Firm-style anomalies testable on daily OHLC. Judged by testbot_baseline.judge.
| Variant | IS Sharpe | IS maxDD | OOS Sharpe | OOS maxDD | OOS ret | turn/yr |
|---|---|---|---|---|---|---|
| ON1 BASE-overnight | 1.50 | -30.1% | 2.04 | -17.7% | +148.4% | 8.5 |
judge: ON1 BASE-overnight vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 1.50 | -54.7% | -30.1% | | OOS | 1.74 | 2.04 | -28.5% | -17.7% | IS bootstrap p=0.550 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on
| ID1 BASE-intraday | -0.01 | -47.8% | 0.51 | -29.4% | +26.2% | 8.5 |
judge: ID1 BASE-intraday vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | -0.01 | -54.7% | -47.8% | | OOS | 1.74 | 0.51 | -28.5% | -29.4% | IS bootstrap p=1.000 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on
| ONMOM overnight-rank | 1.88 | -28.7% | 1.89 | -21.1% | +133.1% | 6.5 |
judge: ONMOM overnight-rank vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 1.88 | -54.7% | -28.7% | | OOS | 1.74 | 1.89 | -28.5% | -21.1% | IS bootstrap p=0.048 | IS p<0.05: OK | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: PASS — advance to forward shadow
| ONMOM-rank/CC-hold | 0.90 | -65.1% | 1.58 | -32.8% | +190.5% | 6.5 |
judge: ONMOM-rank/CC-hold vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.90 | -54.7% | -65.1% | | OOS | 1.74 | 1.58 | -28.5% | -32.8% | IS bootstrap p=0.107 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on
| ONMOM-full-churn | -1.98 | -98.8% | -1.95 | -64.9% | -61.9% | 510.5 |
judge: ONMOM-full-churn vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | -1.98 | -54.7% | -98.8% | | OOS | 1.74 | -1.95 | -28.5% | -64.9% | IS bootstrap p=1.000 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: NO verdict: FAIL — move on
| ONWKND Fri-Mon-only | -0.90 | -70.0% | -0.54 | -25.5% | -13.0% | 98.4 |
judge: ONWKND Fri-Mon-only vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | -0.90 | -54.7% | -70.0% | | OOS | 1.74 | -0.54 | -28.5% | -25.5% | IS bootstrap p=1.000 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on
| HS same-month | 0.85 | -42.0% | 1.68 | -27.0% | +152.4% | 16.3 |
judge: HS same-month vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.85 | -54.7% | -42.0% | | OOS | 1.74 | 1.68 | -28.5% | -27.0% | IS bootstrap p=0.755 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on
| TOM overlay | 0.69 | -39.0% | 1.14 | -21.4% | +54.6% | 17.3 |
judge: TOM overlay vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.69 | -54.7% | -39.0% | | OOS | 1.74 | 1.14 | -28.5% | -21.4% | IS bootstrap p=0.999 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on
Verdicts (2026-09-09)
Scoreboard: 38 judged, 0 adopted, 1 watchlisted.
- ONMOM — PASS downgraded to WATCHLIST.* The signal is real (guard
clean, IS p=0.048, OOS confirms, ON1-strong/ID1-zero internal
consistency). But the gate's cost model didn't cover the strategy's
defining trade, and the follow-ups are decisive:
- full-churn (daily round trips at 10bps): Sharpe -1.98. Lethal.
- CC-hold of overnight-ranked names: FAIL (the edge lives in the overnight session itself, not in the names).
- weekend-only harvest: FAIL (-0.90; weekend overnights carry nothing — consistent with the classic weekend effect). Harvesting needs daily round-trips; at 1-2bps institutional costs it might survive, at this stack's 10bps it cannot. Not forwarded to shadow (shadowing the untradable is theater). Revisit only with cheaper execution. p=0.048 across ~38 tested variants also deserves multiple- testing humility.
- HS seasonality — FAIL (p=0.755, turnover 16.3/yr churn for nothing).
- TOM overlay — FAIL decisively (p=0.999 against). Sitting out 60% of days kills compounding; ToM timing actively harmful here.
- ID1 ≈ 0 IS is the round's cleanest finding: BASE's entire profit accrues overnight; the intraday leg contributes nothing. Explains why stops and intraday timing never helped in earlier rounds.