ClaimReplication target
P5-3(a) / p13 — the demoted headline
0.333→0.750 under the frozen rule, but no family control; exact-gate family is underpowered. Replication target.
Under the historical seat-level rule, persona p13 moved from 0.333 cooperation at δ=.10 to 0.750 at δ=.90 and passed a per-candidate lower-bound test. The rule searched multiple candidates and fired on any pass without declared family-level error control — the defect external review exposed.
Three 200,000-permutation constructions
| Gate construction | Eligible max | Familywise p (add-one) | Monte Carlo 95% |
|---|---|---|---|
| Historical seat-level gate | p13/s2a (+0.4167) | 0.059230 | [0.058194, 0.060268] |
| Percentile episode-cluster bootstrap | p13/s2a (+0.4167) | 0.043455 | [0.042561, 0.044353] |
| Conservative exact-episode gate | p05/s2a (+0.0833; p13 ineligible) | 0.773206 | [0.771363, 0.775039] |
The exact family cannot reach 0.05 at all: with six episodes per condition, the minimum attainable familywise p is 0.075040. So the archived record neither prospectively confirms nor decisively disconfirms p13 — it is a replication target whose next test (Phase 6) must preregister the candidate family, gate, statistic, decision rule, and sample size before any data are collected. The favorable 0.043 variant cannot create prospective confirmation; the sealed Branch-2 discussion that celebrated p13 is preserved with its correction table.