DETERMINISTIC BLACKJACK DECISION ENGINE RESEARCH
PHASE 3.100 — TECHNICAL / RESEARCH FINDINGS

TYPE
Paired deterministic decomposition.
No new intervention.
No threshold tuning.

RELEASE DEFINITION
Unconditional TRAIL160/25 stops.
Selective £140 gate does not stop.

PRIMARY
Released N=111 = 2.22% all sessions.
Selective £200 90.99%.
Selective £0 0%.
Selective final > unconditional retained final 95.50%.
Lower 0.90%.
Same 3.60%.
Mean final:
Reference £206.757;
Unconditional £155.811;
Selective £206.757.
Mean selective-unconditional delta +£50.946.
Extra wager +£79.865.
Extra max DD +£1.351.

HOLDOUT
Released N=100 = 2.00%.
£200 88.00%.
£0 0%.
Higher final 93.00%.
Lower 2.00%.
Same 5.00%.
Mean:
Reference £207.313;
Unconditional £154.363;
Selective £207.313.
Delta +£52.950.
Extra wager +£80.000.
Extra DD +£1.200.

INTERPRETATION
The predeclared severity gate released a small subgroup with exceptionally high continuation value in both cohorts.
The selective continuation path equals reference after release, providing an implementation consistency check.

LIMIT
No claim that £140 is optimal.
No population profitability claim.
Do not fine-tune neighbouring thresholds from these results.

MILESTONE
Phases 3.94–3.100 now provide:
hard-exit causal evidence;
trailing candidate mapping;
predeclared trailing causal test;
paired decomposition;
observable state separation;
predeclared selective-gate causal test;
released-session paired decomposition.

PLATE
Recommended: compact branch-summary plate with explicit negative-return / no-guaranteed-profitability boundary.
