "sharpe_oos_min": 0.0,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "662816b63c4feadf",
+      "fitness_is": 0.20828744896100454,
+      "sharpe_is": 0.20788153582834326,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.09737359146830797,
+          "dsr": 2.4288846142332667e-14,
+          "dsr_pvalue": 0.9999999999999757,
+          "return": 0.0030306913996627216,
+          "max_dd": 0.007271637996670698,
+          "n_trades": 2,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": 0.024343397867076993,
+      "sharpe_oos_min": 0.0,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "cf698463450b5daf",
+      "fitness_is": 0.20181196371847498,
+      "sharpe_is": 0.3623641433398033,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.2266655258766632,
+          "sharpe": 0.2999909794832819,
+          "dsr": 6.8141661859725965e-12,
+          "dsr_pvalue": 0.9999999999931858,
+          "return": 0.018823783888960888,
+          "max_dd": 0.01731422516010631,
+          "n_trades": 4,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.44955543499020745,
+          "dsr": 9.481303288744627e-40,
+          "dsr_pvalue": 1.0,
+          "return": 0.049849479597766866,
+          "max_dd": 0.004685656497556025,
+          "n_trades": 1,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.04410712874184496,
+          "dsr": 4.2725586842847596e-14,
+          "dsr_pvalue": 0.9999999999999573,
+          "return": 0.004410420254631564,
+          "max_dd": 0.01996463311196068,
+          "n_trades": 2,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.4450404801682897,
+          "dsr": 1.4117362531214693e-27,
+          "dsr_pvalue": 1.0,
+          "return": 0.0526088849839641,
+          "max_dd": 0.005792329453230141,
+          "n_trades": 1,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0566663814691658,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.2266655258766632,
+      "fitness_oos_std": 0.09814905178567468,
+      "sharpe_oos_mean": 0.30967350584590597,
+      "sharpe_oos_min": 0.04410712874184496,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "ddbfbef72f07d76a",
+      "fitness_is": 0.2007403940095307,
+      "sharpe_is": 0.3527941850815434,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.25816377963939535,
+          "sharpe": 0.5144509577982044,
+          "dsr": 8.645767195175167e-13,
+          "dsr_pvalue": 0.9999999999991355,
+          "return": 0.07304433130107668,
+          "max_dd": 0.01915225360494227,
+          "n_trades": 5,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.44955543499020745,
+          "dsr": 9.481303288744627e-40,
+          "dsr_pvalue": 1.0,
+          "return": 0.049849479597766866,
+          "max_dd": 0.004685656497556025,
+          "n_trades": 1,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.23195306903928886,
+          "sharpe": 0.32337815238365364,
+          "dsr": 8.912615856501225e-14,
+          "dsr_pvalue": 0.9999999999999108,
+          "return": 0.033935014244555584,
+          "max_dd": 0.010486324211451901,
+          "n_trades": 3,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.35283933132790596,
+          "dsr": 6.658494490709348e-14,
+          "dsr_pvalue": 0.9999999999999334,
+          "return": 0.05344386510725729,
+          "max_dd": 0.022762761145810856,
+          "n_trades": 1,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.12252921216967105,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.25816377963939535,
+      "fitness_oos_std": 0.12287913982320425,
+      "sharpe_oos_mean": 0.41005596912499287,
+      "sharpe_oos_min": 0.32337815238365364,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "2e49f8e8e6cec686",
+      "fitness_is": 0.19435604112910157,
+      "sharpe_is": 0.3713282033185845,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.26072350293525676,
+          "sharpe": 0.5284128848027025,
+          "dsr": 2.486680252274251e-13,
+          "dsr_pvalue": 0.9999999999997513,
+          "return": 0.05881454144506426,
+          "max_dd": 0.016516772761264658,
+          "n_trades": 5,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.44955543499020745,
+          "dsr": 9.481303288744627e-40,
+          "dsr_pvalue": 1.0,
+          "return": 0.049849479597766866,
+          "max_dd": 0.004685656497556025,
+          "n_trades": 1,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.21422619151698888,
+          "sharpe": 0.2269482258874523,
+          "dsr": 3.6571308232390715e-13,
+          "dsr_pvalue": 0.9999999999996343,
+          "return": 0.02568538444204793,
+          "max_dd": 0.019558622265454236,
+          "n_trades": 3,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.4450404801682897,
+          "dsr": 1.4117362531214693e-27,
+          "dsr_pvalue": 1.0,
+          "return": 0.0526088849839641,
+          "max_dd": 0.005792329453230141,
+          "n_trades": 1,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.11873742361306142,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.26072350293525676,
+      "fitness_oos_std": 0.11987003696674377,
+      "sharpe_oos_mean": 0.412489256462163,
+      "sharpe_oos_min": 0.2269482258874523,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "9852e8e4e480ef21",
+      "fitness_is": 0.1922948882127238,
+      "sharpe_is": 0.15925760158030436,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.1494822926244717,
+          "sharpe": 0.0804180937424291,
+          "dsr": 2.443102889370683e-16,
