research(alt): sweep 104 strategie alternative su Deribit (153 agenti) + marginal scorer
Ondata di ricerca onesta a largo spettro su BTC/ETH+DVOL certificati: 104 ipotesi distinte (11 famiglie), un agente-finder per ipotesi, verifica avversariale a 3 scettici sui promettenti, sintesi (153 agenti totali). Esito: NIENTE di nuovo regge -> conferma del soffitto strutturale ~1.3 BTC/ETH-direzionale; lo stack TP01+XS01+VRP01 resta imbattuto. - altlib.py: harness condiviso vettoriale leak-free (eval_weights/study_weights, fee-sweep, both-asset + hold-out 2025+). Riproduce i numeri canonici di TP01. - MARGINAL SCORER (study_marginal/marginal_vs_tp01): Sharpe INCREMENTALE vs baseline TP01 (corr, blend uplift OOS, alpha residua) + jackknife OOS (clean-year + drop-best-month). earns_slot = abs!=FAIL & ADDS & robust_oos. Smaschera gli overlay su TSMOM con PASS assoluti fasulli (CMB04, VOL11, ...) e il falso positivo KAMA (ADDS ma muore al jackknife). - runs/*.py (104) script riproducibili per ipotesi; wf_altstrat.js workflow. - Verdetto: 0 candidati deployabili; 2 LEAD fragili (VOL08, STA05_LS) da forward-monitor. - test_marginal_scorer.py blocca baseline + invarianti. Suite: 32 verde. Diario: docs/diary/2026-06-20-alt-strategies-100agent-sweep.md Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Adriano Dal Pastro
@@ -0,0 +1,153 @@
|
||||
"""VOL07 — DVOL spike contrarian long (capitulation timing).
|
||||
|
||||
HYPOTHESIS: When DVOL > 90th expanding percentile (fear/capitulation), buy at close,
|
||||
hold ~1 week (max_bars). The idea: implied vol spikes coincide with panic bottoms,
|
||||
and the subsequent reversion offers a contrarian long edge.
|
||||
|
||||
Signals style (discrete entry/exit), 1d only.
|
||||
DVOL history starts 2021-03, so the full period is reduced to ~5 years.
|
||||
|
||||
Small grid:
|
||||
- dvol_pct threshold: 85th or 90th expanding percentile
|
||||
- max_bars (hold period): 5 or 7 days
|
||||
Total: 2 x 2 = 4 configs x 1 TF = 4 backtests.
|
||||
Best config selected by min(BTC holdout sharpe, ETH holdout sharpe).
|
||||
"""
|
||||
import sys
|
||||
sys.path.insert(0, "/opt/docker/PythagorasGoal/scripts/research/alt")
|
||||
import altlib as al
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
|
||||
TFS = ("1d",)
|
||||
|
||||
def make_entries(dvol_pct_threshold: float, max_bars: int, cooldown: int = 3):
|
||||
"""
|
||||
Entry: when DVOL crosses above the expanding `dvol_pct_threshold`-th percentile
|
||||
(i.e., DVOL[i] > expanding_pct and DVOL[i-1] <= expanding_pct — fresh spike).
|
||||
No TP/SL — exit by max_bars only.
|
||||
Cooldown: no new entry within `cooldown` bars of a previous entry.
|
||||
"""
|
||||
def entries_fn(df: pd.DataFrame):
|
||||
dv = al.dvol(df, "BTC") # will be overridden per-asset below — but we need asset
|
||||
|
||||
# This placeholder is overridden by the per-asset wrapper in run()
|
||||
return _compute_entries(df, dv, dvol_pct_threshold, max_bars, cooldown)
|
||||
|
||||
return entries_fn
|
||||
|
||||
|
||||
def _compute_entries(df: pd.DataFrame, dv: np.ndarray, dvol_pct_threshold: float,
|
||||
max_bars: int, cooldown: int):
|
||||
n = len(df)
|
||||
entries = [None] * n
|
||||
|
||||
# Expanding percentile of DVOL (causal — uses only data up to i)
|
||||
# To avoid bias: require min 60 observations before triggering
|
||||
min_obs = 60
|
||||
last_entry_bar = -999
|
||||
|
||||
dvol_series = pd.Series(dv)
|
||||
|
||||
for i in range(min_obs, n):
|
||||
if np.isnan(dv[i]) or np.isnan(dv[i - 1]):
|
||||
continue
|
||||
|
||||
# Expanding pct up to i (inclusive, causal)
|
||||
hist = dvol_series.iloc[:i + 1].dropna()
|
||||
if len(hist) < min_obs:
|
||||
continue
|
||||
threshold = float(np.percentile(hist.values, dvol_pct_threshold))
|
||||
|
||||
# Fresh spike: DVOL crosses above threshold
|
||||
prev_hist = dvol_series.iloc[:i].dropna()
|
||||
prev_threshold = float(np.percentile(prev_hist.values, dvol_pct_threshold)) if len(prev_hist) >= min_obs else np.nan
|
||||
|
||||
if np.isnan(prev_threshold):
|
||||
continue
|
||||
|
||||
crossed_up = (dv[i] > threshold) and (dv[i - 1] <= prev_threshold)
|
||||
|
||||
if crossed_up and (i - last_entry_bar >= cooldown):
|
||||
entries[i] = {"dir": +1, "tp": None, "sl": None, "max_bars": max_bars}
|
||||
last_entry_bar = i
|
||||
|
||||
return entries
|
||||
|
||||
|
||||
def make_entries_per_asset(asset: str, dvol_pct_threshold: float, max_bars: int, cooldown: int = 3):
|
||||
"""Per-asset wrapper: uses the correct DVOL for each asset."""
