feat(fade): swap filtro live hurst->trend_max 3.0 (gate PORT06 sul path live)

Punto 7 roadmap sweep. Gate trendmax_port06_impact.py (engine exit16_port06_impact
riusato, parita' canonica 1.00000; maschera hurst IDENTICA al live via
fade_base.hurst_skip_mask; PORT06 pesi cap, path live EXIT-16):

- CANDIDATO (hurst+trend) BOCCIATO: over-filtering (FULL Sharpe 7.23->7.11,
  meta' dei trade) nonostante DD 2.68->2.06.
- SCOPERTA: il loss-guard Hurst e' ridondante-DANNOSO post-EXIT-16 (NESSUNO
  batte LIVE: FULL Sh 8.07 vs 7.23). EXIT-16 ha eliminato i wick-stop che hurst
  evitava -> gli ingressi saltati (66% delle barre) sono tornati vincenti.
  Il test che promosse hurst (2026-06-02) era sull'engine PRE-EXIT-16.
- TREND-ONLY domina LIVE su tutte le metriche (FULL Sh 7.89 DD 2.46, OOS Sh 9.91
  DD 1.20) ed e' la config che la ricerca EXIT-16 aveva davvero promosso (entries
  trend-filtrate, no hurst) mai eseguita dal live. Plateau 2.5/3.0/3.5 robusto.

Decisione (utente): SWAP. _defs.py: trend_max=3.0 + ema_long=200 nelle 6 fade,
hurst_max rimosso (hurst_skip_mask resta in fade_base). hourly_report: monitor
stop-rate per epoca PRE->HURST->TREND (verdetto a n>=30). CLAUDE.md aggiornato
(paragrafo hurst marcato storico). Diario docs/diary/2026-06-07-trendmax-gate.md.

Lezione: ri-gateare ogni filtro quando cambia l'exit engine.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Adriano Dal Pastro
2026-06-07 10:29:57 +00:00
parent 11ace196c7
commit e3bb622b90
5 changed files with 370 additions and 19 deletions
+259
View File
@@ -0,0 +1,259 @@
"""GATE PORT06: trend_max=3.0/ema_long=200 sulle 6 fade LIVE (improvement-sweep punto 7).
Contesto. Il backtest canonico (build_everything) applica il filtro trend alle fade,
ma in produzione le SleeveSpec di _defs.py NON lo passano -> le fade live fadano anche
i trend/crolli estesi (es. MR01/MR02_ETH long ripetuti nel crash ETH del 2026-06-05).
Anche il test PORT06 del loss-guard Hurst (fade_lossguard_port_test) fu misurato su
entries GIA' trend-filtrate -> la config live attuale (hurst SENZA trend) non e' mai
stata gateata. hurst e trend si sovrappongono (entrambi tagliano il regime trending):
il delta marginale va misurato sul path live, non assunto dai numeri canonici.
Confronto, a livello PORT06 (stessa matematica pesi cap di Portfolio.backtest):
LIVE = fade con hurst_max=0.55 + EXIT-16 (sl_confirm 0.5 ATR), trend OFF [prod oggi]
CANDIDATO = LIVE + trend_max=3.0 / ema_long=200 [proposta]
TREND-ONLY= EXIT-16 + trend, senza hurst [diagnostica overlap]
Parita' preliminare: il replay (mode=orig, trend ON, no hurst, no EXIT-16) deve
riprodurre le equity fade canoniche (come exit16_port06_impact).
La maschera Hurst e' quella ESATTA del live: fade_base.hurst_skip_mask (close-only,
window=100, step=6) — non la cache di regime_lab.
