SPY 2006-08-14..2026-08-11 | vrp = z252(log(VIX/RV21)) | long/flat | held [t,t+1] | IS<2016-01-01/OOS>= | 2bps/flip | 21d block placebo x2000.
| leg | exposure | FULL net SR | IS net SR | OOS net SR | OOS B&H | FULL DD | flips |
|---|---|---|---|---|---|---|---|
| vrp_z>0 (PRIMARY) | 49% | 0.71 | 0.68 | 0.76 | 0.80 | -27% | 434 |
| bvol_z>0 (VIX level) | 42% | 0.26 | 0.16 | 0.37 | 0.80 | -55% | 372 |
| rv_z>0 (RV level) | 46% | 0.35 | 0.16 | 0.55 | 0.80 | -52% | 175 |
| vrp_z<0 (reverse) | 48% | 0.08 | -0.23 | 0.39 | 0.80 | -65% | 435 |
| bvol_z<0 (reverse) | 56% | 0.56 | 0.19 | 0.94 | 0.80 | -20% | 373 |
Placebo (vrp_z>0, 21d block shuffle)
- OOS obs 0.76 vs null 0.52/p95 0.89 -> p 0.152
- IS obs 0.68 vs null 0.21 -> p 0.013
- partial IC(vrp_z, fwd1 | SPY 252d mom) = +0.0294
- DSR(K=5) OOS net vs B&H 0.80: 0.44 | haircut +0.56
Verdict: REJECT
- PRIMARY vrp_z>0: OOS net-vs-BH -0.05, placebo p 0.152, IS edge +0.35, partial-IC|mom +0.029
- REJECT -> the equity VRP does not time SPY at the daily horizon net of the placebo + momentum bars. VIX feeds stay infrastructure.
Unregistered observation (do NOT adopt)
- The
bvol_z<0control (hold SPY when VIX < its 252d mean) shows OOS net SR 0.94 vs B&H 0.80 with DD -20% vs the VRP legs' -52..-65%. It was NOT pre-registered, has no placebo here, and low-vol timing is a known momentum/regime proxy. If pursued, it needs its own pre-registered study (placebo + momentum control + IS/OOS), not a post-hoc promotion from this table.