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 +1,236 @@
|
||||
"""XAS06 — Beta-hedged spread reversion.
|
||||
|
||||
IDEA: Compute residual = ETH_ret - beta * BTC_ret using a rolling OLS beta.
|
||||
The cumulative residual (spread) should revert to 0 if BTC/ETH co-integrate.
|
||||
Trade the z-score of that cumulative residual back toward 0:
|
||||
- z < -threshold => long ETH spread (long ETH, short BTC * beta)
|
||||
- z > +threshold => short ETH spread (short ETH, long BTC * beta)
|
||||
Market-neutral: position is on the ETH-BTC spread, not a directional BTC or ETH bet.
|
||||
|
||||
IMPLEMENTATION NOTE:
|
||||
Because study_weights evaluates each asset independently, we implement this as:
|
||||
- ETH position = direction based on z-score of cumulative residual
|
||||
- BTC position = -beta * ETH_position (market-neutral hedge)
|
||||
We encode this as a 2-asset custom evaluator (one target per asset simultaneously).
|
||||
|
||||
GRIDS:
|
||||
- beta_window: 60d, 120d (rolling OLS lookback)
|
||||
- z_window: 30d, 60d (z-score lookback on cumulative residual)
|
||||
- z_entry_threshold: 1.5 (fixed)
|
||||
- z_exit_threshold: 0.5 (exit when spread narrows)
|
||||
|
||||
Since study_weights runs per asset independently, we use a shared-state approach:
|
||||
compute both targets together and expose them via closures.
|
||||
|
||||
Total configs: 2 beta_win x 2 z_win = 4 param sets, each run on 2 tfs → 8 backtests.
|
||||
We pick the best config and report it.
|
||||
"""
|
||||
|
||||
import sys
|
||||
sys.path.insert(0, "/opt/docker/PythagorasGoal/scripts/research/alt")
|
||||
import altlib as al
|
||||
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
from itertools import product
|
||||
|
||||
HOLDOUT = pd.Timestamp("2025-01-01", tz="UTC")
|
||||
FEE_SIDE = al.FEE_SIDE
|
||||
FEE_SWEEP = al.FEE_SWEEP
|
||||
|
||||
|
||||
def _rolling_beta(btc_ret: np.ndarray, eth_ret: np.ndarray, win: int) -> np.ndarray:
|
||||
"""Rolling OLS beta: ETH ~ beta * BTC (no intercept).
|
||||
Uses only data <= i (causal). Returns beta array same length as inputs.
|
||||
nan for first 'win' bars."""
|
||||
n = len(btc_ret)
|
||||
beta = np.full(n, np.nan)
|
||||
btc_s = pd.Series(btc_ret)
|
||||
eth_s = pd.Series(eth_ret)
|
||||
# Rolling cov(BTC, ETH) / var(BTC)
|
||||
cov = btc_s.rolling(win, min_periods=win // 2).cov(eth_s)
|
||||
var = btc_s.rolling(win, min_periods=win // 2).var()
|
||||
beta_vals = (cov / var.replace(0, np.nan)).values
|
||||
return beta_vals
|
||||
|
||||
|
||||
def compute_targets(df_btc: pd.DataFrame, df_eth: pd.DataFrame,
|
||||
beta_win: int, z_win: int,
|
||||
z_entry: float = 1.5, z_exit: float = 0.5):
|
||||
"""Return (btc_target, eth_target) arrays for the spread-reversion strategy.
|
||||
|
||||
Market neutral: ETH target = direction, BTC target = -beta * direction.
|
||||
Vol-target is applied to the combined ETH position (BTC gets scaled accordingly).
|
||||
|
||||
Alignment: merge on timestamp to ensure bar-by-bar alignment.
