PHASE 3.18 SECOND INTERVENTION RESULTS

PRIMARY
Average ending bankroll Textbook/Selective: 99.956 / 98.167
Maximum bankroll observed Textbook/Selective: 126.0 at hand 29 (session 1777) / 128.0 at hand 26 (session 3571)
Positive sessions Textbook/Selective: 48.14% / 38.36%
Paired winner Textbook/Selective/tie: 57.82% / 38.96% / 3.22%
Doubles Textbook/Selective: 15716 / 31892; splits Textbook/Selective: 3808 / 21706
Mean cards consumed Textbook/Selective: 165.674 / 175.190
Ending bankroll >=2x (200) Textbook/Selective: 0.00% / 0.00%
Ending bankroll >=3x (300) Textbook/Selective: 0.00% / 0.00%
Ending bankroll >=5x (500) Textbook/Selective: 0.00% / 0.00%
Ending bankroll >=10x (1000) Textbook/Selective: 0.00% / 0.00%

HOLDOUT
Average ending bankroll Textbook/Selective: 99.972 / 98.109
Maximum bankroll observed Textbook/Selective: 122.5 at hand 24 (session 1273) / 125.5 at hand 29 (session 3625)
Positive sessions Textbook/Selective: 47.42% / 37.78%
Paired winner Textbook/Selective/tie: 58.60% / 38.48% / 2.92%
Doubles Textbook/Selective: 15412 / 31596; splits Textbook/Selective: 3747 / 21486
Mean cards consumed Textbook/Selective: 165.647 / 174.989
Ending bankroll >=2x (200) Textbook/Selective: 0.00% / 0.00%
Ending bankroll >=3x (300) Textbook/Selective: 0.00% / 0.00%
Ending bankroll >=5x (500) Textbook/Selective: 0.00% / 0.00%
Ending bankroll >=10x (1000) Textbook/Selective: 0.00% / 0.00%


PHASE 3.18 — SELECTIVE-AGGRESSION INSTRUMENTATION SUMMARY

PRIMARY resolved selective-aggression hands: 171706; mean net/hand -0.0534; mean committed stake 1.1857; bust rate 16.07%; third-card bust rate when a third card was taken 20.93%.
Textbook would have chosen on those same visible first-decision states: HIT 70234, STAND 68056, DOUBLE 15822.

HOLDOUT resolved selective-aggression hands: 171486; mean net/hand -0.0551; mean committed stake 1.1842; bust rate 16.06%; third-card bust rate when a third card was taken 20.99%.
Textbook would have chosen on those same visible first-decision states: HIT 70136, STAND 67905, DOUBLE 15755.

INTERPRETATION
Phase 3.18 carries forward split-child HARD 20 STAND, HARD 19 STAND, and split-child HARD 18 STAND. It adds one new family only: HARD 9 vs dealer 9-Ace now HIT instead of automatically DOUBLING. All other Maximum Aggression behaviour remains frozen.

