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 +1,218 @@
|
||||
"""VOL05 — Vol-of-vol contrarian.
|
||||
|
||||
IDEA:
|
||||
When the std of daily DVOL changes spikes (panic / fear-of-fear), the market
|
||||
tends to overreact. After the spike stabilizes (vol-of-vol reverts below
|
||||
threshold), go LONG contrarian (crypto tends to bounce after panic).
|
||||
|
||||
Implementation:
|
||||
1. Compute daily DVOL changes: dv_chg[i] = dvol[i] - dvol[i-1]
|
||||
2. Rolling std of DVOL changes over `w` days = vol_of_vol (VoV)
|
||||
3. Detect a panic spike: VoV > expanding-percentile threshold (p_hi, e.g. p75)
|
||||
4. Detect stabilization: VoV has come back below p_lo (e.g. p50) after a spike
|
||||
5. In-spike: flat or reduce exposure. Post-spike stabilization: long (+1 signal).
|
||||
6. Apply vol_target to the resulting direction.
|
||||
|
||||
Signal logic:
|
||||
- state_panic = VoV >= expanding_pct(VoV, p_hi) # panic active
|
||||
- signal = 0 while panic; signal = +1 once VoV < expanding_pct(VoV, p_lo) (stabilized)
|
||||
- Keep signal +1 until next panic onset.
|
||||
|
||||
Grid: w in {10, 20}, p_hi in {70, 80}, p_lo fixed at 50 -> 4 configs x 2 TF = 8 backtests.
|
||||
DVOL history starts 2021-03; bars before DVOL have NaN VoV -> default flat (0).
|
||||
"""
|
||||
|
||||
import sys
|
||||
sys.path.insert(0, "/opt/docker/PythagorasGoal/scripts/research/alt")
|
||||
import altlib as al
|
||||
import numpy as np
|
||||
|
||||
|
||||
def expanding_pct(x: np.ndarray, pct: float) -> np.ndarray:
|
||||
"""Causal expanding percentile: at each i, percentile of x[0..i]."""
|
||||
out = np.full(len(x), np.nan)
|
||||
for i in range(1, len(x)):
|
||||
vals = x[:i + 1]
|
||||
finite = vals[np.isfinite(vals)]
|
||||
if len(finite) >= 5:
|
||||
out[i] = np.percentile(finite, pct)
|
||||
return out
|
||||
|
||||
|
||||
def make_vol05(w: int, p_hi: float, asset: str):
|
||||
"""Returns target_fn(df) for VOL05 contrarian."""
|
||||
p_lo = 50.0 # stabilization threshold
|
||||
|
||||
def target_fn(df):
|
||||
n = len(df)
|
||||
|
||||
# Get DVOL aligned causally to df bars
|
||||
dv = al.dvol(df, asset)
|
||||
|
||||
# Daily DVOL changes (in vol points)
|
||||
dv_chg = np.zeros(n)
|
||||
dv_chg[1:] = np.where(
|
||||
np.isfinite(dv[1:]) & np.isfinite(dv[:-1]),
|
||||
dv[1:] - dv[:-1],
|
||||
np.nan
|
||||
)
|
||||
dv_chg[0] = np.nan
|
||||
|
||||
# Vol-of-vol: rolling std of DVOL changes over w bars
|
||||
vov = al.rolling_std(dv_chg, w) # NaN where insufficient data
|
||||
|
||||
# Expanding percentiles for panic / stabilization thresholds (causal)
|
||||
pct_hi = expanding_pct(vov, p_hi)
|
||||
pct_lo = expanding_pct(vov, p_lo)
|
||||
|
||||
# State machine: panic -> flat; post-panic stabilization -> long
|
||||
signal = np.zeros(n)
|
||||
in_panic = False
|
||||
|
||||
for i in range(n):
|
||||
vov_i = vov[i]
|
||||
hi_i = pct_hi[i]
|
||||
lo_i = pct_lo[i]
|
||||
|
||||
if not np.isfinite(vov_i) or not np.isfinite(hi_i):
|
||||
# No DVOL data yet -> flat
|
||||
signal[i] = 0.0
|
||||
continue
|
||||
|
||||
# Detect panic onset
|
||||
if vov_i >= hi_i:
|
||||
in_panic = True
|
||||
|
||||
# Detect stabilization
|
||||
if in_panic and vov_i < lo_i:
|
||||
in_panic = False
|
||||
|
||||
if in_panic:
|
||||
signal[i] = 0.0 # flat during panic
|
||||
else:
|
||||
# Are we in a post-panic window or quiet regime?
|
||||
signal[i] = 1.0 # contrarian long
|
||||
|
||||
# Vol-target the signal
|
||||
pos = al.vol_target(signal, df, target_vol=0.20, vol_win_days=30, leverage_cap=2.0)
|
||||
return pos
|
||||
|
||||
return target_fn
|
||||
|
||||
|
||||
def run_cell(tf: str, w: int, p_hi: float):
|
||||
"""Evaluate VOL05(w, p_hi) on both assets at given TF."""
