PHASE 3.03 — SELECTIVE-AGGRESSION INSTRUMENTATION SUMMARY

PRIMARY resolved selective-aggression hands: 171688; mean net/hand -0.5019; mean committed stake 1.8509; bust rate 31.35%; third-card bust rate when a third card was taken 36.85%.
Textbook would have chosen on those same visible first-decision states: HIT 70230, STAND 68109, DOUBLE 15836.

HOLDOUT resolved selective-aggression hands: 171773; mean net/hand -0.5022; mean committed stake 1.8492; bust rate 31.31%; third-card bust rate when a third card was taken 36.88%.
Textbook would have chosen on those same visible first-decision states: HIT 70403, STAND 68001, DOUBLE 15767.

INTERPRETATION
Phase 3.03 changes one policy family only: split-child HARD 20 now STANDS instead of automatically DOUBLING. All other Maximum Aggression behaviour remains frozen. The purpose is to measure whether removing this single replicated damage mechanism improves the 30-hand bankroll distribution.
