Two dice produce 36 equally likely outcomes, and every bet on a craps table is priced from that one table of 36. Enumerated across all of them — the same figures Wizard of Odds publishes independently — the pass line keeps 1.4141% of what is staked on it and the any-seven bet keeps 16.6667% — a twelve-fold difference on the same roll of the same dice. Learning how to play craps is mostly learning which side of that spread you are standing on, because the layout does not tell you.
The 36 outcomes everything is priced from
Two six-sided dice give 6 × 6 = 36 ordered results. Nothing about craps is more important than the shape of that distribution, because the totals are not equally likely even though the outcomes are. Seven can be made six ways and two only one, which is why seven is simultaneously the best total and the worst one depending on when it arrives.
| Total | Ways | Probability | True odds against |
|---|---|---|---|
| 2 or 12 | 1 | 2.78% | 35 to 1 |
| 3 or 11 | 2 | 5.56% | 17 to 1 |
| 4 or 10 | 3 | 8.33% | 11 to 1 |
| 5 or 9 | 4 | 11.11% | 8 to 1 |
| 6 or 8 | 5 | 13.89% | 6.2 to 1 |
| 7 | 6 | 16.67% | 5 to 1 |
Those eleven totals and their counts are the entire probability model of the game. Every figure later in this guide is computed from them by enumeration rather than quoted, which means you can rebuild any of it with a sheet of paper and the table above.
How a round is played
A round has two phases, and the same total means opposite things in each. New Jersey’s rules of the game set out the sequence formally at N.J.A.C. 13:69F-1.4; in practice it runs like this.
with point p, chance of repeating before a 7 = ways(p) ÷ (ways(p) + 6)
point 6: 5 ÷ 11 = 45.45% · point 4: 3 ÷ 9 = 33.33%
total win probability = 244/495 = 49.29%
house edge = 251/495 − 244/495 = 7/495 = 1.4141%
Every bet, priced
Running the same enumeration across the layout produces the spread the table hides. The bets nearest the shooter’s hands — the proposition and hardway boxes in the middle — are the ones that cost the most.
Four things stand out. Free odds is exactly zero, because it is paid at the true odds computed in the table above. The don’t pass is fractionally better than the pass line at 1.3636% against 1.4141%, since barring the 12 removes a loss without adding a win. Buying the 4 or 10 for a 5% commission cuts that bet from 6.6667% to 1.6667%, which is the single largest improvement available from changing nothing but how the same wager is booked. And any seven, the most tempting box on the layout because seven is the most common total, is the worst bet on the table at 16.6667% — the payout of 4 to 1 against true odds of 5 to 1.
Why free odds matter more than anything else
The pass line’s 1.4141% is charged on the pass-line stake. Free odds add money to the round at no edge at all, so the edge on the combined wager falls as the odds multiple rises. That is the only lever in craps that changes the price rather than the volatility.
same expected cost, spread over a larger total wager:
+ 1× odds (£10): £0.141 ÷ £20 = 0.71% of money at risk
+ 2× odds (£20): £0.141 ÷ £30 = 0.47%
+ 5× odds (£50): £0.141 ÷ £60 = 0.24%
That distinction gets flattened in a lot of craps advice, so it is worth stating plainly: taking maximum odds does not reduce what you expect to lose on the round. It buys a larger swing in both directions at the same price. Whether that is worth having depends on bankroll rather than on arithmetic.
Does the enumeration hold up in practice?
Everything above is arithmetic, and arithmetic is only worth as much as its assumptions — here, that the dice are fair and every outcome equally likely. Casinos are regulated on exactly that point: New Jersey’s dice standards at N.J.A.C. 13:69F-1.2 specify dimensions, tolerances and markings precisely so the 36 outcomes stay equally likely.
CasinoDrop does not track a live craps table, so we cannot show you 36 outcomes converging on our own tape. What we can show is the same method applied to a game we do record. Single-zero roulette is enumerated the same way — 37 pockets, every bet priced at a 2.7027% edge — and we log every settled spin on the original Lightning Roulette table.