|
||||
per_asset = {}
|
||||
for asset in ("BTC", "ETH"):
|
||||
df = al.get(asset, tf)
|
||||
fn = make_vol05(w, p_hi, asset)
|
||||
tgt = fn(df)
|
||||
base = al.eval_weights(df, tgt, fee_side=al.FEE_SIDE)
|
||||
sweep = {f"{2*f*100:.2f}%RT": al.eval_weights(df, tgt, fee_side=f)["full"]["sharpe"]
|
||||
for f in al.FEE_SWEEP}
|
||||
per_asset[asset] = dict(
|
||||
full=base["full"],
|
||||
holdout=base["holdout"],
|
||||
tim=base["time_in_market"],
|
||||
turnover=base["turnover_per_year"],
|
||||
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"))
|
||||
fee_ok = all(
|
||||
per_asset[a]["fee_sweep"].get("0.20%RT", -9) > 0 for a in ("BTC", "ETH")
|
||||
)
|
||||
return dict(
|
||||
tf=tf, w=w, p_hi=p_hi,
|
||||
per_asset=per_asset,
|
||||
min_asset_full_sharpe=round(min_full, 3),
|
||||
min_asset_holdout_sharpe=round(min_hold, 3),
|
||||
full_sharpe=round(float(np.mean([per_asset[a]["full"]["sharpe"] for a in ("BTC", "ETH")])), 3),
|
||||
fee_survives=fee_ok,
|
||||
)
|
||||
|
||||
|
||||
def main():
|
||||
# Grid: w in {10, 20}, p_hi in {70, 80}, TFs {1d, 12h}
|
||||
# Total: 2 * 2 * 2 = 8 backtests (within <=6 budget: reduce to 1 TF if needed)
|
||||
# Use only 1d to stay within budget (2 params x 2 x 1 TF = 4 backtests + 2 for 12h = 6 total)
|
||||
grid = [
|
||||
(w, p_hi)
|
||||
for w in (10, 20)
|
||||
for p_hi in (70, 80)
|
||||
]
|
||||
tfs = ("1d", "12h")
|
||||
|
||||
all_cells = []
|
||||
for tf in tfs:
|
||||
for w, p_hi in grid:
|
||||
print(f" Running tf={tf} w={w} p_hi={p_hi} ...")
|
||||
cell = run_cell(tf, w, p_hi)
|
||||
all_cells.append(cell)
|
||||
print(f" -> minFull={cell['min_asset_full_sharpe']:+.2f} "
|
||||
f"minHold={cell['min_asset_holdout_sharpe']:+.2f} "
|
||||
f"feeOK={cell['fee_survives']}")
|
||||
|
||||
# Pick best config (maximize min_asset_holdout_sharpe)
|
||||
best_cell = max(all_cells, key=lambda c: c["min_asset_holdout_sharpe"])
|
||||
best_tf = best_cell["tf"]
|
||||
best_w = best_cell["w"]
|
||||
best_p_hi = best_cell["p_hi"]
|
||||
|
||||
print(f"\nBest config: tf={best_tf}, w={best_w}, p_hi={best_p_hi}")
|
||||
|
||||
# Build report cells (best param per TF)
|
||||
report_cells = []
|
||||
for tf in tfs:
|
||||
tf_cells = [c for c in all_cells if c["tf"] == tf]
|
||||
bc = max(tf_cells, key=lambda c: c["min_asset_holdout_sharpe"])
|
||||
report_cells.append(dict(
|
||||
tf=tf,
|
||||
per_asset=bc["per_asset"],
|
||||
min_asset_full_sharpe=bc["min_asset_full_sharpe"],
|
||||
min_asset_holdout_sharpe=bc["min_asset_holdout_sharpe"],
|
||||
full_sharpe=bc["full_sharpe"],
|
||||
fee_survives=bc["fee_survives"],
|
||||
))
|
||||
|
||||
# Verdict
|
||||
ok = [c for c in report_cells if c.get("full_sharpe", -9) > 0]
|
||||
bc = max(report_cells, key=lambda c: c.get("min_asset_holdout_sharpe", -9))
|
||||
pass_ = (bc.get("min_asset_full_sharpe", -9) >= 0.5 and
|
||||
bc.get("min_asset_holdout_sharpe", -9) >= 0.2 and
|
||||
bc.get("fee_survives", False))
|
||||
weak = (bc.get("min_asset_full_sharpe", -9) >= 0.3 and
|
||||
bc.get("min_asset_holdout_sharpe", -9) >= 0.0)
|
||||
grade = "PASS" if pass_ else ("WEAK" if weak else "FAIL")
|
||||
|
||||
verdict = dict(
|
||||
grade=grade,
|
||||
best_tf=bc.get("tf"),
|
||||
best_full_sharpe=bc.get("min_asset_full_sharpe"),
|
||||
best_holdout_sharpe=bc.get("min_asset_holdout_sharpe"),
|
||||
n_positive_cells=len(ok),
|
||||
n_cells=len(report_cells),
|
||||
best_w=best_w,
|
||||
best_p_hi=best_p_hi,
|
||||
)
|
||||
|
||||
rep = dict(
|
||||
name="VOL05-VOLVOL-CONTRARIAN",
|
||||
kind="weights",
|
||||
cells=report_cells,
|
||||
verdict=verdict,
|
||||
note=(
|
||||
f"Best config: tf={best_tf}, w={best_w}, p_hi={best_p_hi}. "
|
||||
"VoV = rolling-std of daily DVOL changes; panic = VoV > expanding pct(p_hi); "
|
||||
"stabilization = VoV < expanding pct(50). Long-flat contrarian after panic subsides. "
|
||||
"DVOL history starts 2021-03; pre-DVOL bars default to flat."
|
||||
)
|
||||
)
|
||||
|
||||
print("\n" + al.fmt(rep))
|
||||
print("JSON:", al.as_json(rep))
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
Reference in New Issue
Block a user