PRIMARY — WORST SELECTIVE-AGGRESSION VISIBLE STATES BY MEAN NET/HAND
Rank | Visible state | Hands | Mean net | Loss% | Bust% | 3rd-card bust% of 3rd-card draws | Mean stake
 1 | HARD 4 vs 11 [split-child] | textbook=HIT     |     12 |  -2.000 | 100.00% |   0.00% |   0.00% | 2.000
 2 | HARD 4 vs 9 [split-child] | textbook=HIT      |      9 |  -2.000 | 100.00% |   0.00% |   0.00% | 2.000
 3 | HARD 4 vs 7 [split-child] | textbook=HIT      |      9 |  -1.556 |  88.89% |   0.00% |   0.00% | 2.000
 4 | HARD 6 vs 4 [split-child] | textbook=HIT      |     32 |  -1.125 |  78.13% |   0.00% |   0.00% | 2.000
 5 | HARD 4 vs 10 [split-child] | textbook=HIT     |     36 |  -1.111 |  77.78% |   0.00% |   0.00% | 2.000
 6 | HARD 6 vs 3 [split-child] | textbook=HIT      |     29 |  -1.034 |  75.86% |   0.00% |   0.00% | 2.000
 7 | SOFT 19 vs 11 [split-child] | textbook=STAND  |     10 |  -1.000 |  70.00% |   0.00% |   0.00% | 2.000
 8 | HARD 5 vs 3 | textbook=HIT                    |    122 |  -0.787 |  69.67% |   0.00% |   0.00% | 2.000
 9 | HARD 4 vs 4 [split-child] | textbook=HIT      |     13 |  -0.769 |  69.23% |   0.00% |   0.00% | 2.000
10 | HARD 5 vs 2 [split-child] | textbook=HIT      |     30 |  -0.667 |  66.67% |   0.00% |   0.00% | 2.000
11 | HARD 4 vs 2 [split-child] | textbook=HIT      |      6 |  -0.667 |  66.67% |   0.00% |   0.00% | 2.000
12 | HARD 4 vs 3 [split-child] | textbook=HIT      |      6 |  -0.667 |  66.67% |   0.00% |   0.00% | 2.000
13 | HARD 5 vs 9 [split-child] | textbook=HIT      |     23 |  -0.652 |  78.26% |  65.22% |   0.00% | 1.000
14 | HARD 6 vs 8 [split-child] | textbook=HIT      |     28 |  -0.643 |  75.00% |  60.71% |   0.00% | 1.000
15 | HARD 6 vs 5 | textbook=HIT                    |    141 |  -0.624 |  65.25% |   0.00% |   0.00% | 2.000
16 | HARD 5 vs 3 [split-child] | textbook=HIT      |     20 |  -0.600 |  65.00% |   0.00% |   0.00% | 2.000
17 | SOFT 19 vs 4 [split-child] | textbook=STAND   |     27 |  -0.593 |  59.26% |   0.00% |   0.00% | 2.000
18 | HARD 5 vs 2 | textbook=HIT                    |    127 |  -0.583 |  64.57% |   0.00% |   0.00% | 2.000
19 | HARD 4 vs 6 [split-child] | textbook=HIT      |     14 |  -0.571 |  64.29% |   0.00% |   0.00% | 2.000
20 | SOFT 18 vs 11 [split-child] | textbook=HIT    |     14 |  -0.571 |  57.14% |   0.00% |   0.00% | 2.000
21 | HARD 16 vs 11 [split-child] | textbook=HIT    |    150 |  -0.567 |  75.33% |  67.33% |  67.33% | 1.000
22 | HARD 16 vs 10 [split-child] | textbook=HIT    |    869 |  -0.560 |  75.14% |  63.29% |  63.29% | 1.000
23 | HARD 7 vs 11 [split-child] | textbook=HIT     |     29 |  -0.552 |  68.97% |  41.38% |   0.00% | 1.000
24 | HARD 14 vs 11 [split-child] | textbook=HIT    |    189 |  -0.540 |  73.02% |  62.43% |  50.26% | 1.000
25 | HARD 15 vs 10 | textbook=HIT                  |   2934 |  -0.528 |  72.67% |  58.79% |  53.65% | 1.000

