Commit Graph

7 Commits

Author SHA1 Message Date
Adriano 922947d2aa research: verify options sleeve on REAL Deribit quotes (spread+skew haircut)
- options_real_quote_check.py: fetches real weekly BTC put chain, measures premium
  haircut (bid vs BS@DVOL-ATM), re-runs CSP sleeve with real haircut
- KEY FINDING (reverses a prior critique): backtest UNDER-prices the OTM put by using
  ATM DVOL; real skew (+28% gross) exceeds the ~4% bid/ask spread -> real bid premium
  = 1.29x modeled. Sleeve premium is conservative at current (calm) quotes.
- Real risk SHIFTS to the tail + roll-liquidity in stress (skew = market pricing fat
  tail), not premium magnitude. Breakpoint: sleeve dies below ~70% premium capture.
- updated eval diary with the verification
2026-06-19 21:48:12 +02:00
Adriano 58fc10de77 research tracks H+I: volume/vol/range + alt-momentum/reversal (both NEGATIVE for alpha)
- trackH volume_vol: no uncorrelated additive edge; profitable signals are trend-in-disguise
  (corr 0.6-0.75); MR/declining-volume fade dead even at fee 0; OBV-up filter is a defensive
  DD overlay only (13.3->10.1% DD but -CAGR), not new alpha
- trackI momentum/reversal: no formulation beats 1-3-6m sign-blend OOS on both assets;
  z-score continuous momentum = same edge (corr 0.96), lower DD 8.4% but lower CAGR;
  long-horizon reversal not bankable (negative/flat standalone). ~1.3 Sharpe ceiling holds.
- TP01 (12h sign-blend) remains the deployable winner
2026-06-19 21:22:49 +02:00
Adriano eac2aa1d00 audit+fix: anti-look-ahead audit, migrate deployable config to >=12h
- trackD_lookahead_audit.py: relabel test (left==right, no labeling leak) + execution-lag
  stress -> our trend pipeline is CLEAN (4h Sharpe 1.36 robust to +1 bar lag, label-invariant)
- ADOPT conservative conclusion: deploy at 12h (sub-12h: costs/overfit dominate, slight Sharpe
  bump unreliable). 12h: Sharpe 1.32, DD 13.3%, CAGR 16.2% ~ identical and robust
- trend_portfolio: DEPLOY_TF=12h, resample_tf(rule); paper trader + tests on 12h
- calendar research (NEGATIVE, both): trackF seasonality (spurious), trackG prior-levels
  (breakouts continue, fade dead; only long-drift survivor, redundant with TP01)
- gitignore data/paper_trend runtime state
2026-06-19 21:13:57 +02:00
Adriano 3b6ff02197 fix: resolution-safe timestamp in trackD_timing resample (pandas 2.x datetime64[ms])
- combination study: PORT LF4h (BTC+ETH) Sharpe 1.32 DD 12.3% remains best
- RV ETH/BTC market-neutral sleeve is genuinely uncorrelated (~0.05) but too weak
  (Sharpe 0.27) to raise portfolio Sharpe; combining the two TF configs is redundant
  (same-asset cross-config corr 0.80)
2026-06-19 20:18:03 +02:00
Adriano 8dbdadd509 research: add per-year trade counts + turnover to trackD_timing 2026-06-19 20:09:32 +02:00
Adriano 33267584d9 research: Track D winner across timeframes (15m-1d) + per-year PnL/DD
- trackD_timing.py: same TSMOM 1-3-6m blend config sampled at 15m/1h/4h/1d
- robust plateau across all TFs; 4h marginally best (LF Sharpe 1.36, DD 13.8%)
- per-year PnL and per-year max drawdown tables
2026-06-19 19:39:02 +02:00
Adriano dc2b5697da research wave 1: 5 honest tracks on certified BTC/ETH + synthesis
- trackA trend, trackB ML, trackC mean-rev, trackD trend-portfolio, trackE xsec/ensemble
- VERDICT: Track D vol-targeted BTC+ETH trend portfolio is the one robust deployable
  earner (Sharpe 1.0-1.32, DD 13-19%, positive every year 2019-2026)
- mean-reversion confirmed dead on clean data; weak-but-real ML/trend residuals
- honest: EUR50/day on 2000 in 1-2y is not reachable (needs ~137k capital or ruinous DD)
2026-06-19 19:14:53 +02:00