PHASE 3.16 — NEXT DESIGN NOTE

AUDIT RESULT
The strongest coherent audited late-stage family was SOFT 13–17 vs dealer 7–Ace:
Primary n=5,819, aggregate net=-2,256.
Holdout n=5,776, aggregate net=-2,514.

INTERVENTION RESULT
Average bankroll:
Primary 97.319 -> 97.632 (+0.313).
Holdout 97.293 -> 97.683 (+0.390).

Positive sessions:
Primary +0.90 pp.
Holdout +0.94 pp.

PHASE 3.17
Re-audit the remaining families and explicitly investigate why HARD 16 strong-dealer and split-child HARD 16 states remain visible despite the earlier HARD 16 protection. Determine whether this is due to first-decision versus subsequent/split-child architecture before changing policy.

POLICY CHECKPOINT RULE
A later multiplier programme must inherit the best validated Phase 3.xx policy checkpoint, not automatically the latest tested rule. Negative or neutral late-stage experiments are evidence for rejection, not damage to the earlier validated architecture.

Wager sizing remains frozen throughout Phase 3.xx.
