503-name panel (2006-08-15..2026-08-11), trailing 21d cross-sectional stats, z(252), long/flat SPY held [t,t+1], 2bps/flip | IS<2016-01-01/OOS>= | Bonferroni bar p < 0.0063 over 8 legs | OOS SPY B&H 0.80.
| signal | exposure | IS SR | OOS SR | OOS vs B&H | partial-IC | placebo p | candidate |
|---|---|---|---|---|---|---|---|
| corr_lvl:low | 56% | 0.25 | 1.05 | +0.24 | -0.004 | 0.0312 | - |
| disp_lvl:low | 55% | 0.29 | 0.63 | -0.17 | -0.017 | 0.2375 | - |
| disp_chg:low | 49% | -0.26 | 0.62 | -0.19 | +0.021 | 0.0950 | - |
| corr_chg:low | 49% | -0.09 | 0.51 | -0.29 | -0.025 | 0.2600 | - |
| disp_lvl:high | 42% | -0.03 | 0.33 | -0.47 | -0.017 | 0.6150 | - |
| corr_chg:high | 48% | 0.20 | 0.27 | -0.53 | -0.025 | 0.6925 | - |
| corr_lvl:high | 41% | 0.16 | 0.26 | -0.54 | -0.004 | 0.8337 | - |
| disp_chg:high | 49% | 0.22 | 0.05 | -0.76 | +0.021 | 0.8762 | - |
Candidates: 0/8
Reading
- Cross-sectional correlation/dispersion is a genuinely different signal class (not price level / VIX). If nothing clears the bar, the regime carries no index-timing edge net of placebo + momentum.