312 FOMC events. Overnight prev close->open, net 2bps RT. SHRINK=1.5 if prior-day VIX in expanding top quartile (PIT), else 1.0. IS<2016-01-01<=OOS.
| variant | n | mean/event | IS SR | OOS ret | OOS SR | OOS maxDD |
|---|
| SPY flat | 198 | +0.194% | 4.13 | +15.3% | 5.90 | -1.3% |
| SPY VIX-Q4x1.5 | 198 | +0.263% | 4.01 | +20.7% | 5.63 | -1.8% |
| SPY VIX-Q4/1.5 | 198 | +0.147% | 4.15 | +11.8% | 6.01 | -1.0% |
| VUG flat | 196 | +0.251% | 4.97 | +23.4% | 6.41 | -1.6% |
| VUG VIX-Q4x1.5 | 196 | +0.333% | 4.71 | +32.0% | 6.18 | -1.9% |
| VUG VIX-Q4/1.5 | 196 | +0.197% | 5.10 | +18.0% | 6.40 | -1.4% |
| MTUM flat | 110 | +0.245% | 6.70 | +24.9% | 6.70 | -2.4% |
| MTUM VIX-Q4x1.5 | 110 | +0.299% | 6.86 | +32.3% | 6.51 | -2.9% |
| MTUM VIX-Q4/1.5 | 110 | +0.209% | 6.58 | +20.2% | 6.59 | -2.0% |
NOTE: the sleeve is flat ~97% of days, so the daily-return Sharpe column is inflated by the many zero days and is NOT the decision statistic. Compare mean/event and the paired tests below.
| instrument | multiplier | sized mean/event | flat mean/event | delta | p(sized<=flat) | verdict |
|---|
| SPY | Q4x1.5 | +0.263% | +0.194% | +0.070% | 0.0004 | sized wins |
| SPY | Q4/1.5 | +0.147% | +0.194% | -0.046% | 0.9996 | no difference |
| VUG | Q4x1.5 | +0.333% | +0.251% | +0.082% | 0.0000 | sized wins |
| VUG | Q4/1.5 | +0.197% | +0.251% | -0.055% | 1.0000 | no difference |
| MTUM | Q4x1.5 | +0.299% | +0.245% | +0.054% | 0.0006 | sized wins |
| MTUM | Q4/1.5 | +0.209% | +0.245% | -0.036% | 0.9994 | no difference |
| instrument | lag | Q4 mean | notQ4 mean | delta | p(sized<=flat) | q4-share | verdict |
|---|
| SPY | 1 | +0.445% | +0.079% | +0.070% | 0.0004 | 31% | pass |
| SPY | 2 | +0.426% | +0.085% | +0.068% | 0.0004 | 32% | pass |
| VUG | 1 | +0.518% | +0.128% | +0.082% | 0.0000 | 32% | pass |
| VUG | 2 | +0.500% | +0.134% | +0.080% | 0.0000 | 32% | pass |
| MTUM | 1 | +0.497% | +0.174% | +0.054% | 0.0006 | 22% | pass |
| MTUM | 2 | +0.478% | +0.176% | +0.054% | 0.0006 | 23% | pass |
| instrument | obs mean/event | null mean | placebo p | reads |
|---|
| SPY | +0.263% | (see null) | 0.0000 | conditional |
| VUG | +0.333% | (see null) | 0.0000 | conditional |
| MTUM | +0.299% | (see null) | 0.0112 | conditional |
- Instrument: adopt VUG/MTUM over SPY only if the OOS per-event
mean is higher AND lag-2 holds. (Flat-instrument result, not sizing.)
- Multiplier: adopt VIX-Q4 sizing only if A2 shows 'sized wins'
(p<0.05, higher per-event mean) at BOTH lag-1 and lag-2. If it wins on
mean but loses on per-event variance (Sharpe), a static high-beta
instrument is the lazier, stabler choice.
- The amplifier is OPT-IN and cannot ship before the forward shadow
(24/24 events) clears; this study only picks the config to vote on.
Time 1s.