SPY 2004-01-02..2026-08-27 (23y). Overnight into the event day, net 2bps RT. IS<2016<=OOS. Placebo = random nights (matched count, 5000); Bonferroni bar p < 0.0038 across 13 legs. 'surv' = p0.
| leg | n | /yr | mean ON | t | net/ev | IS | OOS | placebo p | surv |
|---|---|---|---|---|---|---|---|---|---|
| fomc | 197 | 8.7 | +0.213% | +4.34 | +0.193% | +0.236% | +0.183% | 0.0002 | YES |
| quarter_end | 90 | 4.0 | +0.102% | +1.90 | +0.082% | +0.113% | +0.088% | 0.1764 | no |
| pre_holiday | 209 | 9.2 | +0.058% | +1.52 | +0.038% | +0.031% | +0.086% | 0.3024 | no |
| opex_tue | 263 | 11.6 | +0.036% | +0.82 | +0.016% | +0.006% | +0.070% | 0.4512 | no |
| opex_mon | 238 | 10.5 | +0.037% | +0.73 | +0.017% | +0.033% | +0.040% | 0.4570 | no |
| opex | 266 | 11.7 | +0.127% | +2.92 | +0.107% | +0.216% | +0.026% | 0.0116 | no |
| first_friday | 254 | 11.2 | +0.041% | +0.95 | +0.021% | +0.081% | -0.003% | 0.4104 | no |
| month_end | 272 | 12.0 | +0.018% | +0.52 | -0.002% | +0.040% | -0.007% | 0.6114 | no |
| triple_witch | 88 | 3.9 | +0.180% | +2.03 | +0.160% | +0.352% | -0.017% | 0.0220 | no |
| post_holiday | 209 | 9.2 | -0.024% | -0.45 | -0.044% | -0.001% | -0.049% | 0.8582 | no |
| fomc_plus1 | 201 | 8.9 | -0.051% | -0.63 | -0.071% | +0.009% | -0.128% | 0.9206 | no |
Survivors (p<0.0038 AND OOS>0): fomc.
Reading
- Only legs that survive the Bonferroni bar AND stay positive OOS are candidates to add; everything else is in-sample noise or decayed.
- The OpEx-week rows locate where the OpEx effect sits (which weekday carries the overnight drift).
- A leg that is positive IS but <=0 OOS is decayed, not tradeable.
Time 5s.