26 names, 1158 days, 2020-08-14..2026-09-10. OOS >= 2025-01-01. N_perms=2000. Rule: REJECT if random-subset or permuted-RV OOS p>=0.05.
Ablation (OOS Sharpe)
| Arm | net2 | net10 |
|---|---|---|
| V1 bullish-only | 0.26 | -1.60 |
| V2 +RV>=1.0 | 1.25 | 0.01 |
Placebos (net2, same per-day count)
| Null | null mean | null p95 | observed | empirical p |
|---|---|---|---|---|
| random-subset | 0.58 | 1.28 | 1.25 | 0.058 |
| permuted-RV | 0.58 | 1.24 | 1.25 | 0.045 |
Verdict: REJECT
- random-subset p=0.058; permuted-RV p=0.045.
- REJECT -> the RV>=1.0 selection is indistinguishable from picking the same number of names at random; D1 stays parked.