PHASE 3.09 — LAYMAN RESULTS EXPLANATION
SEVENTH CONTROLLED STABILISATION INTERVENTION

WHAT CHANGED
Phase 3.09 keeps every validated protection from 3.03–3.08.

It changes one structured decision family only:
FIRST-DECISION HARD 16 now follows the corresponding Textbook action for the dealer's visible upcard instead of Maximum Aggression automatically DOUBLING.

That means:
- HARD 16 vs dealer 2–6 -> STAND.
- HARD 16 vs dealer 7–Ace -> HIT.

This is deliberately different from the HARD 17 intervention in 3.08. HARD 17 could use one action, STAND, across the family. HARD 16 requires a dealer-dependent mapping, so the experiment preserves that distinction.

WHY THIS WAS TESTED
After 3.08, HARD 16 occupied many prominent rows in the replicated damage table, with substantial exposure. The table also explicitly showed both HIT and STAND as the correct Textbook first action depending on dealer upcard. A blanket HARD 16 rule would therefore have destroyed information already present in our evidence.

RESULT
Primary average ending bankroll rises from 90.594 in 3.08 to 91.986 in 3.09: +1.392 units.
Holdout rises from 90.478 to 91.741: +1.263 units.

Positive sessions rise from 16.14% to 19.56% Primary and from 16.14% to 18.66% Holdout.

The direction replicates strongly in the independent Holdout cohort. The improvement is smaller than the HARD 17 family intervention but remains substantial.

PEAK BANKROLL
Selective Aggression maximum observed bankroll:
- Primary: 130.0 units at Hand 30.
- Holdout: 130.0 units at Hand 18.

No session finished at 2x, 3x, 5x or 10x.

The peak metric again behaves differently from the average distribution: bankroll stability improves materially without producing a larger extreme maximum.

WHAT THIS TEACHES US
Phase 3.09 demonstrates that a coherent intervention family does not have to mean one identical action.

The scientifically coherent unit can instead be:
'SAME PLAYER STATE FAMILY + CORRECT VISIBLE-DEALER MAPPING.'

That lets us remove a damaging aggressive behaviour while retaining the blackjack decision structure already represented by the frozen Textbook control.

WHAT REMAINS
After HARD 16 is protected, the remaining table shifts toward HARD 15, HARD 14 and several low-total states. Many of the most exposed remaining states prescribe HIT under Textbook.

This is another change in the evidence landscape. We should not automatically convert every remaining hand to Textbook play, because that would simply collapse Maximum Aggression into the control policy and destroy the purpose of controlled optimisation.

NEXT QUESTION
Phase 3.10 should determine which remaining family has the greatest combination of replicated damage and exposure while remaining a coherent, attributable intervention. HARD 15 is an especially important candidate because high-exposure HARD 15 rows now appear prominently.