GATE (sweep): PROMOSSO se OOS Sharpe non peggiora (>= base-0.02) E il DD scende,
e in FULL non degrada.
uv run python scripts/analysis/trendmax_port06_impact.py
"""
from __future__ import annotations
import sys
from pathlib import Path
import numpy as np
import pandas as pd
PROJECT_ROOT = Path(__file__).resolve().parents[2]
sys.path.insert(0, str(PROJECT_ROOT))
from src.data.downloader import load_data
from src.strategies.fade_base import hurst_skip_mask
from scripts.analysis.strategy_research import atr
from scripts.analysis.risk_management import strats_for, FEE_RT, LEV, POS, INIT
from scripts.analysis.combine_portfolio import (
_norm, IDX, port_returns, metrics, SPLIT, OOS_DATE,
)
from scripts.portfolios._defs import PORTFOLIOS
from src.portfolio import weighting as W
BUFFER = 0.5 # EXIT-16 close-confirm (come in produzione)
HURST_MAX = 0.55 # loss-guard live
TREND_MAX = 3.0
EMA_LONG = 200
def build_trades_variant(ents, df, mode, trend_max, hurst_mask=None,
buffer=BUFFER, lev=LEV, fee_rt=FEE_RT, ema_long=EMA_LONG):
"""Engine di exit16_port06_impact.build_trades_variant + skip Hurst opzionale.
mode="orig" : SL intrabar al livello (SL prima del TP) == canonico.
mode="exit16" : SL intrabar OFF; TP intrabar al livello (priorita' nel bar);
SL solo se il CLOSE sfonda sl0 -/+ buffer*ATR14[j], fill a close[j].
trend_max : None = filtro OFF (live attuale); 3.0 = candidato.
hurst_mask : bool[i]=True -> salta l'ingresso (loss-guard live).
"""
h, l, c = df["high"].values, df["low"].values, df["close"].values
n = len(c)
a = atr(df, 14)
el = pd.Series(c).ewm(span=ema_long, adjust=False).mean().values
fee = fee_rt * lev
out = []
last = -1
for e in ents:
i, d = e["i"], e["d"]
if i <= last or i + 1 >= n:
continue
if hurst_mask is not None and hurst_mask[i]:
continue
if trend_max is not None and a[i] and abs(c[i] - el[i]) / a[i] > trend_max:
continue
entry = c[i]
tp, sl0, mb = e["tp"], e["sl"], e["max_bars"]
exit_p = c[min(i + mb, n - 1)]
j = min(i + mb, n - 1)
for k in range(1, mb + 1):
j = i + k
if j >= n:
exit_p = c[n - 1]
break
if mode == "orig":
hs = (d == 1 and l[j] <= sl0) or (d == -1 and h[j] >= sl0)
ht = (d == 1 and h[j] >= tp) or (d == -1 and l[j] <= tp)
if hs:
exit_p = sl0
break
if ht:
exit_p = tp
break
if k == mb:
exit_p = c[j]
else: # exit16
ht = (d == 1 and h[j] >= tp) or (d == -1 and l[j] <= tp)
if ht:
exit_p = tp
break
aj = a[j] if np.isfinite(a[j]) else 0.0
confirm = (d == 1 and c[j] < sl0 - buffer * aj) or \
(d == -1 and c[j] > sl0 + buffer * aj)
if confirm:
exit_p = c[j]
break
if k == mb:
exit_p = c[j]
ret = (exit_p - entry) / entry * d * lev - fee
out.append((i, j, ret))
last = j
return out
def equity_from_trades(df, trades):
ts = pd.to_datetime(df["timestamp"], unit="ms", utc=True)
n = len(df)
eq = np.full(n, INIT, dtype=float)
cap = INIT
for i, j, ret in sorted(trades, key=lambda t: t[1]):
cap = max(cap + cap * POS * ret, 10.0)
eq[j:] = cap
s = pd.Series(eq, index=ts).resample("1D").last().reindex(IDX).ffill().bfill()
return _norm(s)
def port_metrics(members: dict[str, pd.Series], p):
ids = p.sleeve_ids
dr = pd.DataFrame({i: members[i].pct_change().fillna(0.0) for i in ids})
w = W.weight_vector(p.weighting, ids, dr, weights=p.weights,
caps=p.caps, clusters=p.clusters, lookback=p.vol_lookback)
drp = port_returns({i: members[i] for i in ids}, w)
return metrics(drp), metrics(drp, lo=SPLIT)
def _dd(s):
pk = s.cummax()
return float(((pk - s) / pk).max() * 100)
def main():
p = PORTFOLIOS["PORT06"]
fade_ids = [s.sid for s in p.sleeves if s.sid.startswith("MR")]
print("=" * 100)
print(" GATE PORT06 — trend_max=3.0/ema200 sulle fade LIVE (hurst 0.55 + EXIT-16 gia' attivi)")
print(f" fade sleeve: {fade_ids} | caps={p.caps}")
print("=" * 100)
print("\n[1] build_everything() canonico (cache)...")