+          "dsr_pvalue": 0.9999999999999998,
+          "return": 0.0014913672331040573,
+          "max_dd": 0.003691144778103467,
+          "n_trades": 3,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.021783719225659744,
+          "sharpe": -1.222169028737873,
+          "dsr": 2.0819357051846864e-80,
+          "dsr_pvalue": 1.0,
+          "return": -0.018262798724970497,
+          "max_dd": 0.018262798724970497,
+          "n_trades": 16,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.042816502962532865,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.1494822926244717,
+      "fitness_oos_std": 0.06222233226062855,
+      "sharpe_oos_mean": -0.285437733748861,
+      "sharpe_oos_min": -1.222169028737873,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "16a1761c8da604e7",
+      "fitness_is": 0.19121238216295305,
+      "sharpe_is": 0.15255013590499958,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": 0.0,
+      "sharpe_oos_min": 0.0,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "add18cb7382296fd",
+      "fitness_is": 0.1879996666147147,
+      "sharpe_is": 0.28550736974923013,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.0,
+          "sharpe": 0.0,
+          "dsr": 0.0,
+          "dsr_pvalue": 1.0,
+          "return": 0.0,
+          "max_dd": 0.0,
+          "n_trades": 0,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.44955543499020745,
+          "dsr": 9.481303288744627e-40,
+          "dsr_pvalue": 1.0,
+          "return": 0.049849479597766866,
+          "max_dd": 0.004685656497556025,
+          "n_trades": 1,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.150691158813143,
+          "dsr": 4.263193477728488e-17,
+          "dsr_pvalue": 1.0,
+          "return": 0.009379645576139906,
+          "max_dd": 0.008611115582095263,
+          "n_trades": 1,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.46456569507517853,
+          "dsr": 9.733950727895419e-26,
+          "dsr_pvalue": 1.0,
+          "return": 0.056366295538783584,
+          "max_dd": 0.005792329453230141,
+          "n_trades": 1,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.0,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.0,
+      "fitness_oos_std": 0.0,
+      "sharpe_oos_mean": 0.26620307221963224,
+      "sharpe_oos_min": 0.0,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "4ee566eb5e132d8c",
+      "fitness_is": 0.18745393246665415,
+      "sharpe_is": 0.31006907324235156,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.2166522335961314,
+          "sharpe": 0.23875257586420637,
+          "dsr": 4.3199206915491695e-12,
+          "dsr_pvalue": 0.9999999999956801,
+          "return": 0.01599067972499668,
+          "max_dd": 0.01736213317014228,
+          "n_trades": 3,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:55:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.44955543499020745,
+          "dsr": 9.481303288744627e-40,
+          "dsr_pvalue": 1.0,
+          "return": 0.049849479597766866,
+          "max_dd": 0.004685656497556025,
+          "n_trades": 1,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:55:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 0.04410712874184496,
+          "dsr": 4.2725586842847596e-14,
+          "dsr_pvalue": 0.9999999999999573,
+          "return": 0.004410420254631564,
+          "max_dd": 0.01996463311196068,
+          "n_trades": 2,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:55:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 0.46456569507517853,
+          "dsr": 9.733950727895419e-26,
+          "dsr_pvalue": 1.0,
+          "return": 0.056366295538783584,
+          "max_dd": 0.005792329453230141,
+          "n_trades": 1,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:55:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.05416305839903285,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.2166522335961314,
+      "fitness_oos_std": 0.09381316904044511,
+      "sharpe_oos_mean": 0.2992452086678593,
+      "sharpe_oos_min": 0.04410712874184496,
+      "robust_score": 0.0
+    }
+  ]
+}
@@ -0,0 +1,634 @@
+{
+  "run_id": "e263651598894da688d95fda90a34a96",
+  "run_name": "phase1-extended-001",
+  "n_folds": 4,
+  "top_k_requested": 10,
+  "top_k_evaluated": 10,
+  "symbol": "BTC-PERPETUAL",
+  "timeframe": "1h",
+  "start": "2018-09-01T00:00:00+00:00",
+  "end": "2026-01-01T00:00:00+00:00",
+  "ohlcv_bars": 64297,
+  "results": [
+    {
+      "genome_id": "fe6e01eb690d3960",
+      "fitness_is": 0.3513762485888574,
+      "sharpe_is": 0.9011072752402621,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.18429629882863333,
+          "sharpe": 0.18578971215949266,
+          "dsr": 0.022696217108110216,
+          "dsr_pvalue": 0.9773037828918898,
+          "return": 0.03138679502492736,
+          "max_dd": 0.19089436189057732,
+          "n_trades": 90,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.5261711065682141,
+          "sharpe": 1.875154815895858,
+          "dsr": 0.34945842578669783,
+          "dsr_pvalue": 0.6505415742133022,
+          "return": 0.684094224950746,
+          "max_dd": 0.26051011671170043,
+          "n_trades": 108,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.29904777517998976,
+          "sharpe": 0.45027968136531676,
+          "dsr": 0.040236883094469954,
+          "dsr_pvalue": 0.9597631169055301,
+          "return": 0.16192920610625539,
+          "max_dd": 0.25615601205401484,
