PEAD leg = validated base_topQ H60 (expand2 construction). Series truncated to 2020+ (earnings usable from 2020): IS''=2020-2022 stress window, OOS>=2023. Judged by testbot_baseline.judge.
| Variant (2020+) | IS'' Sharpe | IS'' maxDD | OOS Sharpe | OOS maxDD | OOS ret |
|---|---|---|---|---|---|
| Q1 PEAD top-Q alone | 0.29 | -50.4% | 1.65 | -20.1% | +115.6% |
judge: Q1 PEAD top-Q alone vs frozen baseline v1 (overlap 1661d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 1.06 | 0.29 | -34.1% | -50.4% | | OOS | 1.74 | 1.65 | -28.5% | -20.1% | IS bootstrap p=0.888 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on
| Q2 book 50/50 | 0.69 | -42.1% | 1.81 | -24.1% | +166.7% |
judge: Q2 book 50/50 vs frozen baseline v1 (overlap 1661d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 1.06 | 0.69 | -34.1% | -42.1% | | OOS | 1.74 | 1.81 | -28.5% | -24.1% | IS bootstrap p=0.888 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on
| Q3 book 70/30 | 0.86 | -38.9% | 1.80 | -25.9% | +188.4% |
judge: Q3 book 70/30 vs frozen baseline v1 (overlap 1661d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 1.06 | 0.86 | -34.1% | -38.9% | | OOS | 1.74 | 1.80 | -28.5% | -25.9% | IS bootstrap p=0.888 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on
Verdicts (2026-09-09)
ADOPT: nothing. Scoreboard: 29 judged variants, 0 adopted.
- Q1 PEAD alone — FAIL. Stress-window IS'' 0.29 (COVID/2022 maul earnings drift), OOS 1.65 ≈ BASE. Standalone PEAD is no replacement.
- Q2/Q3 books — FAIL on the bar, best OOS prints of the program (1.81/1.80 vs 1.74, better DD both). But the bar is the bar: IS'' p=0.888.
- Why all three share p=0.888: Q2/Q3 diffs vs BASE are positive scalar multiples of the Q1 diff (blends contain BASE itself), and bootstrap-p is scale-invariant — all three test the same direction. Not a bug; the books' fate is decided by PEAD-vs-BASE in 2020-22, where PEAD bled (0.29 vs 1.06).
- Structural note: sleeve-level combination beats rank-level blending (U1 failed everywhere; Q2/Q3 at least match OOS), but "matches with better DD" is insurance, and insurance doesn't auto-pass. The 2020-22 stress window is doing its job: anything that bleeds there can't advance without overwhelming evidence.