PHASE 3.05 — LAYMAN RESULTS EXPLANATION
THIRD CONTROLLED STABILISATION INTERVENTION

WHAT CHANGED
Phase 3.05 keeps both earlier protections:
1. If a split child forms HARD 20, STAND rather than DOUBLE.
2. Any HARD 19 at its first decision STANDS rather than DOUBLE.

It adds one new rule only:
3. If a SPLIT CHILD forms HARD 18 at its first decision, STAND rather than DOUBLE.

Ordinary HARD 18 is deliberately NOT changed in 3.05. Everything else in Maximum Aggression remains frozen.

WHY THIS WAS TESTED
After Phase 3.04, HARD 18 dominated the remaining replicated damage evidence. The publication next-design decision narrowed the next test to split-child HARD 18 so we could add one tightly defined post-split protection and see exactly what it contributed.

RESULT
Primary average ending bankroll rises from 86.287 in 3.04 to 86.922 in 3.05: +0.635 units.
Holdout rises from 86.023 to 86.706: +0.683 units.

Positive sessions rise from 9.40% to 10.18% Primary and from 8.76% to 9.80% Holdout.

The improvement is therefore smaller than the HARD 20 and HARD 19 interventions, but it points in the same direction in both independent cohorts.

MULTIPLIER OUTCOMES
No session finished at 2x, 3x, 5x or 10x in either cohort. These remain observational measures only and do not influence the policy.

PEAK BANKROLL — IMPORTANT CORRECTION AND NEW MEASUREMENT
Phase 3.05 now measures actual bankroll after every hand.

Selective Aggression maximum observed bankroll:
- Primary: 127.0 units at Hand 30 (session 1543).
- Holdout: 129.0 units at Hand 18 (session 3473).

WHAT THE EVIDENCE NOW SAYS
The third intervention helps, but only modestly. This is useful evidence: not every additional safety layer produces the large gains seen in 3.03 and 3.04.

The remaining replicated ranking still contains ordinary HARD 18 states very prominently. Because ordinary HARD 18 was intentionally left untouched in 3.05, it remains available as a clean next experimental question.

CAUTION
A single maximum-bankroll trajectory is an extreme observation, not proof of a repeatable route. Peak trajectories should be recorded and compared across later phases before any policy is designed around them.
