DETERMINISTIC BLACKJACK DECISION ENGINE RESEARCH
PHASE 1.29 FINDINGS SUMMARY

RESEARCH DIRECTOR
Amit Amlani

PUBLICATION AUTHOR
ChatGPT (OpenAI)

WHAT WAS TESTED
---------------
Four fixed personalities were benchmarked over 5,000 primary sessions and
5,000 independent holdout sessions at each of two starting bankrolls: £100 and
£1,000. Every session had a maximum of 30 hands.

Each actual session was paired with a counterfactual replay in which the same
profile was forced to continue to hand 30. This made it possible to measure the
money retained by walking away rather than merely recording the balance when a
profile stopped.

PRIMARY FINDINGS
----------------
1. The three textbook-based profiles produced average walk-away balances close
   to their starting bankrolls in both the primary and holdout cohorts.

2. The aggressive profile usually triggered its emergency loss boundary very
   early. Its average walk-away hand was approximately 5.5, and its average
   retained bankroll was roughly one third of the starting bankroll.

3. The aggressive exit rule nevertheless protected a large amount relative to
   forcing that same aggressive player to continue. This must not be mistaken
   for good overall performance: the walk-away balance remained very poor.

4. Controlled-loss exits below the original bankroll occurred genuinely, but
   they were uncommon under the Phase 1.29 thresholds. They occurred most often
   for the Recovery Player, at approximately 1.4% to 1.7% of sessions.

5. The controlled-ambition profile showed the largest average amount protected
   by walking away among the three textbook-based profiles, while the
   disciplined and recovery profiles generally retained slightly more of the
   original bankroll. This is a trade-off rather than a single winner.

6. Results at £100 and £1,000 were directionally similar, supporting scale
   consistency of the percentage-based rules.

IMPORTANT INTERPRETATION
------------------------
Phase 1.29 does not prove one universally optimal balance. It demonstrates that
walk-away policies can be benchmarked as complete personalities and that a
balance below the starting bankroll can become a rational exit only after a
substantial drawdown and recovery.

The rarity of controlled-loss exits is itself a useful result. Phase 1.30
should deliberately vary the drawdown and recovery thresholds to determine
whether the Phase 1.29 rule was too strict, appropriately selective, or merely
one point within a wider optimum region.

MAIN PUBLICATION MESSAGE
------------------------
The same balance does not always mean the same thing. £90 after falling from
£100 may represent continuing decline. £90 after recovering from £50 may
represent a valuable escape from severe loss. The player history, not only the
current balance, determines whether walking away is rational.
