Balance data

Historical measurements by the balance simulator on the older main snapshot; they do not measure current Poker Head builds or J-54. Each table names the command, seed, balance profile and commits it comes from.

Reading the numbers

• Player EV / bet: the average bankroll change per round divided by the bet, for a basic-strategy player. +10% means the player gains a tenth of the bet per round on average; negative numbers favour the house.
• Std. error: the measurement's uncertainty. Differences smaller than about two standard errors are noise.
• EV gain per use: EV with the item used once per hand minus EV without it, on the same deals.
• Bots do not exploit chaos rules beyond moving their bust line, so a skilled player can do better than these numbers under chaos.

Player EV under each chaos rule

The first row is plain blackjack. Rules with parameters are simulated with specific values, shown in the variant name.

python -m bjsim edge --rules all · seed 1 · balance default@b574fce2d1a5 · game 108b5e1 · simulator d11d84e · Player EV per round (basic strategy, 20000 rounds)

Item value

Measured without chaos. A Shield only helps while a rule is active, so it measures nothing here; an information item shows no gain when the bots' item policy cannot act on what it reveals.

The current bot policy does not act on Danger Radar's next-hit warning. Its zero here measures that policy's limitation, not the item's usefulness to a player.

python -m bjsim items --rules none · seed 1 · balance default@b574fce2d1a5 · game 108b5e1 · simulator d11d84e · Item value under `none` (baseline EV +0.95%, 20000 rounds)

Regression bands

python -m bjsim check · seed 1 · balance default@b574fce2d1a5 · game 108b5e1 · simulator d11d84e · Balance bands