HOLDOUT — WORST SELECTIVE-AGGRESSION VISIBLE STATES BY MEAN NET/HAND
Rank | Visible state | Hands | Mean net | Loss% | Bust% | 3rd-card bust% of 3rd-card draws | Mean stake
 1 | HARD 4 vs 5 [split-child] | textbook=HIT      |      6 |  -2.000 | 100.00% |   0.00% |   0.00% | 2.000
 2 | HARD 4 vs 7 [split-child] | textbook=HIT      |      5 |  -2.000 | 100.00% |   0.00% |   0.00% | 2.000
 3 | HARD 4 vs 8 [split-child] | textbook=HIT      |      7 |  -1.429 |  85.71% |   0.00% |   0.00% | 2.000
 4 | HARD 4 vs 11 [split-child] | textbook=HIT     |     13 |  -1.385 |  84.62% |   0.00% |   0.00% | 2.000
 5 | HARD 5 vs 5 [split-child] | textbook=HIT      |     11 |  -1.273 |  81.82% |   0.00% |   0.00% | 2.000
 6 | HARD 4 vs 10 [split-child] | textbook=HIT     |     35 |  -1.200 |  80.00% |   0.00% |   0.00% | 2.000
 7 | HARD 4 vs 9 [split-child] | textbook=HIT      |     10 |  -1.200 |  80.00% |   0.00% |   0.00% | 2.000
 8 | HARD 6 vs 2 [split-child] | textbook=HIT      |     29 |  -1.172 |  79.31% |   0.00% |   0.00% | 2.000
 9 | SOFT 18 vs 11 [split-child] | textbook=HIT    |     12 |  -1.000 |  75.00% |   0.00% |   0.00% | 2.000
10 | HARD 4 vs 4 [split-child] | textbook=HIT      |     11 |  -0.909 |  72.73% |   0.00% |   0.00% | 2.000
11 | SOFT 14 vs 6 [split-child] | textbook=DOUBLE  |     15 |  -0.800 |  66.67% |   0.00% |   0.00% | 2.000
12 | HARD 6 vs 6 [split-child] | textbook=HIT      |     34 |  -0.765 |  67.65% |   0.00% |   0.00% | 2.000
13 | SOFT 14 vs 11 [split-child] | textbook=HIT    |     10 |  -0.700 |  80.00% |  60.00% |   0.00% | 1.000
14 | HARD 5 vs 3 | textbook=HIT                    |    125 |  -0.624 |  65.60% |   0.00% |   0.00% | 2.000
15 | HARD 6 vs 2 | textbook=HIT                    |    128 |  -0.609 |  64.84% |   0.00% |   0.00% | 2.000
16 | HARD 16 vs 9 [split-child] | textbook=HIT     |    229 |  -0.603 |  78.17% |  64.19% |  64.19% | 1.000
17 | HARD 5 vs 4 [split-child] | textbook=HIT      |     20 |  -0.600 |  65.00% |   0.00% |   0.00% | 2.000
18 | HARD 5 vs 2 | textbook=HIT                    |    130 |  -0.585 |  64.62% |   0.00% |   0.00% | 2.000
19 | HARD 6 vs 3 | textbook=HIT                    |    130 |  -0.585 |  63.08% |   0.00% |   0.00% | 2.000
20 | HARD 5 vs 5 | textbook=HIT                    |    131 |  -0.565 |  64.12% |   0.00% |   0.00% | 2.000
21 | HARD 5 vs 11 [split-child] | textbook=HIT     |      9 |  -0.556 |  77.78% |  44.44% |   0.00% | 1.000
22 | HARD 16 vs 11 | textbook=HIT                  |    521 |  -0.555 |  74.09% |  64.49% |  64.49% | 1.000
23 | HARD 14 vs 9 [split-child] | textbook=HIT     |    259 |  -0.544 |  75.29% |  61.39% |  53.28% | 1.000
24 | HARD 16 vs 10 [split-child] | textbook=HIT    |    784 |  -0.543 |  74.23% |  63.01% |  63.01% | 1.000
25 | HARD 5 vs 4 | textbook=HIT                    |    132 |  -0.515 |  62.88% |   0.00% |   0.00% | 2.000


PHASE 3.18 — REMAINING TRUE POLICY-DIVERGENCE STATES AFTER HARD 9 vs 9-ACE HIT FAMILY

Filter: aggregate ONLY hands where Selective Aggression's recorded first action differs from the frozen Textbook first action; then require the same visible state >=100 times in BOTH cohorts and negative mean net/hand in BOTH cohorts. This prevents already-protected states with negative realised outcomes from being mislabelled as remaining policy damage.

