iteration 3 · 2026-07-22 · laptop-drives-bigblack
The leakage-free scoreboard every out-of-sample number rides on. On 870,304 real bars, 6 of 7 substrate gates pass — including the crucial proof that purge/embargo is load-bearing. One gate (the PBO-null calibration) has an open residual, so #0 does not graduate yet. The substrate gate stays pending.
readonly=2. Compute wall ~34 s, peak RSS 521 MB (caps 5c/5G/no-swap, never hit). Empirical: 34.85% of consecutive bars close within 3s → ample real label-overlap leakage surface.#0 is the substrate gate — everything downstream rides on it, so it does not graduate on a partial or hand-waved pass. Six gates are solid; the seventh (PBO-null ∈ [0.45,0.55]) sits at 0.595. Rather than p-hack the frozen criterion or drop it, #0 is checkpointed and resumes next firing with a pre-registered null. Default = not-grounded.
| Gate | Result | |
|---|---|---|
| Fold generator | 0 label-window overlaps survive purge+embargo (naive has many) | PASS |
| Leakage gate | naive random k-fold straddles 252,512 overlapping pairs, leak t = 169; CPCV+purge leaves 0 straddling | PASS |
| Positive recovery | look-ahead IC* recovered by CPCV OOS, 0.55% bias, power 1.0 | PASS |
| N_min(H) surface | events/fold 290,101; N_min lag5 ρ0.14→30k, lag20 ρ0.079→100k; dialed-IC → PARK | PASS |
| Harden (zoo→substrate) | KP recovered · KN silent (0/15 folds) · KD identical | PASS |
| Null FPR | family-wise OOS-IC FPR 0.000 ≤ 0.05 | PASS |
| Null PBO ∈ [.45,.55] | 0.595 (S=10, 252 splits) | OPEN |
Because 34.85% of consecutive bars close within the 3s label window, a naïve random k-fold splits 252,512 overlapping-label pairs across train/test — and their leaked label correlation reads a t-stat of 169 (a torrent of spurious skill). The CPCV contiguous-group + purge + embargo partition removes every straddling overlap (0 pairs, 0 contamination) while preserving genuine signal (0.55% recovery bias). The leakage-free scoreboard works.