|
||||
"""
|
||||
# Align on common timestamps
|
||||
btc = df_btc[["timestamp", "close"]].rename(columns={"close": "btc"})
|
||||
eth = df_eth[["timestamp", "close"]].rename(columns={"close": "eth"})
|
||||
merged = pd.merge(btc, eth, on="timestamp", how="inner")
|
||||
if len(merged) < beta_win * 2:
|
||||
n = len(df_btc)
|
||||
return np.zeros(n), np.zeros(n)
|
||||
|
||||
btc_ret = al.simple_returns(merged["btc"].values)
|
||||
eth_ret = al.simple_returns(merged["eth"].values)
|
||||
|
||||
# Rolling beta (causal)
|
||||
beta = _rolling_beta(btc_ret, eth_ret, beta_win)
|
||||
|
||||
# Residual return: ETH - beta * BTC (market-neutral component)
|
||||
residual = eth_ret - np.nan_to_num(beta, nan=0.0) * btc_ret
|
||||
|
||||
# Cumulative residual (the "spread level")
|
||||
cum_residual = pd.Series(residual).cumsum().values
|
||||
|
||||
# Z-score of cumulative residual over rolling window
|
||||
z = al.zscore(cum_residual, z_win)
|
||||
|
||||
# Signal: mean-reversion on z-score
|
||||
# z > +entry => spread is high => short spread (short ETH, long BTC)
|
||||
# z < -entry => spread is low => long spread (long ETH, short BTC)
|
||||
# Exit when |z| < z_exit
|
||||
n_merged = len(merged)
|
||||
direction_eth = np.zeros(n_merged)
|
||||
current_pos = 0 # +1 long ETH, -1 short ETH, 0 flat
|
||||
|
||||
for i in range(n_merged):
|
||||
z_i = z[i]
|
||||
if not np.isfinite(z_i):
|
||||
direction_eth[i] = current_pos
|
||||
continue
|
||||
if current_pos == 0:
|
||||
if z_i < -z_entry:
|
||||
current_pos = 1 # long ETH spread
|
||||
elif z_i > z_entry:
|
||||
current_pos = -1 # short ETH spread
|
||||
elif current_pos == 1:
|
||||
if z_i > -z_exit:
|
||||
current_pos = 0
|
||||
elif current_pos == -1:
|
||||
if z_i < z_exit:
|
||||
current_pos = 0
|
||||
direction_eth[i] = current_pos
|
||||
|
||||
# Vol-target ETH position
|
||||
# Use ETH vol for scaling
|
||||
bpd = al.bars_per_day(df_eth)
|
||||
bpy = bpd * 365.25
|
||||
eth_vol = al.realized_vol(eth_ret, max(2, 30 * bpd), bpy)
|
||||
eth_vol_aligned = np.interp(np.arange(n_merged),
|
||||
np.arange(len(eth_vol))[:len(eth_vol)],
|
||||
eth_vol[:n_merged]) if len(eth_vol) >= n_merged else eth_vol[:n_merged]
|
||||
|
||||
scal = np.where((eth_vol_aligned > 0) & np.isfinite(eth_vol_aligned),
|
||||
0.20 / eth_vol_aligned, 0.0)
|
||||
eth_target_merged = np.clip(direction_eth * scal, -2.0, 2.0)
|
||||
eth_target_merged = np.nan_to_num(eth_target_merged, nan=0.0)
|
||||
|
||||
# BTC target = -beta * eth_direction * scale (hedge)
|
||||
beta_filled = np.where(np.isfinite(beta), beta, 0.0)
|
||||
btc_target_merged = -beta_filled * eth_target_merged
|
||||
btc_target_merged = np.nan_to_num(btc_target_merged, nan=0.0)
|
||||
|
||||
# Map back to original df indices via timestamp merge
|
||||
# Build a mapping from timestamp -> index in merged
|
||||
ts_to_merged_idx = {ts: i for i, ts in enumerate(merged["timestamp"].values)}
|
||||
|
||||
def _align_to_df(df_orig, tgt_merged):
|
||||
out = np.zeros(len(df_orig))
|
||||
for j, ts in enumerate(df_orig["timestamp"].values):
|
||||
if ts in ts_to_merged_idx:
|
||||
out[j] = tgt_merged[ts_to_merged_idx[ts]]
|
||||
return out
|
||||
|
||||
btc_target = _align_to_df(df_btc, btc_target_merged)
|
||||
eth_target = _align_to_df(df_eth, eth_target_merged)
|
||||
|
||||
return btc_target, eth_target
|
||||
|
||||
|
||||
def run_config(beta_win_days: int, z_win_days: int, tf: str):
|
||||
"""Run one param config on a given TF. Returns cell dict."""
