Why trust this data
The simulation code and data are open source, so the assumptions behind the published metrics can be inspected instead of taken on faith. Review the source artifacts in the blackjack repository.
Methodology
How the Dataset Is Produced
Simulation Design
Fixed game configurations are run across deck count, penetration, rules, and strategy systems. Outputs are expectation values from repeated randomized runs.
RNG Description
Randomized shoe sequences use deterministic seeding and reproducible execution pipelines so dataset versions can be rebuilt and compared.
Convergence Approach
Each configuration runs until EV and variance estimates stabilize within a predefined tolerance band before publication.
Rulesets Tested
Rules include H17/S17 behavior and DAS. RSA and surrender are tracked as ruleset concepts, but they are currently not simulated in the EV data viewer dataset. Side bets and non-standard payouts are excluded.
Confidence Interpretation
Results are expectation values; variance and drawdown still apply. Real-world table conditions, dealer procedures, and player execution errors can shift outcomes.
Validation Notes
Validation compares aggregate metrics against benchmark references and internal consistency checks before publication.
Limitations
What This Data Does Not Claim
Expectation Values
Long-run averages describe simulated conditions. Individual sessions remain volatile and can diverge sharply from the mean.
Table Reality
Dealer procedures, penetration consistency, rule enforcement, heat, and execution errors are outside the static dataset.
Source Artifacts
Simulation source artifacts are published in the blackjack repository.