from src.portfolio.sleeves import all_sleeve_equities
eq_base = dict(all_sleeve_equities())
print(f" sleeve totali: {len(eq_base)}")
# dati + entries + maschera hurst (identica al live: close-only, w=100, step=6)
dfs, masks, entries = {}, {}, {}
for asset in ("BTC", "ETH"):
dfs[asset] = load_data(asset, "1h")
masks[asset] = hurst_skip_mask(dfs[asset], HURST_MAX)
print(f" {asset}: {len(dfs[asset])} barre, hurst-skip {masks[asset].mean()*100:.1f}% delle barre")
for nm, (fn, params) in strats_for(asset).items():
sid = f"{nm}_{asset}"
if sid in fade_ids:
entries[sid] = (asset, fn(dfs[asset], **params))
# --- [2] PARITA': mode=orig, trend ON, no hurst, no exit16 == canonico ---
print("\n[2] PARITA' replay (orig, trend ON, no hurst) vs canonico:")
print(f" {'sleeve':<10s}{'corr':>10s}{'ret_canon%':>14s}{'ret_replay%':>14s}{'diff%':>9s}")
parity_ok = True
for sid in fade_ids:
asset, ents = entries[sid]
rep = equity_from_trades(dfs[asset], build_trades_variant(
ents, dfs[asset], mode="orig", trend_max=TREND_MAX))
base = eq_base[sid]
corr = base.pct_change().fillna(0).corr(rep.pct_change().fillna(0))
rb = (base.iloc[-1] / base.iloc[0] - 1) * 100
rr = (rep.iloc[-1] / rep.iloc[0] - 1) * 100
flag = "" if (corr > 0.999 and abs(rr - rb) <= max(1.0, abs(rb) * 0.01)) else " <-- MISMATCH"
if flag:
parity_ok = False
print(f" {sid:<10s}{corr:>10.5f}{rb:>14.1f}{rr:>14.1f}{rr-rb:>+9.2f}{flag}")
print(f"\n PARITA' {'OK' if parity_ok else 'FALLITA'}.")
if not parity_ok:
print(" >>> STOP: non procedo col gate su un engine non in parita'.")