Across 5,600 recorded spins, all ten groupings we checked landed inside the range chance produces at that sample size, including the zero at 2.61% against a designed 2.70%. That is not evidence about craps dice, and it is not offered as such. It is evidence that counting a finite outcome space and comparing it against what a real table produces is a method that survives contact with reality — and it is the same method used for the craps figures above.
What a betting plan can and cannot do
- Which edge you pay. Sticking to the line bets and place 6 or 8 keeps every wager under 1.6%.
- How a bet is booked — buying the 4 or 10 for commission cuts that edge from 6.67% to 1.67%.
- Variance, through the odds multiple, at an unchanged expected cost.
- Exposure per round, by not covering boxes you have no price for.
- The 36 outcomes. Every roll is independent, so no total is ever due.
- The sign of the expected value. Every bet except free odds is priced below true odds.
- The result of a progression. Doubling after losses meets the table maximum, and the edge is unchanged on every stake along the way.
- Anything about how the dice are thrown at a regulated table.
Betting systems are the most-discussed and least-useful part of craps. A progression rearranges when losses arrive; it does not alter the price of any individual bet, because each roll is settled against the same 36 outcomes as the last. The one genuine improvement available is choosing cheaper bets, and that choice is made by reading the chart above rather than the layout.
We do not track a live craps table, so the honest thing to hand you is the method rather than a craps feed. The Lightning Roulette board carries every settled spin we record against the 37-pocket wheel — the same finite-outcome arithmetic used above, and the one place you can watch it hold up round by round. The crash board does the same for a game whose payout rule is a single published formula.
Open the Lightning Roulette tracker →
Or the crash board, with hit rate against the published model →
- CasinoDrop does not track craps. Every craps figure here is enumerated arithmetic over the 36-outcome space, not a measurement of a real table, and none of it is evidence about any particular casino’s dice.
- The enumeration assumes fair dice and equally likely outcomes. That assumption is what regulators test for, and it is an assumption rather than something we verified.
- House edges move with the paytable. The figures here use the payouts named beside each bet; a table paying 2:1 rather than 3:1 on the field, or a different commission on buy bets, produces a different number for that bet.
- The roulette comparison is a check on the method, not on craps. It shows that counting a finite outcome space matches what a real table produces on a game we record. It says nothing about dice.
- That roulette sample is the original Lightning Roulette table only, 5,600 spins over about three days. It describes those spins rather than the game’s lifetime or any other table of it.
- House edge is a long-run average per unit staked. It does not describe a session, and the spread around it over an evening is far wider than the edge itself.
- Nothing here predicts a roll. Every roll is independent of the last, and no total becomes due because it has been absent.
FAQs
Bet the pass line before the come-out roll. A 7 or 11 wins immediately, a 2, 3 or 12 loses, and any other total becomes the point. The shooter then tries to repeat that point before rolling a 7. Once a point is set you may back the bet with free odds, which is paid at true odds and carries no house edge.
Exactly 7/495, or 1.4141%, enumerated over all 36 two-dice outcomes. The pass line wins 244 times in 495 and pays even money, and the gap between those two facts is the whole edge. The don’t pass is slightly better at 3/220, or 1.3636%, because barring the 12 removes a loss without adding a win.
Free odds, which carries no house edge at all because it is paid at the true odds of the point repeating. It can only be placed behind an existing line bet once a point is set. Among the bets you can make on their own, the don’t pass at 1.3636% and the pass line at 1.4141% are the cheapest, followed by place 6 or 8 at 1.5152%.
Any seven, at 16.6667%. Seven is the most common total, which is what makes the bet tempting, but it is paid at 4 to 1 against true odds of 5 to 1. The 2 or 12 at 30 to 1 is next at 13.8889%, then any craps and the hard 4 or hard 10 at 11.1111%. The proposition boxes in the middle of the layout are the expensive ones.
Not in money terms. Free odds is priced fairly, so adding it leaves your expected loss unchanged at 1.4141% of the pass-line stake while increasing the total you have at risk. That lowers the edge measured against money wagered — from 1.41% to about 0.24% at 5x odds — but it buys a larger swing in both directions rather than a cheaper game.