+          "n_trades": 87,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.06652893772008044,
+          "sharpe": -0.7859068293026578,
+          "dsr": 0.0016949251764253048,
+          "dsr_pvalue": 0.9983050748235747,
+          "return": -0.1801701961968295,
+          "max_dd": 0.3050931306970407,
+          "n_trades": 89,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.2690110295742294,
+      "fitness_oos_min": 0.06652893772008044,
+      "fitness_oos_max": 0.5261711065682141,
+      "fitness_oos_std": 0.16971232602043682,
+      "sharpe_oos_mean": 0.43132934502950243,
+      "sharpe_oos_min": -0.7859068293026578,
+      "robust_score": 0.06652893772008044
+    },
+    {
+      "genome_id": "d98739b2ba8d65e8",
+      "fitness_is": 0.3581811122056351,
+      "sharpe_is": 1.5316294902683918,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.03897301517910094,
+          "sharpe": -1.1213338499931884,
+          "dsr": 0.0007106609032094727,
+          "dsr_pvalue": 0.9992893390967905,
+          "return": -0.18358503193809717,
+          "max_dd": 0.24053109269341416,
+          "n_trades": 65,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.13868820192362147,
+          "sharpe": 0.1139203144076397,
+          "dsr": 0.018885390702584475,
+          "dsr_pvalue": 0.9811146092974156,
+          "return": 0.019694298831973045,
+          "max_dd": 0.1528666578131679,
+          "n_trades": 21,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.43784574830368517,
+          "sharpe": 1.922993791724599,
+          "dsr": 0.36463909799020594,
+          "dsr_pvalue": 0.6353609020097941,
+          "return": 0.31046814355338936,
+          "max_dd": 0.09604869735695161,
+          "n_trades": 48,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.377910610011883,
+          "sharpe": 1.3313322991701542,
+          "dsr": 0.17322722607861918,
+          "dsr_pvalue": 0.8267727739213808,
+          "return": 0.12021252899342505,
+          "max_dd": 0.04712452925993322,
+          "n_trades": 32,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.24835439385457264,
+      "fitness_oos_min": 0.03897301517910094,
+      "fitness_oos_max": 0.43784574830368517,
+      "fitness_oos_std": 0.1647414807695958,
+      "sharpe_oos_mean": 0.5617281388273011,
+      "sharpe_oos_min": -1.1213338499931884,
+      "robust_score": 0.03897301517910094
+    },
+    {
+      "genome_id": "0dd6619fdcbe37f4",
+      "fitness_is": 0.3765498201912705,
+      "sharpe_is": 0.9388498977535525,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.03550929012857722,
+          "sharpe": -1.1093436310362703,
+          "dsr": 0.0005959199839818188,
+          "dsr_pvalue": 0.9994040800160182,
+          "return": -0.24600377155172415,
+          "max_dd": 0.38950670401585935,
+          "n_trades": 122,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.049414033104104804,
+          "sharpe": -0.9995738302518242,
+          "dsr": 0.0007695523975576842,
+          "dsr_pvalue": 0.9992304476024423,
+          "return": -0.15162579158457645,
+          "max_dd": 0.21485725923368693,
+          "n_trades": 79,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.4634730710112187,
+          "sharpe": 1.2076100738049764,
+          "dsr": 0.15231462029944348,
+          "dsr_pvalue": 0.8476853797005566,
+          "return": 0.3072592298707053,
+          "max_dd": 0.15464658494822137,
+          "n_trades": 153,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.1215451856323226,
+          "sharpe": -0.22288474503734051,
+          "dsr": 0.00830898358713792,
+          "dsr_pvalue": 0.9916910164128621,
+          "return": -0.035113534418310444,
+          "max_dd": 0.17145164314561556,
+          "n_trades": 74,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.16748539496905585,
+      "fitness_oos_min": 0.03550929012857722,
+      "fitness_oos_max": 0.4634730710112187,
+      "fitness_oos_std": 0.17398113830545858,
+      "sharpe_oos_mean": -0.28104803313011467,
+      "sharpe_oos_min": -1.1093436310362703,
+      "robust_score": 0.03550929012857722
+    },
+    {
+      "genome_id": "00545b157923dc6b",
+      "fitness_is": 0.34960092770407836,
+      "sharpe_is": 0.9785144736247469,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.027180640826810924,
+          "sharpe": -1.26345444662059,
+          "dsr": 0.00030602669379661417,
+          "dsr_pvalue": 0.9996939733062034,
+          "return": -0.2411006699210636,
+          "max_dd": 0.36676673786660746,
+          "n_trades": 79,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.5208756815430423,
+          "sharpe": 1.770946043023605,
+          "dsr": 0.3128216779179025,
+          "dsr_pvalue": 0.6871783220820975,
+          "return": 0.60240576625387,
+          "max_dd": 0.2331907909757441,
+          "n_trades": 45,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.2959791903863346,
+          "sharpe": 0.3683437692731339,
+          "dsr": 0.033871432018897314,
+          "dsr_pvalue": 0.9661285679811027,
+          "return": 0.134452978611995,
+          "max_dd": 0.19964940601028408,
+          "n_trades": 58,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.15883382666451054,
+          "sharpe": 0.031160944245222175,
+          "dsr": 0.015769276637227145,
+          "dsr_pvalue": 0.9842307233627728,
+          "return": 0.006899603676653765,
+          "max_dd": 0.1947452948903118,
