DETERMINISTIC BLACKJACK DECISION ENGINE RESEARCH
PHASE 1.21 ? DOMINANT RECENT-CARD IMPRESSION

LAYMAN PURPOSE
--------------
Real players may not require three or four identical card categories in a row. They may see one contrary card as an exception and still feel that low cards or high cards are dominating. This phase tests that more human short-window impression.

WINDOW RULES
------------
2-of-3: at least two of the latest three visible cards are low or high.
3-of-4: at least three of the latest four visible cards are low or high.
Low cards are 2?7. High cards are 8?Ace. One contrary card is allowed.

FOUR BELIEF TYPES
-----------------
Low continuation: mostly low cards make another low card feel likely.
Low reversal: mostly low cards make a high card feel due.
High continuation: mostly high cards make another high card feel likely.
High reversal: mostly high cards make a low card feel due.

PRIMARY SESSION RESULTS
-----------------------
Textbook control              | avg ending ? 98.75 | W/L/E 2224/2682/94 | avg drawdown ?13.38 | activations 0
Low continuation 2-of-3       | avg ending ? 97.71 | W/L/E 2020/2890/90 | avg drawdown ?13.83 | activations 24856
Low continuation 3-of-4       | avg ending ? 98.53 | W/L/E 2193/2725/82 | avg drawdown ?13.48 | activations 7063
Low reversal 2-of-3           | avg ending ? 97.93 | W/L/E 2099/2800/101 | avg drawdown ?13.78 | activations 24032
Low reversal 3-of-4           | avg ending ? 98.41 | W/L/E 2149/2756/95 | avg drawdown ?13.54 | activations 8044
High continuation 2-of-3      | avg ending ? 96.58 | W/L/E 1882/3028/90 | avg drawdown ?14.52 | activations 95753
High continuation 3-of-4      | avg ending ? 97.27 | W/L/E 1987/2942/71 | avg drawdown ?14.12 | activations 67459
High reversal 2-of-3          | avg ending ? 98.70 | W/L/E 2211/2691/98 | avg drawdown ?13.41 | activations 1189
High reversal 3-of-4          | avg ending ? 98.71 | W/L/E 2214/2690/96 | avg drawdown ?13.40 | activations 814

DECISION AND PREDICTION EVIDENCE
--------------------------------
Low continuation 2-of-3       | prediction correct 46.12% | changed better/worse/same 5623/7919/11314 | dealer hidden high 54.64% | dealer next high 53.78% | dealer bust 39.14%
Low continuation 3-of-4       | prediction correct 47.42% | changed better/worse/same 1678/2142/3243 | dealer hidden high 53.39% | dealer next high 52.80% | dealer bust 39.23%
Low reversal 2-of-3           | prediction correct 53.64% | changed better/worse/same 4688/7384/11960 | dealer hidden high 54.07% | dealer next high 53.86% | dealer bust 23.56%
Low reversal 3-of-4           | prediction correct 52.37% | changed better/worse/same 1493/2594/3957 | dealer hidden high 54.04% | dealer next high 52.40% | dealer bust 22.50%
High continuation 2-of-3      | prediction correct 53.64% | changed better/worse/same 16616/24876/54261 | dealer hidden high 53.88% | dealer next high 53.58% | dealer bust 20.45%
High continuation 3-of-4      | prediction correct 54.25% | changed better/worse/same 11723/17393/38343 | dealer hidden high 54.07% | dealer next high 54.20% | dealer bust 20.52%
High reversal 2-of-3          | prediction correct 45.84% | changed better/worse/same 253/389/547 | dealer hidden high 55.42% | dealer next high 53.87% | dealer bust 38.69%
High reversal 3-of-4          | prediction correct 44.35% | changed better/worse/same 166/273/375 | dealer hidden high 55.41% | dealer next high 54.59% | dealer bust 38.82%

HITTING 15 OR 16 AFTER EXPECTING A LOW CARD
--------------------------------------------
Low continuation 2-of-3       | cases 24856 | survived 42.86% | busted 57.14% | hand improved 42.86%
Low continuation 3-of-4       | cases 7063 | survived 44.15% | busted 55.85% | hand improved 44.15%
High reversal 2-of-3          | cases 1189 | survived 38.52% | busted 61.48% | hand improved 38.52%
High reversal 3-of-4          | cases 814 | survived 37.59% | busted 62.41% | hand improved 37.59%

HOLDOUT SESSION RESULTS
-----------------------
Textbook control              | avg ending ? 98.55 | W/L/E 2187/2733/80 | avg drawdown ?13.56 | activations 0
Low continuation 2-of-3       | avg ending ? 97.51 | W/L/E 2003/2902/95 | avg drawdown ?14.03 | activations 25079
Low continuation 3-of-4       | avg ending ? 98.28 | W/L/E 2129/2785/86 | avg drawdown ?13.68 | activations 7203
Low reversal 2-of-3           | avg ending ? 97.73 | W/L/E 2048/2876/76 | avg drawdown ?13.95 | activations 24292
Low reversal 3-of-4           | avg ending ? 98.24 | W/L/E 2131/2801/68 | avg drawdown ?13.71 | activations 8114
High continuation 2-of-3      | avg ending ? 96.42 | W/L/E 1829/3090/81 | avg drawdown ?14.64 | activations 96039
High continuation 3-of-4      | avg ending ? 97.15 | W/L/E 1967/2941/92 | avg drawdown ?14.26 | activations 67746
High reversal 2-of-3          | avg ending ? 98.51 | W/L/E 2185/2740/75 | avg drawdown ?13.58 | activations 1152
High reversal 3-of-4          | avg ending ? 98.52 | W/L/E 2192/2735/73 | avg drawdown ?13.57 | activations 797

LAYMAN INTERPRETATION
--------------------
This phase does not ask whether a player can remember a perfect run. It asks whether a dominant impression?two of three cards or three of four cards?creates useful decisions. Continuation and reversal beliefs are tested separately because two people can see the same cards and form opposite expectations. Prediction accuracy, decision quality and final session outcome are reported separately.

ETHICAL AND INFORMATION BOUNDARY
--------------------------------
Only already-visible cards can form the impression. Hidden cards are evaluated afterwards. No card counting, remaining-shoe knowledge, concealed information or live gambling assistance is used. This is deterministic behavioural research, not a gambling recommendation.

DETERMINISTIC REPLAY: PASS