Because it is the most likely total, at 6 ways in 36, and the round changes what that means. On the come-out roll a 7 wins the pass line immediately. Once a point is set, the same 7 ends the round and loses it. Nothing about the dice changes between those two moments; only the phase of the round does.
The payout and the commission. Placing the 4 or 10 pays 9 to 5 and carries a 6.6667% house edge. Buying it pays true odds of 2 to 1 with a 5% commission, which brings the edge down to 1.6667%. You are betting on exactly the same outcome; only the way the wager is booked changes, and it is the largest single improvement available on the layout.
No. A progression changes when losses arrive rather than what any bet costs, because every roll is settled against the same 36 outcomes as the last. Doubling after a loss also runs into the table maximum, and the house edge is unchanged on every stake along the way. The only real improvement available is choosing bets with a lower edge.
Regulated tables are built to prevent it. New Jersey’s dice standards at N.J.A.C. 13:69F-1.2 specify dimensions, tolerances and markings, and table rules require the dice to strike the far wall. The arithmetic in this guide assumes fair dice and equally likely outcomes, which is exactly the property regulators test for rather than something a player can verify at the table.
Out of 36 equally likely outcomes: 7 comes up 6 ways, 6 and 8 five ways each, 5 and 9 four ways, 4 and 10 three ways, 3 and 11 two ways, and 2 and 12 one way each. That works out to 16.67% for a 7 down to 2.78% for a 2 or a 12, and every price on the layout is computed from those counts.
Playing within a limit
Craps is a negative-expectation game on every bet except free odds, and craps tables settle rounds quickly, so turnover accumulates faster than most players picture. At a 1.4141% edge, £10 on the pass line costs about 14 pence per resolved round on average — which is a small number until it is multiplied by an evening. Decide the loss limit before the first come-out. CasinoDrop’s responsible gambling guide covers deposit caps, timeouts and self-exclusion, and the National Council on Problem Gambling lists support in the US.
- Everything comes from 36 outcomes. Seven can be made six ways and two only one. Every price on the layout follows from that table, and you can rebuild any of them yourself.
- The spread is twelve-fold. Pass line 1.4141%, don’t pass 1.3636%, place 6 or 8 1.5152% — against 16.6667% on any seven. Same dice, same roll.
- Free odds is the only zero-edge bet in the casino. It lowers the edge measured against total money wagered without lowering the expected loss in pounds. That is variance, not a discount.
- How a bet is booked matters. Buying the 4 or 10 for a 5% commission moves it from 6.6667% to 1.6667% without changing what you are betting on.
- Systems rearrange losses; they do not reprice bets. Each roll is settled against the same 36 outcomes, and the table maximum ends every progression eventually.
- CasinoDrop Lightning Roulette tracker — first-party, the 5,600 recorded spins used to check the enumeration method; original table, 2026-08-15 to 2026-08-18
- Lightning Roulette round feed — first-party, the JSON behind the observed grouping shares
- N.J.A.C. 13:69F-1.2 — primary, New Jersey Administrative Code: dice standards, dimensions and tolerances
- N.J.A.C. 13:69F-1.3 — primary, New Jersey Administrative Code: receipt, storage and inspection of dice
- N.J.A.C. 13:69F-1.4 — primary, New Jersey Administrative Code: craps rules of the game, and the come-out and point phases
- N.J.A.C. 13:69F-1.6 — primary, New Jersey Administrative Code: permissible wagers and payout odds
- Craps basics — reported, Wizard of Odds: independent house-edge figures that the enumeration here reproduces
- Craps appendix 1 — reported, Wizard of Odds: bet-by-bet analysis
- Craps appendix 2 — reported, Wizard of Odds: combined pass line and odds figures
- Dice probability basics — reported, Wizard of Odds: the two-dice outcome distribution
- The craps pass line and odds — reported, University of Wisconsin–Madison statistics department: an independent derivation of the 244/495 pass-line probability
- Help and treatment — primary, National Council on Problem Gambling