|
||||
def entries_fn(df: pd.DataFrame):
|
||||
dv = al.dvol(df, asset)
|
||||
return _compute_entries(df, dv, dvol_pct_threshold, max_bars, cooldown)
|
||||
return entries_fn
|
||||
|
||||
|
||||
# Grid
|
||||
CONFIGS = [
|
||||
{"dvol_pct": 85, "max_bars": 5},
|
||||
{"dvol_pct": 85, "max_bars": 7},
|
||||
{"dvol_pct": 90, "max_bars": 5},
|
||||
{"dvol_pct": 90, "max_bars": 7},
|
||||
]
|
||||
|
||||
best_rep = None
|
||||
best_score = -np.inf
|
||||
|
||||
for cfg in CONFIGS:
|
||||
name = f"VOL07_p{cfg['dvol_pct']}_h{cfg['max_bars']}"
|
||||
print(f"\n--- Config: pct={cfg['dvol_pct']} max_bars={cfg['max_bars']} ---")
|
||||
|
||||
# We need per-asset entries — study_signals calls entries_fn(df) without knowing asset.
|
||||
# Workaround: create a closure that wraps per-asset logic by detecting via df length/dates.
|
||||
# Better: run each asset separately and build the report manually.
|
||||
|
||||
cells = []
|
||||
tf = "1d"
|
||||
per_asset = {}
|
||||
fee_ok_all = True
|
||||
|
||||
for a in ("BTC", "ETH"):
|
||||
df = al.get(a, tf)
|
||||
ent_fn = make_entries_per_asset(a, cfg["dvol_pct"], cfg["max_bars"])
|
||||
ent = ent_fn(df)
|
||||
n_entries = sum(1 for e in ent if e is not None)
|
||||
print(f" {a}: {n_entries} entries")
|
||||
|
||||
base = al.eval_signals(df, ent, fee_rt=2 * al.FEE_SIDE, leverage=1.0, asset=a, tf=tf)
|
||||
sweep = {
|
||||
f"{2*f*100:.2f}%RT": al.eval_signals(df, ent, fee_rt=2 * f, leverage=1.0)["full"]["sharpe"]
|
||||
for f in al.FEE_SWEEP
|
||||
}
|
||||
fee_ok = sweep.get("0.20%RT", -9) > 0
|
||||
fee_ok_all = fee_ok_all and fee_ok
|
||||
per_asset[a] = dict(
|
||||
full=base["full"], holdout=base["holdout"],
|
||||
n_trades=base["n_trades"], win_rate=base["win_rate"],
|
||||
fee_sweep=sweep, yearly=base["yearly"]
|
||||
)
|
||||
|
||||
min_full = min(per_asset[a]["full"]["sharpe"] for a in ("BTC", "ETH"))
|
||||
min_hold = min(per_asset[a]["holdout"].get("sharpe", 0.0) for a in ("BTC", "ETH"))
|
||||
cell = dict(
|
||||
tf=tf, per_asset=per_asset,
|
||||
min_asset_full_sharpe=round(min_full, 3),
|
||||
min_asset_holdout_sharpe=round(min_hold, 3),
|
||||
full_sharpe=round(np.mean([per_asset[a]["full"]["sharpe"] for a in ("BTC", "ETH")]), 3),
|
||||
fee_survives=fee_ok_all
|
||||
)
|
||||
cells.append(cell)
|
||||
|
||||
# Build a verdict-compatible report
|
||||
rep = dict(name=name, kind="signals", cells=cells, verdict=al._verdict(cells))
|
||||
print(al.fmt(rep))
|
||||
print("JSON:", al.as_json(rep))
|
||||
|
||||
score = min_hold
|
||||
if score > best_score:
|
||||
best_score = score
|
||||
best_rep = rep
|
||||
|
||||
print("\n\n=== BEST CONFIG ===")
|
||||
print(al.fmt(best_rep))
|
||||
print("JSON:", al.as_json(best_rep))
|
||||
Reference in New Issue
Block a user