+          "n_trades": 36,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.2507173348551746,
+      "fitness_oos_min": 0.027180640826810924,
+      "fitness_oos_max": 0.5208756815430423,
+      "fitness_oos_std": 0.1826508968673572,
+      "sharpe_oos_mean": 0.2267490774803428,
+      "sharpe_oos_min": -1.26345444662059,
+      "robust_score": 0.027180640826810924
+    },
+    {
+      "genome_id": "eea882db55f8dd5e",
+      "fitness_is": 0.3181543783162782,
+      "sharpe_is": 1.3214317725559044,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.02689139923003529,
+          "sharpe": -1.3020990219857191,
+          "dsr": 0.00036878360260577075,
+          "dsr_pvalue": 0.9996312163973943,
+          "return": -0.22280992288117984,
+          "max_dd": 0.28735423155244094,
+          "n_trades": 69,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.1064185213723357,
+          "sharpe": -0.3110958507357913,
+          "dsr": 0.006879713400940769,
+          "dsr_pvalue": 0.9931202865990593,
+          "return": -0.057958925555868235,
+          "max_dd": 0.19529025657454543,
+          "n_trades": 33,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.37938211463073945,
+          "sharpe": 1.490827385000592,
+          "dsr": 0.2224878866287407,
+          "dsr_pvalue": 0.7775121133712593,
+          "return": 0.2563445845249124,
+          "max_dd": 0.1053935530975972,
+          "n_trades": 59,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.4175480977898098,
+          "sharpe": 1.6336803280087653,
+          "dsr": 0.260079721684657,
+          "dsr_pvalue": 0.739920278315343,
+          "return": 0.157201259896103,
+          "max_dd": 0.04639072483391873,
+          "n_trades": 43,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.23256003325573005,
+      "fitness_oos_min": 0.02689139923003529,
+      "fitness_oos_max": 0.4175480977898098,
+      "fitness_oos_std": 0.16881097094204098,
+      "sharpe_oos_mean": 0.3778282100719617,
+      "sharpe_oos_min": -1.3020990219857191,
+      "robust_score": 0.02689139923003529
+    },
+    {
+      "genome_id": "24a3e592397d4bda",
+      "fitness_is": 0.3181543783162782,
+      "sharpe_is": 1.3214317725559044,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.02689139923003529,
+          "sharpe": -1.3020990219857191,
+          "dsr": 0.00036878360260577075,
+          "dsr_pvalue": 0.9996312163973943,
+          "return": -0.22280992288117984,
+          "max_dd": 0.28735423155244094,
+          "n_trades": 69,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.1064185213723357,
+          "sharpe": -0.3110958507357913,
+          "dsr": 0.006879713400940769,
+          "dsr_pvalue": 0.9931202865990593,
+          "return": -0.057958925555868235,
+          "max_dd": 0.19529025657454543,
+          "n_trades": 33,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.37938211463073945,
+          "sharpe": 1.490827385000592,
+          "dsr": 0.2224878866287407,
+          "dsr_pvalue": 0.7775121133712593,
+          "return": 0.2563445845249124,
+          "max_dd": 0.1053935530975972,
+          "n_trades": 59,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.4175480977898098,
+          "sharpe": 1.6336803280087653,
+          "dsr": 0.260079721684657,
+          "dsr_pvalue": 0.739920278315343,
+          "return": 0.157201259896103,
+          "max_dd": 0.04639072483391873,
+          "n_trades": 43,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.23256003325573005,
+      "fitness_oos_min": 0.02689139923003529,
+      "fitness_oos_max": 0.4175480977898098,
+      "fitness_oos_std": 0.16881097094204098,
+      "sharpe_oos_mean": 0.3778282100719617,
+      "sharpe_oos_min": -1.3020990219857191,
+      "robust_score": 0.02689139923003529
+    },
+    {
+      "genome_id": "7d6636945c3fdee5",
+      "fitness_is": 0.4143633094822953,
+      "sharpe_is": 1.0518304581395579,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.024065408548587387,
+          "sharpe": -1.362275778196952,
+          "dsr": 0.0002459819911643212,
+          "dsr_pvalue": 0.9997540180088357,
+          "return": -0.27317643202118824,
+          "max_dd": 0.283710961847723,
+          "n_trades": 76,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.10523269843210308,
+          "sharpe": -0.32050989428963983,
+          "dsr": 0.006397259401103736,
+          "dsr_pvalue": 0.9936027405988963,
+          "return": -0.058925951660568465,
+          "max_dd": 0.19264026914858506,
+          "n_trades": 115,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.37331175980311637,
+          "sharpe": 0.6268880704230287,
+          "dsr": 0.05821057212982157,
+          "dsr_pvalue": 0.9417894278701784,
+          "return": 0.13285431586579133,
+          "max_dd": 0.11992643530986899,
+          "n_trades": 66,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.058003508089270274,
+          "sharpe": -0.9065054278150867,
+          "dsr": 0.0011919907667399207,
+          "dsr_pvalue": 0.9988080092332601,
+          "return": -0.11553575296643481,
+          "max_dd": 0.21946607715577302,
+          "n_trades": 108,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.14015334371826926,
+      "fitness_oos_min": 0.024065408548587387,
+      "fitness_oos_max": 0.37331175980311637,
+      "fitness_oos_std": 0.13766562991922562,
+      "sharpe_oos_mean": -0.4906007574696624,
+      "sharpe_oos_min": -1.362275778196952,
+      "robust_score": 0.024065408548587387
+    },
+    {
