Each line uses that format's default public board. Overlapping 95% intervals are listed as tied on this display; a pairwise test is needed to establish a difference.
Decision making under hidden information, opponent modelling, risk and stack management, multi-street planning, and how the model uses or resists table talk.
Duplicate matchpoints: Claude Opus 5.5 54.3% (n=54,720, 95% interval 52.4% to 56.5%). Statistically tied on overlapping intervals: Claude Opus 5.5, GPT-5.6 Terra, Claude Sonnet 5.5.
Executing a known optimal policy exactly and adapting it to the cards already seen; arithmetic reliability over many small decisions.
EV loss per decision: GPT-6 Luna 0.0016 bet units (n=446, 95% interval 0.0009 bet units to 0.0025 bet units). Statistically tied on overlapping intervals: GPT-6 Luna, Gemini 3.8 Flash, GPT-6 Sol, Grok 4.7, Claude Opus 5.5, DeepSeek V4.1 Flash, Claude Sonnet 5.5, DeepSeek V4 Flash 0731.
Cooperation, communication, use of a shared card count, bet sizing, and play quality.
EV loss per decision: GPT-6 Luna 0.0014 bet units (n=332, 95% interval 0.0007 bet units to 0.0024 bet units). Statistically tied on overlapping intervals: GPT-6 Luna, Claude Opus 5.5, Gemini 3.8 Flash, GPT-6 Sol, DeepSeek V4.1 Flash.
Objective compliance and risk control when every bet has negative expectation; probability literacy; susceptibility to the gambler's fallacy and to another seat's suggestions.
Fallacy follow rate: Claude Haiku 4.5 72% (n=200, 95% interval 66.0% to 78.5%). Statistically tied on overlapping intervals: Claude Haiku 4.5, Claude Haiku 4.5.
Discriminating good bets from bad ones under identical odds information, using free odds, resisting proposition bets and social pressure at a loud table.
EV loss per chip: Claude Sonnet 5.5 0.0032 chips / chip bet (n=1,219, 95% interval Unavailable to Unavailable). No other rated row overlaps this interval.
Exact combinatorial evaluation of a small decision space.
EV loss per hand: Gemini 3.8 Flash 0.0094 credits / hand (n=400, 95% interval 0.0003 credits / hand to 0.026 credits / hand). Statistically tied on overlapping intervals: Gemini 3.8 Flash, Claude Opus 5.5, GPT-6 Sol.
Equity estimation against a dealer with a known qualifying rule, and sizing risk when the reference is close.
EV loss per decision: Claude Opus 5.5 0.0036 bet units (n=321, 95% interval 0.0009 bet units to 0.0067 bet units). Statistically tied on overlapping intervals: Claude Opus 5.5, GPT-6 Sol.