-
LIVE101/D0: OOS ret +1240.8%, SR 1.10, maxDD -35.5%, worst1mo -16.2%, boot-p 0.000
-
LIVE101/D1: OOS ret +1982.1%, SR 1.53, maxDD -16.7%, worst1mo -12.6%, boot-p 0.000 SR(D1)-SR(D0) bootstrap on daily diff: median +0.19, 2.5% -0.34, 97.5% +0.76
-
ALL500/D0: OOS ret +968.3%, SR 1.05, maxDD -38.6%, worst1mo -19.0%, boot-p 0.001
-
ALL500/D1: OOS ret +1668.4%, SR 1.52, maxDD -18.9%, worst1mo -12.4%, boot-p 0.000 SR(D1)-SR(D0) bootstrap on daily diff: median +0.24, 2.5% -0.31, 97.5% +0.80
D1 remains a shadow candidate: paired bootstrap on the 101-name leg is the only honest number because ALL500 with top_k=50 runs CAGR-style blended; both IS and OOS in the broad scan must stay sign-stable.
ALL DONE
Read
- D1 survives the broad 500-name robustness check numerically (SR 1.52 vs 1.05, maxDD -18.9% vs -38.6%, worst month -12.4% vs -19.0%).
- BUT the paired daily-diff bootstrap of SR(D1)-SR(D0) on the 101-name leg has median +0.19 with a 95% CI crossing zero ([-0.34, +0.76]). It is NOT a statistically distinct uplift at 5%; the documented benefit is themaxDD/worst-month drawdown cut at similar-order horizon, not a significance-chain-on-SR win. Per Phase-2 house rule above, that means D1 stays shadow-only until the 30d+ live shadow window clears, not promoted.