Sub-window robustness
filtered admission · selection entries 27 Sep–3 Oct 2026 UTC · source cutoff 4 Oct 06:00 UTC
Research onlyThe filtered arm requires a historical v1 h60 CLEAN score available at entry and vetoes causal BUNDLED scores. Raw means no additional 335 gate; the frozen RH universe already had safety gates. The pre-window is held out from this re-selection but overlaps earlier model development. The post-window has at most one complete entry hour and is underpowered. Scorer coverage is in the report.
Frozen comparison
| Admission / portfolio | Period | PnL | Max DD | Positive days | Entries |
|---|---|---|---|---|---|
| incumbent_core_v2 | pre | +20,230 | -4,974 | 89% | 4155 |
| incumbent_core_v2 | selection | +3,189 | -5,416 | 71% | 2216 |
| incumbent_core_v2 | post | +0 | 0 | 0% | 0 |
| A6_robust_subwindows | pre | +16,983 | -4,070 | 78% | 781 |
| A6_robust_subwindows | selection | +6,846 | -3,589 | 71% | 506 |
| A6_robust_subwindows | post | +0 | 0 | 0% | 0 |
| raw / separately selected A6_robust_subwindows | pre | +42,671 | -9,981 | 89% | 3676 |
| raw / separately selected A6_robust_subwindows | selection | +15,029 | -7,871 | 86% | 1746 |
| raw / separately selected A6_robust_subwindows | post | +0 | 0 | 0% | 0 |
| raw / same frozen roster | pre | +36,459 | -14,784 | 89% | 2963 |
| raw / same frozen roster | selection | +12,648 | -9,470 | 86% | 1381 |
| raw / same frozen roster | post | +0 | 0 | 0% | 0 |
Composer controls
Fee 0 as in the frozen RH composer; impact, participation, token caps and joint settlement use its published engine. Changing controls is exploratory and does not change the frozen tables.
Exit-time equity
Member scales
Edit a scale, or switch presets to restore its frozen value.
| Frozen book | Scale |
|---|
Daily PnL
| UTC day | Period | Net USD |
|---|---|---|
| 2026-09-18 | pre | +6,659 |
| 2026-09-19 | pre | -526 |
| 2026-09-20 | pre | +530 |
| 2026-09-21 | pre | +2,799 |
| 2026-09-22 | pre | +3,642 |
| 2026-09-23 | pre | +887 |
| 2026-09-24 | pre | +2,540 |
| 2026-09-25 | pre | -1,177 |
| 2026-09-26 | pre | +1,627 |
| 2026-09-27 | selection | -1,252 |
| 2026-09-28 | selection | +878 |
| 2026-09-29 | selection | -2,405 |
| 2026-09-30 | selection | +1,536 |
| 2026-10-01 | selection | +2,458 |
| 2026-10-02 | selection | +2,548 |
| 2026-10-03 | selection | +3,082 |
| 2026-10-04 | post | +0 |
Selection member correlation
Solo member daily net PnL over seven UTC days. Correlations from so few days are uncertain.
| Member | f11 | z04 |
|---|---|---|
| f11 | 1.00 | 0.66 |
| z04 | 0.66 | 1.00 |
Source: sealed 56-chain 330u composer lineage; score snapshot SHA 439ce57443a068fb…; selection freeze 253cf21f719bea6e…; Python/JS parity 120 scenarios, worst $0.0000. Complete cohort attribution by entry; equity ordered by exit.