Monthly trailing-252d-Sharpe-weighted legs {BASE, NM1, U1} (ADAPT) + static equal-weight (EQ3). Judged by testbot_baseline.judge.
| Variant | IS Sharpe | IS maxDD | OOS Sharpe | OOS maxDD | OOS ret |
|---|---|---|---|---|---|
| ADAPT sharpe-blend | 0.90 | -53.0% | 1.77 | -24.6% | +183.8% |
judge: ADAPT sharpe-blend vs frozen baseline v1 (overlap 4765d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.91 | 0.90 | -54.7% | -53.0% | | OOS | 1.74 | 1.77 | -28.5% | -24.6% | IS bootstrap p=0.529 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on
| EQ3 equal-weight | 0.93 | -52.0% | 1.84 | -24.5% | +192.3% |
judge: EQ3 equal-weight vs frozen baseline v1 (overlap 4765d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.91 | 0.93 | -54.7% | -52.0% | | OOS | 1.74 | 1.84 | -28.5% | -24.5% | IS bootstrap p=0.341 | 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: 40 judged, 0 adopted, 1 watchlisted.
- ADAPT — FAIL. Monthly Sharpe-chasing across legs adds nothing (IS p=0.529, OOS tied). Adaptation to trailing performance is noise-chasing at this horizon: by the time a leg's 252d Sharpe proves itself, the regime has moved on.
- EQ3 — FAIL on the bar, and the round's real lesson. Static equal-weight (OOS 1.84, DD -24.5%) beats the adaptive blender (1.77) built from the same legs. Fourth independent confirmation of the parked RL-allocator verdicts: adaptation does not beat equal-weight net of costs in this program. The mechanism differs (Sharpe-weighted legs vs RL sleeve allocator); the outcome rhymes exactly.
- Overlap starts Nov 2008 (blender needs 252d leg history) — the GFC bottom is excluded for all legs equally; comparison stays fair.