+      "genome_id": "94413564f2f6c03e",
+      "fitness_is": 0.4504680245893446,
+      "sharpe_is": 1.931509193896414,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.1454684156938894,
+          "sharpe": -0.1847503153026734,
+          "dsr": 0.010432618759993285,
+          "dsr_pvalue": 0.9895673812400068,
+          "return": -0.006041376713751601,
+          "max_dd": 0.028928093840880797,
+          "n_trades": 10,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.7414755036663777,
+          "dsr": 0.032908166066855475,
+          "dsr_pvalue": 0.9670918339331446,
+          "return": 0.024390796509991652,
+          "max_dd": 0.01848789755136504,
+          "n_trades": 3,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.6298210284045589,
+          "sharpe": 2.8227093941852077,
+          "dsr": 0.783195485997482,
+          "dsr_pvalue": 0.216804514002518,
+          "return": 0.1153800539734966,
+          "max_dd": 0.009172322853563938,
+          "n_trades": 13,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": -0.3705731110916588,
+          "dsr": 0.005298299731142135,
+          "dsr_pvalue": 0.9947017002688578,
+          "return": -0.01618834464316521,
+          "max_dd": 0.050704690156295765,
+          "n_trades": 8,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.1938223610246121,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.6298210284045589,
+      "fitness_oos_std": 0.2586344704657744,
+      "sharpe_oos_mean": 0.7522153678643133,
+      "sharpe_oos_min": -0.3705731110916588,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "9deb48ad9a1d6a01",
+      "fitness_is": 0.38499883511982724,
+      "sharpe_is": 1.7309660717184492,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.11865902601061347,
+          "sharpe": -0.33840465028379796,
+          "dsr": 0.006577366557928182,
+          "dsr_pvalue": 0.9934226334420718,
+          "return": -0.012941050841296264,
+          "max_dd": 0.03402877313520886,
+          "n_trades": 17,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.5508017623316874,
+          "dsr": 0.028490827818300125,
+          "dsr_pvalue": 0.9715091721816999,
+          "return": 0.01831303651843519,
+          "max_dd": 0.017308885097101046,
+          "n_trades": 7,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.0,
+          "sharpe": 2.1433702644761987,
+          "dsr": 0.4071948898247041,
+          "dsr_pvalue": 0.5928051101752959,
+          "return": 0.10827348451282881,
+          "max_dd": 0.008755897100329497,
+          "n_trades": 8,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 1.0955591683486918,
+          "dsr": 0.07909942404164044,
+          "dsr_pvalue": 0.9209005759583595,
+          "return": 0.01372945300669004,
+          "max_dd": 0.006329669977511941,
+          "n_trades": 5,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.029664756502653367,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.11865902601061347,
+      "fitness_oos_std": 0.05138086545675487,
+      "sharpe_oos_mean": 0.862831636218195,
+      "sharpe_oos_min": -0.33840465028379796,
+      "robust_score": 0.0
+    },
+    {
+      "genome_id": "1bb0f36ae3f178ae",
+      "fitness_is": 0.3289012666358953,
+      "sharpe_is": 1.2144322416340778,
+      "folds": [
+        {
+          "fold": 0,
+          "fitness": 0.03325222099260249,
+          "sharpe": -1.0706470661093082,
+          "dsr": 0.0007248785745670455,
+          "dsr_pvalue": 0.999275121425433,
+          "return": -0.09901039935604483,
+          "max_dd": 0.13711458096989376,
+          "n_trades": 27,
+          "test_start": "2022-05-02 12:00:00+00:00",
+          "test_end": "2023-04-02 08:00:00+00:00"
+        },
+        {
+          "fold": 1,
+          "fitness": 0.0,
+          "sharpe": 0.12138335573748249,
+          "dsr": 0.018713173174332784,
+          "dsr_pvalue": 0.9812868268256673,
+          "return": 0.00555346714044469,
+          "max_dd": 0.02724874190208475,
+          "n_trades": 7,
+          "test_start": "2023-04-02 09:00:00+00:00",
+          "test_end": "2024-03-02 05:00:00+00:00"
+        },
+        {
+          "fold": 2,
+          "fitness": 0.32995249837249846,
+          "sharpe": 0.9733164430116448,
+          "dsr": 0.07297295015940324,
+          "dsr_pvalue": 0.9270270498405968,
+          "return": 0.05047188141942249,
+          "max_dd": 0.02617799182077058,
+          "n_trades": 13,
+          "test_start": "2024-03-02 06:00:00+00:00",
+          "test_end": "2025-01-31 02:00:00+00:00"
+        },
+        {
+          "fold": 3,
+          "fitness": 0.0,
+          "sharpe": 1.3421850442772436,
+          "dsr": 0.05521595540810741,
+          "dsr_pvalue": 0.9447840445918926,
+          "return": 0.06121506651714692,
+          "max_dd": 0.00995276327591354,
+          "n_trades": 3,
+          "test_start": "2025-01-31 03:00:00+00:00",
+          "test_end": "2025-12-31 23:00:00+00:00"
+        }
+      ],
+      "fitness_oos_mean": 0.09080117984127524,
+      "fitness_oos_min": 0.0,
+      "fitness_oos_max": 0.32995249837249846,
+      "fitness_oos_std": 0.13873981434768828,
+      "sharpe_oos_mean": 0.34155944422926565,
+      "sharpe_oos_min": -1.0706470661093082,
+      "robust_score": 0.0
+    }
+  ]
+}
Author	SHA1	Message	Date
