What each row contains
| Field | What it is |
|---|---|
game | The game the bet belongs to |
bet | The specific wager |
house_edge_percent |
The edge as a percentage of the amount staked |
house_edge_exact_fraction |
The same value as an exact rational number, where the game admits one |
page |
The URL of the page that derives the figure |
The cheapest eight, as a sample
| Game | Bet | Exact | House edge |
|---|---|---|---|
| blackjack rule delta | surrender | – | -0.0008% |
| craps | odds pass 4 | 0/1 | 0.0000% |
| craps | odds dont 4 | 0/1 | 0.0000% |
| craps | odds pass 5 | 0/1 | 0.0000% |
| craps | odds dont 5 | 0/1 | 0.0000% |
| craps | odds pass 6 | 0/1 | 0.0000% |
| craps | odds dont 6 | 0/1 | 0.0000% |
| craps | odds pass 8 | 0/1 | 0.0000% |
The full ranked table is here, generated from this same file.
How the figures were produced
Exact enumeration or closed-form derivation in rational arithmetic, not simulation. Craps walks the 36 dice outcomes; baccarat walks the complete game tree of a real 416-card shoe drawing without replacement; hold'em equities enumerate all 1,712,304 boards per matchup. Each engine checks its own output against independently published figures before anything is exported, and a disagreement stops the build.
Where a figure could not be derived it is attributed rather than claimed, and the page says so. One game is absent on purpose: slot machines have no computable edge from outside, because the reel weightings are not published to anyone but the manufacturer.
The methodology page is the citation for this dataset.
Corrections
If a figure here is wrong it is wrong in a way you can demonstrate, which is the point of publishing the working. Tell us which bet and which number and the correction goes into the engine rather than the file, so it cannot come back.