ETH 2022-12-01..2026-09-18 (OK 827, BLOCKED 561). Long iff ann-funding z(168h)[D-1]>1.5. Placebo = random entry of the same count, N=4000.
Rule: REJECT if OOS placebo p>=0.05.
#Placebo entry
| Segment | observed SR | n_long | n_cand | null mean | null p95 | empirical p |
|---|
| OOS | 0.55 | 22 | 621 | -0.02 | 1.22 | 0.242 |
| FULL | 0.71 | 31 | 827 | -0.00 | 0.82 | 0.080 |
- RIDE 0.55 | FADE -0.33 | buy-hold 0.15
- frequency-matched 7d-momentum 1.26 (mask corr 0.25)
| threshold | n_long | OOS Sharpe |
|---|
| z>1.0 | 49 | 0.34 |
| z>1.25 | 28 | 0.71 |
| z>1.5 <- PRIMARY | 22 | 0.55 |
| z>1.75 | 16 | -0.38 |
| z>2.0 | 16 | -0.38 |
- OOS placebo p=0.242 (observed 0.55 vs null -0.02); FULL p=0.080.
- REJECT -> the funding-z entry is indistinguishable from random entry; the trend leg of funding stays unproven.