PHASE 3.394 — CLEAN PEEK vs NO-PEEK FULL-SESSION COMPARISON

DESIGN
Each numbered session begins from the same fresh deterministic six-deck shoe seed in every arm, £100 bankroll, and the same frozen controller starting state.
After play begins, the architectures are allowed to consume cards and diverge naturally through the complete session (maximum 30 hands, or earlier frozen stopping condition).
GBP, £15 table minimum, £5 ordinary wager grid, player policy and unrelated blackjack rules remain frozen.

IMPORTANT COMPARISON BOUNDARY
Phase 3.394 deliberately uses independent common-start sessions so Peek and No-Peek can be compared cleanly.
The HISTORICAL_FROZEN arm uses the historical round/controller rules but is run under this same independent-session framing; therefore its absolute population totals are not expected to reproduce the persistent cross-session Phase 3.388 population.

PRIMARY
HISTORICAL_FROZEN: n=10,000; mean final £87.5048; median £50.00; mean exposure £379.6005; mean maxDD £88.4295; reached £200 2,850 (28.50%); table exit 4,289 (42.89%); W/L/P 88,998/105,121/17,399; hits 137,275; splits 20,129; doubles 30,126; player naturals 9,031; dealer naturals 9,153; cards 1,097,819.
CLEAN_PEEK: n=10,000; mean final £88.5482; median £55.00; mean exposure £379.8785; mean maxDD £87.9485; reached £200 2,888 (28.88%); table exit 4,224 (42.24%); W/L/P 89,338/104,763/17,412; hits 137,905; splits 19,378; doubles 30,308; player naturals 9,062; dealer naturals 9,183; cards 1,100,463.
CLEAN_NO_PEEK: n=10,000; mean final £86.0028; median £45.00; mean exposure £378.7485; mean maxDD £89.2638; reached £200 2,786 (27.86%); table exit 4,340 (43.40%); W/L/P 88,547/104,475/17,223; hits 145,435; splits 19,983; doubles 31,015; player naturals 8,947; dealer naturals 9,113; cards 1,100,660.

HOLDOUT
HISTORICAL_FROZEN: n=10,000; mean final £86.4625; median £47.50; mean exposure £377.3080; mean maxDD £89.0325; reached £200 2,810 (28.10%); table exit 4,405 (44.05%); W/L/P 88,373/104,584/17,149; hits 136,968; splits 20,069; doubles 30,372; player naturals 8,935; dealer naturals 9,104; cards 1,090,750.
CLEAN_PEEK: n=10,000; mean final £87.5918; median £50.00; mean exposure £378.2050; mean maxDD £88.6100; reached £200 2,852 (28.52%); table exit 4,331 (43.31%); W/L/P 88,882/104,395/17,171; hits 137,777; splits 19,362; doubles 30,531; player naturals 8,997; dealer naturals 9,138; cards 1,095,121.
CLEAN_NO_PEEK: n=10,000; mean final £86.2290; median £42.50; mean exposure £378.4440; mean maxDD £89.3273; reached £200 2,787 (27.87%); table exit 4,408 (44.08%); W/L/P 88,655/104,207/16,979; hits 146,063; splits 20,092; doubles 31,330; player naturals 8,959; dealer naturals 9,157; cards 1,098,692.

DIRECT CLEAN PEEK MINUS CLEAN NO-PEEK — SAME SESSION STARTS
PRIMARY: mean final-bankroll difference +£2.5455 in favour of Peek; median £0.00; final bankroll differed in 3,428/10,000 sessions (34.28%); Peek finished higher in 1,843, No-Peek in 1,585; £200 status flipped in 834; table-exit status flipped in 1,356.
HOLDOUT: mean final-bankroll difference +£1.3627 in favour of Peek; median £0.00; final bankroll differed in 3,505/10,000 sessions (35.05%); Peek finished higher in 1,842, No-Peek in 1,663; £200 status flipped in 871; table-exit status flipped in 1,377.

FINDING
Under common starting conditions, clean Peek produced the stronger aggregate bankroll result in both PRIMARY and HOLDOUT. Clean No-Peek produced lower mean final bankroll and slightly poorer risk/termination outcomes.
The full-session design deliberately retains every later knock-on effect caused by different card consumption. No attempt is made to assign the final difference to a first, second or later dealer natural.

VERDICT: PASS.
Peek and No-Peek can be compared as simple complete-session architectures from identical starts, and PRIMARY/HOLDOUT show the same direction of effect.
