PHASE 3.09 — NEXT DESIGN NOTE

Phase 3.09 validates a dealer-dependent HARD 16 family intervention.

Increment versus 3.08:
Primary average ending bankroll: +1.392 units.
Holdout: +1.263.
Primary positive-session frequency: +3.42 percentage points.
Holdout: +2.52 percentage points.

The remaining replicated evidence now shifts toward HARD 15, HARD 14 and low-total HIT states. HARD 15 deserves particular attention because some rows have very high exposure, including HARD 15 vs dealer 10 with roughly 3,000 observations in each cohort.

PHASE 3.10 DESIGN QUESTION
Select the next intervention using both:
1. replicated damage per encounter; and
2. exposure/frequency across the complete programme.

Do not simply choose Rank 1 if it is rare. Do not blanket-convert all remaining states to Textbook. Preserve one coherent policy family and attribution.

All frozen architecture and information controls remain unchanged.