Rank | State | Primary n / net | Holdout n / net | P bust% / H bust% | Textbook first action
 1 | HARD 5 vs 3                        |   122 / -0.787 |   125 / -0.624 |   0.00% /   0.00% | HIT
 2 | HARD 5 vs 2                        |   127 / -0.583 |   130 / -0.585 |   0.00% /   0.00% | HIT
 3 | HARD 6 vs 3                        |   153 / -0.444 |   130 / -0.585 |   0.00% /   0.00% | HIT
 4 | HARD 6 vs 2                        |   157 / -0.318 |   128 / -0.609 |   0.00% /   0.00% | HIT
 5 | HARD 6 vs 5                        |   141 / -0.624 |   140 / -0.200 |   0.00% /   0.00% | HIT
 6 | HARD 5 vs 5                        |   139 / -0.245 |   131 / -0.565 |   0.00% /   0.00% | HIT
 7 | HARD 6 vs 4                        |   137 / -0.350 |   137 / -0.423 |   0.00% /   0.00% | HIT
 8 | SOFT 18 vs 10                      |   499 / -0.373 |   564 / -0.323 |   0.00% /   0.00% | HIT
 9 | SOFT 19 vs 10                      |   525 / -0.244 |   500 / -0.384 |   0.00% /   0.00% | STAND
10 | HARD 5 vs 4                        |   127 / -0.079 |   132 / -0.515 |   0.00% /   0.00% | HIT
11 | SOFT 18 vs 9                       |   144 / -0.389 |   144 / -0.167 |   0.00% /   0.00% | HIT
12 | HARD 7 vs 2                        |   280 / -0.257 |   284 / -0.289 |   0.00% /   0.00% | HIT
13 | SOFT 16 vs 3                       |   138 / -0.217 |   138 / -0.319 |   0.00% /   0.00% | HIT
14 | HARD 5 vs 6                        |   154 / -0.390 |   157 / -0.115 |   0.00% /   0.00% | HIT
15 | HARD 6 vs 6                        |   140 / -0.257 |   132 / -0.167 |   0.00% /   0.00% | HIT
16 | HARD 7 vs 3                        |   276 / -0.268 |   294 / -0.136 |   0.00% /   0.00% | HIT
17 | HARD 7 vs 5                        |   294 / -0.333 |   273 / -0.066 |   0.00% /   0.00% | HIT
18 | SOFT 14 vs 3                       |   108 / -0.093 |   130 / -0.292 |   0.00% /   0.00% | HIT
19 | HARD 7 vs 6                        |   286 / -0.182 |   275 / -0.116 |   0.00% /   0.00% | HIT
20 | HARD 10 vs 11                      |   277 / -0.173 |   286 / -0.105 |   0.00% /   0.00% | HIT

DESIGN USE
Phase 3.18 carries forward the validated split-child HARD 20, HARD 19 and split-child HARD 18 protections and replaces automatic DOUBLE in one hard-hand family: HARD 9 vs dealer 9-Ace -> HIT. Remaining replicated damage states are retained for later one-rule-at-a-time interventions.

PHASE 3.18 — RUNNING TXT AUDIT RECORD

OBJECTIVE
Make the fourth controlled stabilisation intervention while preserving the frozen architecture. Carry forward the three validated protections, then add one new policy family: HARD 16 follows its corresponding Textbook action rather than taking Maximum Aggression's automatic DOUBLE.

EVIDENCE HIERARCHY
1. Primary design authority: finite six-deck Phase 2.xx evidence and new Phase 3.xx data.
2. Earlier non-depleting/infinite-generator behaviour-profile data: historical corroboration and hypothesis generation only.
3. Old behaviour-profile evidence must not directly define a new policy unless finite-shoe evidence independently supports the same conclusion.

FROZEN VARIABLES
Six decks, deterministic session shuffle, 30 hands, 100-unit start, 1-unit base wager, dealer rules, blackjack payout, split limit, no in-session reshuffle, textbook control and Maximum Aggression policy are unchanged from Phase 2.02.

POLICY INFORMATION BOUNDARY
The future optimiser may use player hand composition/total, pair status, dealer upcard and legal actions. Phase 3.18 deliberately uses no previous-card-history, running count, remaining-rank inventory, hidden card or future-card feature. The accumulated policy uses only current visible states: split-child HARD 20, HARD 19, HARD 18, split-child HARD 16 vs Ace, and HARD 17. No card-history or composition feature is introduced.

PHASE 3.18 DECISION RULE
Carry forward split-child HARD 20 STAND, HARD 19 STAND and split-child HARD 18 STAND. ONE additional Phase 3.18 intervention: if an ordinary first-decision hand is hard (not soft) and totals 18, Selective Aggression STANDS. Otherwise the frozen Maximum Aggression policy is unchanged. Textbook remains the control.

PHASE 3.18 REPLAY FINGERPRINT: 3193501e13b6cdae8a1f56ef53f91fc7c160cd4126333ae1a30b1af73219f2ed
