PHASE 3.07 — LAYMAN RESULTS EXPLANATION
FIFTH CONTROLLED STABILISATION INTERVENTION

WHAT CHANGED
Phase 3.07 keeps every protection validated in 3.03–3.06.

It changes one exact state only:
SPLIT-CHILD HARD 16 versus dealer Ace now HITS instead of automatically DOUBLING.

Why HIT? Because the Phase 3.06 evidence table records HIT as the corresponding Textbook first action for this visible state. We are therefore not inventing a new blackjack rule; we are replacing one damaging Maximum-Aggression decision with the established control decision for that state.

WHY THIS WAS TESTED
After 3.06 this was the leading remaining replicated damage row:
Primary: 162 occurrences, mean net/hand -1.309.
Holdout: 174 occurrences, mean net/hand -1.207.
The same state was damaging independently in both cohorts.

RESULT
Primary average ending bankroll rises from 88.684 in 3.06 to 88.705 in 3.07: +0.021 units.
Holdout rises from 88.461 to 88.482: +0.021 units.

Primary positive sessions move from 12.70% to 12.72%.
Holdout remains 12.18%.

This is a very small effect. That is scientifically useful: the state was individually damaging when encountered, but it is comparatively uncommon, so correcting it has little effect on the population-wide 30-hand bankroll average.

MULTIPLIER OUTCOMES
No session finished at 2x, 3x, 5x or 10x in either cohort.

PEAK BANKROLL
Selective Aggression maximum observed bankroll is unchanged:
Primary: 135.0 units at Hand 26.
Holdout: 132.0 units at Hand 30.

This reinforces our caution about treating isolated high-bankroll trajectories as optimisation targets. The new rule improves a narrow state without creating a new extreme peak.

WHAT THIS TEACHES US
Damage severity and programme-level importance are not the same thing.

A state can have a strongly negative result each time it occurs but still produce only a tiny overall improvement if it is rare. Phase 3.07 therefore adds an important selection lesson for later optimisation: future candidates should be judged using BOTH harm when encountered and how often the state contributes to the complete experiment.

NEXT QUESTION
After removing split-child HARD 16 vs Ace from the candidate set, HARD 17 states become especially prominent. The leading remaining row is split-child HARD 17 vs dealer 6, whose Textbook first action is STAND.

Before 3.08 changes anything, the remaining HARD 17 rows should be considered as a coherent family versus isolated dealer-specific states, while preserving the one-change-at-a-time attribution principle.
