DETERMINISTIC BLACKJACK DECISION ENGINE RESEARCH
PHASE 1.12 — PACKAGE GUIDE

PLAIN-ENGLISH PURPOSE
---------------------
This package asks what a player would actually experience over a realistic
100-hand session. Each simulated player starts with £100. The report shows how
much was staked in total, the current balance after 10, 25, 50, 75 and 100
hands, the ending balance, the gain or loss, and the largest fall from a prior
high point.

WHAT WAS TESTED
---------------
10,000 independent sessions were created, giving 1,000,000 dealt hands. Three
strategies received exactly the same cards:

1. Flat £1 base-unit wagering.
2. £1 normally and £2 after a frozen exact previous-hand fingerprint.
3. £1 normally and £2 after a frozen broader previous-hand class.

The fingerprint lists were discovered and validated using the preserved Phase
1.11 training and validation cohorts. They were not changed after seeing the
Phase 1.12 session results.

IMPORTANT MONEY DEFINITIONS
---------------------------
Starting balance:
The £100 available at the beginning of every session.

Total amount staked:
Every settled wager added together. This can exceed £100 because the same money
may be wagered repeatedly. Doubles and splits can also increase the amount at
risk in one blackjack round.

Ending balance:
The money remaining after 100 hands.

Net gain or loss:
Ending balance minus the original £100. A minus sign means the player lost
money. A plus sign means the player finished above the starting balance.

Maximum drawdown:
The largest fall from the highest balance previously reached during that
session.

MAIN RESULT IN PLAIN ENGLISH
----------------------------
Across 10,000 sessions, flat wagering finished with an average balance of
£99.55. The exact fingerprint model finished with £99.47 and the broad model
with £99.44. Both fingerprint models staked more money and experienced larger
average drawdowns. Therefore the apparent Phase 1.11 holdout improvement did
not persist in this new realistic-session cohort.

This does not prove that fingerprint wagering must always perform worse. It
does show why independent replication and realistic session testing are
necessary before interpreting a statistical pattern as useful.

FAIR-PLAY AND ETHICAL ASSUMPTION
--------------------------------
The simulator assumes genuine fair decks, unbiased shuffling and fixed rules.
No future cards, hidden information, casino systems, live-play electronic aid,
shuffle seed, remaining-shoe composition or evasion method is supplied to the
wager controller. This package is a deterministic research and risk-analysis
experiment, not a promise of gambling profit.

FILES
-----
DeterministicBlackjackPhase112.java
    Complete Phase 1.12 Java source, including preserved earlier research.

Phase_1_12_Results.txt
    Complete console output containing Phase 1.12 and preserved evidence.

Phase_1_12_Session_Outcome_Study.txt
    Main player-centred session report.

Phase_1_12_Sample_Session_Diaries.txt
    Lowest, middle and highest exact-model session examples with balance
    checkpoints and total stakes.

Phase_1_12_Replay_Verification.txt
    Plain-English deterministic replay evidence.

Phase_1_12_README.txt
    This package guide.

PRESERVED SUPPORTING FILES
--------------------------
The Phase 1.11, Phase 1.10 sample-hand and fingerprint-frequency text files are
also included so the frozen trigger origin and prior evidence remain available.

PHASE STATUS
------------
Phase 1.12: PASS
Deterministic complete-session replay: PASS
Policy fingerprint preserved: PASS