return
# --- [3] varianti live-path delle 6 fade ---
VARIANTS = {
"LIVE": dict(trend_max=None, hurst=True), # produzione oggi
"CANDIDATO": dict(trend_max=TREND_MAX, hurst=True), # + trend filter
"TREND-ONLY": dict(trend_max=TREND_MAX, hurst=False), # swap hurst->trend
"NESSUNO": dict(trend_max=None, hurst=False), # solo EXIT-16 (baseline filtri)
"TREND-2.5": dict(trend_max=2.5, hurst=False), # sensibilita' soglia
"TREND-3.5": dict(trend_max=3.5, hurst=False),
}
eqs = {v: {} for v in VARIANTS}
ntr = {v: {} for v in VARIANTS}
for sid in fade_ids:
asset, ents = entries[sid]
for v, cfg in VARIANTS.items():
tr = build_trades_variant(ents, dfs[asset], mode="exit16",
trend_max=cfg["trend_max"],
hurst_mask=masks[asset] if cfg["hurst"] else None)
eqs[v][sid] = equity_from_trades(dfs[asset], tr)
ntr[v][sid] = len(tr)
# --- [4] PORT06 per variante ---
print("\n" + "=" * 100)
print(f" [4] PORT06 (pesi cap, OOS da {OOS_DATE}) — fade in path LIVE (exit16+hurst)")
print("=" * 100)
print(f" {'variante':<12s}{'FULL Sh':>9s}{'FULL DD%':>10s}{'FULL CAGR':>11s}"
f" | {'OOS Sh':>8s}{'OOS DD%':>9s}{'OOS CAGR':>10s}")
print(" " + "-" * 94)
res = {}
for v in VARIANTS:
members = dict(eq_base)
for sid in fade_ids:
members[sid] = eqs[v][sid]
f, o = port_metrics(members, p)
res[v] = (f, o)
print(f" {v:<12s}{f['sharpe']:>9.2f}{f['dd']:>10.2f}{f['cagr']:>10.0f}%"
f" | {o['sharpe']:>8.2f}{o['dd']:>9.2f}{o['cagr']:>9.0f}%")
f_l, o_l = res["LIVE"]
f_c, o_c = res["CANDIDATO"]
print(" " + "-" * 94)
print(f" {'DELTA C-L':<12s}{f_c['sharpe']-f_l['sharpe']:>+9.2f}{f_c['dd']-f_l['dd']:>+10.2f}"
f"{f_c['cagr']-f_l['cagr']:>+10.0f}% | {o_c['sharpe']-o_l['sharpe']:>+8.2f}"
f"{o_c['dd']-o_l['dd']:>+9.2f}{o_c['cagr']-o_l['cagr']:>+9.0f}%")
# --- [5] per-sleeve ---
print("\n Per-sleeve (FULL): ret% | DD% | n trade [LIVE -> CANDIDATO]")
print(f" {'sleeve':<10s}{'ret L%':>10s}{'ret C%':>10s}{'DD L%':>8s}{'DD C%':>8s}"
f"{'ntr L':>7s}{'ntr C':>7s}")
for sid in fade_ids:
el_, ec = eqs["LIVE"][sid], eqs["CANDIDATO"][sid]
rl = (el_.iloc[-1] / el_.iloc[0] - 1) * 100
rc = (ec.iloc[-1] / ec.iloc[0] - 1) * 100
print(f" {sid:<10s}{rl:>10.1f}{rc:>10.1f}{_dd(el_):>8.1f}{_dd(ec):>8.1f}"
f"{ntr['LIVE'][sid]:>7d}{ntr['CANDIDATO'][sid]:>7d}")
# --- GATE ---
print("\n" + "=" * 100)
print(" GATE (sweep): PROMOSSO se OOS Sharpe >= LIVE-0.02 E OOS DD scende, e FULL non degrada")
print("=" * 100)
oos_sh_ok = o_c["sharpe"] >= o_l["sharpe"] - 0.02
oos_dd_ok = o_c["dd"] < o_l["dd"]
full_ok = f_c["sharpe"] >= f_l["sharpe"] - 0.02 and f_c["dd"] <= f_l["dd"] + 0.20
promoted = oos_sh_ok and oos_dd_ok and full_ok
print(f" OOS Sharpe {o_l['sharpe']:.2f} -> {o_c['sharpe']:.2f} ({'OK' if oos_sh_ok else 'KO'})")
print(f" OOS DD% {o_l['dd']:.2f} -> {o_c['dd']:.2f} ({'OK' if oos_dd_ok else 'KO'})")
print(f" FULL Sharpe {f_l['sharpe']:.2f} -> {f_c['sharpe']:.2f} | "
f"FULL DD {f_l['dd']:.2f} -> {f_c['dd']:.2f} ({'OK' if full_ok else 'KO'})")
print("\n VERDETTO: " + (">>> PROMOSSO <<<" if promoted else ">>> BOCCIATO <<<"))
if __name__ == "__main__":
main()