Discovery panel (main _ADJ, 500 names, 2006-2026). Placebo = within-date score permutation, N=60. Rule: REJECT if placebo or block p>=0.05; else VARIANT if p_vs_mom>=0.10; else VIABLE.
| Candidate | Sharpe | placebo p | block p | p_vs_mom | corr_mom | grid med | grid+ | pos yrs | 24mo holdout | verdict |
|---|
| H2 liquidity-shock-reversal | 0.53 | 0.417 | 0.013 | 1.000 | 0.36 | 0.52 | 100% | 15/21 | 2.35 | REJECT |
| H6 beta-hedged-momentum | 0.89 | 0.000 | 0.000 | 0.156 | 0.95 | 0.87 | 100% | 19/21 | 1.69 | VARIANT |
- observed Sharpe 0.53; placebo null mean 0.51, p95 0.56 -> empirical p 0.417
- vs MOM_126 (same hold/k): mom 0.98, corr 0.36, p_vs_mom 1.000
- block(10) p 0.013; iid p 0.009; MinTRL 15293 obs (have 5030)
- halves 0.27 / 0.87; locked 24mo holdout 2.35 (p=0.000)
| control | Sharpe |
|---|
| PLAIN_REV (no gate) | 0.78 |
| SHOCK_ONLY | 0.52 |
| CALM_ONLY | 0.69 |
| H2_RESID_MOM | 0.42 |
| param | value | Sharpe |
|---|
| look | 3 | 0.52 |
| look | 5 | 0.53 |
| look | 10 | 0.46 |
| vol_mult | 1.25 | 0.60 |
| vol_mult | 1.5 | 0.53 |
| vol_mult | 2.0 | 0.32 |
| calm_thr | 0.005 | 0.42 |
| calm_thr | 0.01 | 0.53 |
| calm_thr | 0.02 | 0.50 |
- observed Sharpe 0.89; placebo null mean 0.68, p95 0.74 -> empirical p 0.000
- vs MOM_126 (same hold/k): mom 0.88, corr 0.95, p_vs_mom 0.156
- block(10) p 0.000; iid p 0.000; MinTRL 2098 obs (have 5030)
- halves 0.60 / 1.24; locked 24mo holdout 1.69 (p=0.011)
| control | Sharpe |
|---|
| MOM_126_raw | 0.88 |
| H6_RESID_MOM | 0.59 |
| param | value | Sharpe |
|---|
| beta_look | 63 | 0.82 |
| beta_look | 126 | 0.89 |
| beta_look | 252 | 0.87 |
- H2: REJECT (placebo p=0.417, p_vs_mom=1.000); H6: VARIANT (placebo p=0.000, p_vs_mom=0.156)
- REJECT = ranking is noise vs permutation; VARIANT = real but not independent of momentum; VIABLE = independent edge (fund a shadow next).