DETERMINISTIC BLACKJACK DECISION ENGINE RESEARCH
PHASE 3.212 — CANDIDATE #13 INDEPENDENT REPLICATION / CLOSURE
HARD 12 vs DEALER 9: HIT versus STAND

Fresh HA/HB/HC/HD; 20,000 paired sessions.
Frozen reference HIT; alternate STAND. No hybrid/state/threshold changes.

3.212 RESULT
HIT: mean £84.774; £200 22.090%; terminal £0 39.535%.
STAND: mean £84.195; £200 21.860%; terminal £0 39.845%.
STAND-HIT: bankroll £-0.579; £200 -0.230 pp; terminal £0 +0.310 pp.

Candidate-touched 2,301 sessions:
STAND-HIT bankroll £-5.035;
£200 -1.999 pp;
terminal £0 +2.694 pp.

REPLICATION JUDGMENT
The 3.211 bankroll and terminal-survival HIT signal reproduces strongly.
The previously neutral £200 endpoint now also favours HIT.
All four fresh replication cohorts favour HIT on bankroll and terminal survival; three of four favour HIT on £200.

POOLED 3.211 + 3.212 — 40,000 PAIRS
HIT: mean £84.239; £200 21.795%; terminal £0 39.685%.
STAND: mean £83.751; £200 21.677%; terminal £0 40.005%.
STAND-HIT: bankroll £-0.488; £200 -0.117 pp; terminal £0 +0.320 pp.

PHASE 3.212 VERDICT
CANDIDATE #13 CLOSED.
HARD12_vs9: RETAIN HIT.
Classification: REFERENCE RETAINED — INDEPENDENTLY REPLICATED MULTI-OBJECTIVE HIT ADVANTAGE.
STAND not promoted. No hybrid justified.

This closes the currently planned individual high-rank candidate sequence.

PHASE 3.213 NEXT DESIGN — CROSS-CANDIDATE CONSOLIDATION I
Build the frozen candidate ledger before testing combinations.
Identify which candidate decisions are retained reference decisions, which alternatives earned promotion, and which cases were near-neutral/unstable.
Predeclare the first combination policy from evidence only; do not select combinations after observing outcomes.
Establish a fresh baseline-vs-combined-policy paired test and candidate-interaction instrumentation.
