Weekly W-FRI, signal = noncomm net/OI z(156w), lag 1 week PIT, 10bps/switch, IS<2016<=OOS, Bonferroni bar p<0.00833 over 6 candidates.
| contract | proxy | dir | scope | Sharpe | buy-hold | exposure | placebo p | n | survives |
|---|---|---|---|---|---|---|---|---|---|
| VIX | SPY | long z>0 | IS | 0.38 | 0.49 | 33% | 0.1120 | 518 | no |
| VIX | SPY | long z>0 | OOS | 0.61 | 0.91 | 49% | 0.1745 | 562 | no |
| VIX | SPY | long z<0 | IS | 0.21 | 0.49 | 50% | 0.4315 | 518 | no |
| VIX | SPY | long z<0 | OOS | 0.63 | 0.91 | 51% | 0.1695 | 562 | no |
| VIX | SPY | corr(vs next-wk ret) | — | +0.030 | perm p 0.277 | — | — | 1080 | — |
| 10Y | TLT | long z>0 | IS | 0.14 | 0.58 | 50% | 0.4210 | 624 | no |
| 10Y | TLT | long z>0 | OOS | -0.34 | 0.01 | 31% | 0.5990 | 557 | no |
| 10Y | TLT | long z<0 | IS | 0.53 | 0.58 | 50% | 0.0185 | 624 | no |
| 10Y | TLT | long z<0 | OOS | 0.18 | 0.01 | 69% | 0.0225 | 557 | no |
| 10Y | TLT | corr(vs next-wk ret) | — | -0.013 | perm p 0.624 | — | — | 1181 | — |
| BTC | BTC | long z>0 | IS | 0.00 | 0.00 | 0% | 1.0000 | 0 | no |
| BTC | BTC | long z>0 | OOS | -0.12 | 1.05 | 30% | 0.9930 | 364 | no |
| BTC | BTC | long z<0 | IS | 0.00 | 0.00 | 0% | 1.0000 | 0 | no |
| BTC | BTC | long z<0 | OOS | 0.90 | 1.05 | 35% | 0.0955 | 364 | no |
| BTC | BTC | corr(vs next-wk ret) | — | -0.096 | perm p 0.077 | — | — | 364 | — |
Verdict
REJECT — no contract/direction beats buy-hold AND clears the Bonferroni placebo bar. (Last signal dates: ['2026-08-28', '2026-09-25', '2026-10-02'].)
Reading
- Adoptable only if an overlay beats buy-hold on Sharpe AND placebo p < the Bonferroni bar. Sign alone is not a verdict.
- CME BTC COT starts Dec-2017, so its 'IS' window is only 2017-2018. Runtime 22s.