Adriano Dal Pastro	23b7273e71	feat(paper): ETH a tick 5m + tooling per-year/per-tick analysis scripts/run_paper_trading.py: AssetConfig ETH ora usa timeframe="5m" invece del default 1h. Il winner c04dff7086 e' stato trovato dal GA su dati 5m e a 1h la strategia perde: - ETH @ 5m (native): +359.50% cum 7y, 77% winrate, max DD/yr 19% - ETH @ 1h (precedente): -33.03% cum 7y, 67% winrate, max DD 74% BTC resta a 1h (winner 238e4812 native a 1h, +104% 7y, Sharpe 2+ in 3 anni). Nuovi script di analisi: - scripts/yearly_strategies.py: breakdown per anno (2019-2025) di 4 strategie su tick di discovery (trade/winrate/return/maxDD/Sharpe). - scripts/multi_tick_strategies.py: confronto cross-tick (5m/15m/1h) per i 2 winner correnti. Documenta la divergenza tick-paper di ETH. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-16 22:10:38 +00:00
Adriano Dal Pastro	9c871d1d86	feat(validation): WFA tooling + multi-fold results phase1-100 runs Aggiunge 2 script di analisi per validare i top-K genomi cross-fold: - scripts/analyze_btc_winners.py: per-trade dump (wins/losses/winrate/ avg_win/avg_loss/return/maxDD/Sharpe) per ogni top-K × 4 fold expanding-window WFA. Usato per identificare i winner robusti vs i lucky-shot overfit. - scripts/compare_winners.py: cross-run comparison di 5 winner candidate (BTC 1h + ETH 1h + BTC 5m + ETH 5m) sui medesimi 4 fold, con totali cumulativi. Risultati WFA freezati: - validation-btc-100-001.json: BTC 1h baseline (undertrading=10) - validation-btc-100-001-thr3.json: BTC 1h con threshold=3 (rilassato per strategie ultra-selettive) - validation-btc-100-5m-thr3.json: BTC 5m con threshold=3 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-16 21:48:55 +00:00
Adriano Dal Pastro	8b767da5e7	feat(strategy): swap paper a 238e4812 (BTC) + c04dff7086 (ETH) — phase1-100 winners Risultati phase1-{btc,eth}-100-001 (1h) + phase1-{btc,eth}-100-5m-001 (5m, 770K bars) con 100-agent GA × 10 generazioni × fitness v2 hardened + WFA 0.7 split. BTC (238e4812, meteorologist Gen 7 winner del run 1h): - IS f=0.2604, Sharpe=0.437, 25 trades - OOS: f=0.4184, Sharpe=1.296, +22.15% return fold 3, DD 7.4% - WFA 4-fold: cum_ret +31.65%, 58.6% winrate, 2 fold Sharpe 2+ ETH (c04dff7086, mean-reversion oversold del run 5m): - IS f=0.1881, OOS f=0.1789 (no overfit), Sharpe_OOS=0.171 - WFA 4-fold: cum_ret +23.83%, 77.8% winrate, 2 fold @ 100% winrate - Primo ETH winner che sopravvive hard-kill v2 in tutta la storia progetto Archive: btc_9cf506b8.json (precedente winner hardened-001), eth_facd6af85d5d.json. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-16 21:48:42 +00:00
Adriano Dal Pastro	1c0058ec3b	feat(strategy): swap BTC paper strategy a 9cf506b8 (hardened-001 robust winner) Sostituisce btc_fb63e851 (in produzione dal 13 maggio, net +$9.9k su 7y BTC, Sharpe medio -0.20) con btc_9cf506b8, vincitore del run phase1-hardened-001 (K=50 G=15 7y, fitness-v2 con hard-kill su fees_eat_alpha + negative_net_pnl). Performance comparate su 7y BTC continui: - btc_fb63e851: Net +$9,951 Sharpe medio -0.20 4/8 anni positivi - btc_9cf506b8: Net +$30,538 Sharpe medio +0.31 5/8 anni positivi Best year 2021: +$18,938 (vs prod +$6,835) Best Sharpe annuale: 2023 +1.27 (vs prod 2024 +1.16) Zero adversarial findings su 7y continui. Performance cross-asset (test su ETH 6.8y): - btc_fb63e851 su ETH: -$120 (pseudo-flat, nessun segnale) - btc_9cf506b8 su ETH: +$2,059 (4/7 anni positivi) La strategia generalizza out-of-asset, indicatore di robustezza non-ETH-overfit. DSL: physicist style, lookback 150 bar (~6 giorni). Triple-AND condition su realized_vol + atr_pct + sma_pct per entry (long o short). Exit su sma_pct=0 o vol collasso (<0.1%). Selettivo: 502 trade in 7y = 1.4 trade/giorno medio. btc_fb63e851 archiviato in strategies/archive/ per consultazione futura. Il glob loader `btc_*.json` e' non-ricorsivo, quindi archive/ non viene caricato automaticamente. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-16 13:21:37 +00:00
Adriano Dal Pastro	220e510d5e	fix(orchestrator): escludi prompt_library dal config_dict JSON PromptLibrary e' un frozen dataclass non JSON-serializable. Quando passata in cfg.prompt_library e poi spread via **cfg.__dict__ in config_dict, faceva fallire repo.create_run() con TypeError al primo run dopo refactor. Fix: filtra cfg.__dict__ escludendo prompt_library, e salva separatamente la lista degli stili (prompt_library_styles) per reproducibility. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-16 13:21:08 +00:00
Adriano Dal Pastro	9742df3a1f	fix(fitness): hardening anti-overfit post-7y incident — 3 correzioni Incident: extended run elite (Sharpe IS 1.93) net-negativo su 7y continuous (fees=101% del gross). Multi-fold validation NON sufficiente: ogni fold restarta equity, mascherando accumulo fees compound. Correzioni: 1) Default --start esteso a 2018-09-01 (7.3 anni) - Copre bear 2018-19, halving 2020, COVID, ATH 2021, winter 2022, ETF rally 2024, regime corrente. - Una finestra corta (2y) lasciava il GA libero di overfit single-regime. 2) fees_eat_alpha promosso a hard-kill in fitness v2 - Da soft-penalty 0.4x a hard-kill 0 fitness. - Una strategia con fees > 50% del gross non e' recuperabile via selection: il prodotto del GA non puo' deployare con quel cost burden. 3) Nuovo finding negative_net_pnl (HIGH, hard-kill) - Fires quando sum(trade.net_pnl) < 0 sul training window. - Cattura: gross negativo (no edge direzionale) E gross positivo ma fees > gross (edge insufficiente). - Sintesi del net-after-fees su finestra continua come "deal-breaker" non negoziabile via soft penalty. CLI: - --fitness-hard-kill <csv> per override esplicito. - Default v2: no_trades,degenerate,undertrading,fees_eat_alpha,negative_net_pnl. Test: - 252 pass (251 + 1 nuovo: test_negative_net_pnl_fires_on_negative_gross). - Test e2e WFA aggiornato: passa fitness_hard_kill_findings=('no_trades',) perche' il fixture sintetico non produce strategie profittevoli. - test_no_findings_on_reasonable_strategy rinominato: test_rsi_mean_reversion_loses_money_on_synthetic_data (riflette semantica reale: RSI mean-rev su synthetic ohlcv ha net negativo). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-16 11:07:40 +00:00
