Each gate vs skipping the same number of gate60-long days at random. N perms = 3000. Rule: REJECT if p>=0.05.
| Gate | days | long | skip | base SR | gated SR | obs lift | null mean | null p95 | p | verdict |
|---|---|---|---|---|---|---|---|---|---|---|
| F1 toxicity (abs_obi_z>0.5) | 1387 | 382 | 244 | 1.09 | 0.82 | -0.27 | -0.08 | +0.04 | 0.994 | REJECT |
| F2 OI-collapse (oi_z>=-1 & ex_z<=1) | 1387 | 382 | 247 | 1.09 | 0.73 | -0.36 | -0.08 | +0.04 | 1.000 | REJECT |
| F4 VIX (vix<=252d median) | 1387 | 382 | 119 | 1.09 | 1.20 | +0.12 | -0.04 | +0.08 | 0.014 | SIGNAL |
| F4b credit (spread<=252d median) | 1387 | 382 | 8 | 1.09 | 0.99 | -0.10 | -0.00 | +0.04 | 0.998 | REJECT |
| F9 stablecoins (gz>=-1) | 1387 | 382 | 4 | 1.09 | 1.06 | -0.02 | -0.00 | +0.02 | 0.942 | REJECT |
Summary: 1 of 5 gates SIGNAL
- REJECT -> the gate's skip adds nothing over random skipping (reduced-exposure effect, not information).
- SIGNAL -> the gate's macro/micro state decides which long days to skip.