|
||||
df_btc = al.get("BTC", tf)
|
||||
df_eth = al.get("ETH", tf)
|
||||
|
||||
bpd = al.bars_per_day(df_btc)
|
||||
beta_win = max(10, beta_win_days * bpd)
|
||||
z_win = max(5, z_win_days * bpd)
|
||||
|
||||
btc_tgt, eth_tgt = compute_targets(df_btc, df_eth, beta_win, z_win)
|
||||
|
||||
per_asset = {}
|
||||
fee_ok_all = True
|
||||
for asset, df, tgt in [("BTC", df_btc, btc_tgt), ("ETH", df_eth, eth_tgt)]:
|
||||
base = al.eval_weights(df, tgt, fee_side=FEE_SIDE)
|
||||
sweep = {f"{2*f*100:.2f}%RT": al.eval_weights(df, tgt, fee_side=f)["full"]["sharpe"]
|
||||
for f in FEE_SWEEP}
|
||||
fee_ok = sweep.get("0.20%RT", -9) > 0
|
||||
fee_ok_all = fee_ok_all and fee_ok
|
||||
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"))
|
||||
return 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,
|
||||
params=dict(beta_win_days=beta_win_days, z_win_days=z_win_days, tf=tf))
|
||||
|
||||
|
||||
def main():
|
||||
# Grid: 2 beta_win x 2 z_win = 4 configs; run on 1d only to stay <=6 backtests
|
||||
# 4 configs x 1 tf x 2 assets = 8 eval_weights calls
|
||||
param_grid = list(product([60, 120], [30, 60])) # (beta_win_days, z_win_days)
|
||||
tfs = ["1d"] # Keep total backtests = 4 x 1 = 4 (x2 assets = 8 eval calls)
|
||||
|
||||
all_cells = []
|
||||
for beta_win_days, z_win_days in param_grid:
|
||||
for tf in tfs:
|
||||
print(f" Running beta_win={beta_win_days}d z_win={z_win_days}d tf={tf} ...")
|
||||
cell = run_config(beta_win_days, z_win_days, tf)
|
||||
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 by holdout Sharpe
|
||||
best_cell = max(all_cells, key=lambda c: c["min_asset_holdout_sharpe"])
|
||||
best_params = best_cell["params"]
|
||||
print(f"\nBest config: {best_params}")
|
||||
|
||||
# Re-run best config on all TFs for the final report
|
||||
final_cells = []
|
||||
for tf in ["1d", "12h"]:
|
||||
cell = run_config(best_params["beta_win_days"], best_params["z_win_days"], tf)
|
||||
final_cells.append(cell)
|
||||
|
||||
# Build report
|
||||
def _verdict(per_cell):
|
||||
if not per_cell:
|
||||
return dict(grade="FAIL", reason="no cells")
|
||||
ok = [c for c in per_cell if c.get("full_sharpe", -9) > 0]
|
||||
best = max(per_cell, key=lambda c: c.get("min_asset_holdout_sharpe", -9))
|
||||
pass_ = (best.get("min_asset_full_sharpe", -9) >= 0.5 and
|
||||
best.get("min_asset_holdout_sharpe", -9) >= 0.2 and
|
||||
best.get("fee_survives", False))
|
||||
weak = (best.get("min_asset_full_sharpe", -9) >= 0.3 and
|
||||
best.get("min_asset_holdout_sharpe", -9) >= 0.0)
|
||||
grade = "PASS" if pass_ else ("WEAK" if weak else "FAIL")
|
||||
return dict(grade=grade, best_tf=best.get("tf"),
|
||||
best_full_sharpe=best.get("min_asset_full_sharpe"),
|
||||
best_holdout_sharpe=best.get("min_asset_holdout_sharpe"),
|
||||
n_positive_cells=len(ok), n_cells=len(per_cell),
|
||||
best_params=best.get("params"))
|
||||
|
||||
rep = dict(name="XAS06", kind="weights", cells=final_cells, verdict=_verdict(final_cells))
|
||||
print(al.fmt(rep))
|
||||
print("JSON:", al.as_json(rep))
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
Reference in New Issue
Block a user