Adriano Dal Pastro	21b5cf1eae	feat(validation): add validation data for phase1-extended-001 with detailed results	2026-05-16 10:55:35 +00:00
Adriano Dal Pastro	a29748e3d8	perf(backtest): vectorize engine + parallel LLM propose + multi-fold validation tool - backtest/engine.py: state machine numpy invece di pd.iterrows() - 16.8x speedup su 2y (470ms -> 28ms), 11.3x su 7y (1744ms -> 154ms) - 7 parity test parametrici vs reference iterrows assicurano equivalenza - orchestrator/run.py + run_phase1.py: --llm-concurrency N - ThreadPoolExecutor parallelizza hypothesis_agent.propose() per generazione - 5-8x speedup wall time GA loop (OpenRouter qwen-2.5 regge 6-10 concorrenti) - default 1 = backward-compat sequenziale - scripts/validate_run.py: validation multi-fold standalone - prende run_id + top-K + N-folds expanding-window su dataset esteso (7y) - rivela overfitter mascherati da fitness IS alta (vedi phase1-extended-001: elite IS Sharpe 1.93 collassa OOS a -1.00) - ranking per robust_score = min(fitness_oos) su tutti i fold Test 250/250 pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-16 10:53:48 +00:00
Adriano Dal Pastro	fa11cca2bc	docs(readme): allinea a stato finale post-refactor sessione 2026-05-15 Aggiornato per riflettere i 12 commit di refactor della sessione: - Layout: 3 servizi Docker (strategy-crypto-paper + strategy-crypto-gui + multi-swarm-core-gui), GUI split (core GA / strategy paper) - Test: 238 attesi (era 186) - Architettura prompt: SYSTEM compositor v3.2 documentato (9 sezioni iniettate da scaffold core + contenuto strategy via prompts.json) - Grammar: 8 indicatori (5 base + 3 _pct: atr_pct, sma_pct, macd_pct) - Input MarketSummary: 14 metriche (5 base + 4 regime + 5 geometria) - 7 stili cognitive in prompts.json (era 6, +military_strategist e psychologist), con focus_metrics + lookback consigliato + archetipo - Deploy: 2 URL distinti (/multi_swarm_core_gui + /strategy_crypto_gui), Traefik replacepathregex (NON stripprefix per evitare doppio root_path) - Note operative: chown 1000:1000 anche per strategies/ post git mv - 3 regression guard test ASCII/archetype/lookback documentati Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-15 22:05:06 +00:00
Adriano Dal Pastro	6526c6e6e3	refactor(prompts): strategy_crypto v3.2 — consolidamento + 3 invariant tests Patch di consolidamento post-diagnosi v3.1, divisa in REGRESSIONI ripristinate + MIGLIORAMENTI opportunistici + INVARIANTI permanenti. Regressioni v3.0 -> v3.1 ripristinate: 1. ASCII-strict: caratteri circa-uguale (U+2248) sostituiti con 'vicino a' in 3 stili (historian, engineer, psychologist). Critico per encoding robustness. 2. 'Archetipo dominante: <metafora>': chiusa identitaria reintrodotta in tutte le 7 directive. Ancora semantica resistente a mutate_prompt_llm. 3. 'Lookback consigliato: X-Y barre': hint range differenziato per stile (physicist 150-300, biologist 80-200, historian 200-500, meteorologist 50-150, engineer 60-120, military 100-200, psychologist 50-120) per orientare il parametro evoluto lookback_window del genoma. Miglioramenti opportunistici: 4. Voce attiva rinforzata: +verbi generativi ('costruisci', 'combina', 'cattura', 'diagnostica', 'preferisci') in tutte le directive 5. anti_patterns 5 -> 7 voci: aggiunti (6) chattering crossover same-type same-lookback, (7) soglie hard senza isteresi entry/exit 6. output_priorities 4 -> 5 voci: aggiunta in cima (#1) 'coerenza con lente cognitiva' (fondamento del design swarm) 7. domain_warnings: +frase 'seasonality > 0 non significa significativa, gate a 0.05' 8. NEW _design_invariants metadata: documenta gli invarianti che future versioni DEVONO preservare (utile per chi edita + mutate_prompt_llm) NEW invariant tests (regression guards permanenti): - test_strategy_crypto_directives_ascii_safe - test_strategy_crypto_directives_have_archetype_marker - test_strategy_crypto_directives_have_lookback_hint Statistiche v3.2: - directive: 800-950 char (era 545-614 in v3.1, troppo snellite) - physicist 890, biologist 867, historian 887, meteorologist 884, engineer 904, military_strategist 898, psychologist 909 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-15 21:54:13 +00:00
