Frozen rule: skip top-19% upper-wick-z of the top-decile 126d momentum names, thin half on dvol pct<0.2. 300 matched random-skip draws. Split 2023-01-01.
| statistic | observed | random-skip mean | random p |
|---|---|---|---|
| full diff (bp/day) | +1.14 | +0.15 | 0.013 |
| OOS diff (bp/day) | +1.79 | +0.63 | 0.163 |
| OOS both Sharpe | 1.51 | 1.29 | 0.060 |
Beta audit (both − base vs EW market)
beta -0.01 | raw mean +1.14 bp/day | beta-neutral mean +1.22 bp/day | OOS beta-neutral +1.89 bp/day
Reading
- Placebo p is P(random skip >= observed) at matched skip size; the wall signal must beat random skipping (p<0.05), else it is generic exposure reduction.
- Random-skip null spread (full diff): sd 0.48 bp, 5-95pct [-0.60, +0.93] bp/day. Time 543s.