Adriano Dal Pastro	ccf9f7a33c	refactor(prompts): strategy_crypto v3.1 — pulizia post-diagnosi Pulizia contenuto prompts.json risolvendo le 9 debolezze identificate in diagnosi v3.0 -> v3.1: - agent_role: + frase swarm awareness ("la diversita e' asset critico, esplora territori meno ovvi") - pattern_guidance: riscritto astratto (rimossi tutti i nomi di indicatori; SMA(short)>SMA(long), RSI>70, ecc -> "trend strutturale", "compressione vol", "mean reversion strutturale"). Il GA scopre l'incarnazione, le directive sono BIAS non template. - domain_warnings: riformulato come "NON assumere" (rimosso hint inferenziale su funding rate che avrebbe indotto hallucination) - directive: trimmate tutte sotto 900 char (era 922-975) - focus_metrics: standardizzati a 4 per stile (era 4-5, asimmetria estetica); rimosse ridondanze: * historian: rimosso autocorr_baseline (gia' visibile in Regime recente -> usare solo autocorr_recent come delta proxy) * psychologist: kurt/skew (gia' in base statistics) sostituiti con autocorr_recent + spectral_entropy (piu informativi) - NEW anti_patterns: lista esplicita (5 voci) per ridurre overfitting nel design della strategia (no > 4 AND, no singolo evento, no correlazione=causalita, no stazionarieta perfetta, no temporal features con seasonality < 0.05) - NEW output_priorities: trade-off espliciti (robustezza > ottimalita, semplicita > complessita, selettivita > attivita) - NEW _focus_metrics_design metadata: documenta l'intento delle scelte Effetto atteso al prossimo Phase 1/2 run: ipotesi piu robuste (meno overfitting cross-regime), strategie piu semplici (no 5+ AND), maggior diversita di esplorazione (swarm awareness nel role), zero inferenze su feature non accessibili. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-15 21:40:58 +00:00
Adriano Dal Pastro	19a3592a20	feat(prompt): PromptLibrary v3.1 — anti_patterns + output_priorities Estende il compositor del SYSTEM con 2 sezioni opzionali iniettate DOPO i VINCOLI core e PRIMA dell'EXAMPLE: ANTI-PATTERN DA EVITARE: lista esplicita di cose da evitare (overfitting, correlazione=causalita, > 4 AND, singolo evento estremo, ecc.) PRIORITA' DI OUTPUT: trade-off come "robustezza > ottimalita su singolo regime", "semplicita > complessita raffinata", "selettivita > attivita" Razionale: ridurre la varianza non-utile nelle strategie generate quando il LLM affronta trade-off, e prevenire overfitting nel design. Entrambi i campi sono opzionali (skip se "") -> backward-compatible con prompts.json v3.0. PromptLibrary v3.1: +2 fields top-level (default ""). _build_system_prompt: 2 sezioni condizionali post-VINCOLI. Test: +3 unit (compositor inject/skip + from_json parsing). Tot: 235 test pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-15 21:38:35 +00:00
Adriano Dal Pastro	5202eb517b	refactor(prompt): SYSTEM_TEMPLATE diventa compositor — strategy controlla agent_role/pattern_guidance/instruction Refactor architetturale: il prompt non e' piu' un template hardcoded ma viene COMPOSTO at-runtime da scaffold core + contenuto strategy. CORE scaffold (universal, fisso): - Grammar JSON spec (legato al protocol/compiler) - UNITA' regole semantiche - VINCOLI del validator - Esempio output STRATEGY content (tunable in prompts.json): - agent_role: "Sei un agente generatore di ipotesi ... [crypto/forex/equity]" - pattern_guidance: sezione di archetipi tecnici, ora crypto-specific - instruction: frase finale del USER ("Genera una strategia ... [crypto]") - domain_warnings: NEW opzionale, per disclaimer di dominio (es. crypto 24/7) Implementazione: - PromptLibrary v3.0: 4 nuovi campi top-level (agent_role, pattern_guidance, instruction, domain_warnings), parsati da prompts.json, default fallback in default() - hypothesis.py: SYSTEM_TEMPLATE constant rimossa, sostituita da _build_system_prompt(lib, genome) che compone scaffold + content - USER_TEMPLATE: ultima riga ora ha placeholder {instruction} - prompts.json v3.0 in strategy_crypto: agent_role + pattern_guidance + instruction + domain_warnings popolati con flavor crypto-specific Pattern: "core = framework, strategy = contenuto". Per future strategie (forex, equity), basta creare un prompts.json con flavor diverso, zero modifiche al core. Universal scaffold (grammar, vincoli, units) e' condiviso e mantiene la garanzia di parse/validate. Test: +5 unit (compositor + PromptLibrary fields). Tot: 232 test pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-15 21:18:54 +00:00
Adriano Dal Pastro	898b24b6a3	fix(orchestrator): definisci prompt_library PRIMA di istanziare HypothesisAgent Bug introdotto in `b6f48e4`: HypothesisAgent(prompt_library=prompt_library) era chiamato a riga 109, ma prompt_library veniva definito a riga 123 -> NameError a runtime quando run_phase1 viene eseguito. Spostato il blocco di setup prompt_library + set_cognitive_styles PRIMA della istanziazione di HypothesisAgent. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-15 21:05:39 +00:00
Adriano Dal Pastro	b6f48e46fc	feat(agents): 5 metriche geometrico-frattali + style-aware focus block Aggiunge al MarketSummary 5 metriche regime-aware: - efficiency_ratio (Kaufman): discriminatore trending/ranging - tail_index_left/right (Hill): pesantezza code, robust vs kurtosis - structural_uptrend (HH/HL Dow-style): trend strutturale senza lag MA - compression_ratio: vol coil pre-breakout - spectral_entropy + dominant_cycle (gated): struttura ciclica nel FFT Architettura "Style-aware focus, no filtering": - Tutte le 5 metriche restano visibili a tutti gli stili (landscape GA smooth) - prompts.json v2.2: ogni stile dichiara focus_metrics: list[str] - USER_TEMPLATE renderizza "Focus per la tua lente: ..." con i valori prioritari - Mutation cognitive_style preserva accesso a tutte le metriche (no discontinuita) PromptLibrary esteso con focus field (parsato da JSON entry styles). HypothesisAgent accetta prompt_library nel costruttore; orchestrator lo passa. 7 directive aggiornate per interpretare i 5 nuovi input attraverso la lente: - physicist: efficiency_ratio + dominant_cycle (modi armonici) - engineer: efficiency_ratio < 0.2 = no signal - psychologist: tail_left/right = paura/euforia ricorrente - ecc. Test: +19 unit (metriche + focus rendering), +smoke MarketSummary. Tot: 216 test pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-15 21:04:06 +00:00