Roulette odds: why every bet has the same house edge
Roulette is unusual among casino games: it offers a wide, visually elaborate betting layout in which every single bet is mathematically identical. There is no clever corner of the felt, no bet the table would rather you avoided, and no combination that improves your position. There is exactly one decision that changes your odds, and it happens before you place a chip.
The one number that matters
A European wheel has 37 pockets: the numbers 1 through 36 plus a single zero. An American wheel has 38, adding a double zero. Every payout on both wheels is identical — a straight-up number pays 35 to 1 on either.
That identical pricing is where the house edge lives. The payouts are calculated as though the wheel held only 36 pockets. Back a single number on a European wheel and you are paid 35 to 1 on a 1-in-37 chance; on an American wheel, the same 35 to 1 on a 1-in-38 chance. The zeros are the difference between the true odds and the paid odds, and the second one roughly doubles it.
Every bet, both wheels
The table below is generated from the payout schedule the practice game implements. The right-hand columns are the house edge — the share of every unit staked that the table keeps on average.
| Bet | Numbers covered | Pays | European edge | American edge |
|---|---|---|---|---|
| Straight up | 1 | 35 to 1 | 2.70% | 5.26% |
| Split | 2 | 17 to 1 | 2.70% | 5.26% |
| Street | 3 | 11 to 1 | 2.70% | 5.26% |
| Corner | 4 | 8 to 1 | 2.70% | 5.26% |
| Line | 6 | 5 to 1 | 2.70% | 5.26% |
| Dozen or column | 12 | 2 to 1 | 2.70% | 5.26% |
| Red, black, odd, even, high, low | 18 | 1 to 1 | 2.70% | 5.26% |
Seven bet types, seven identical numbers. This is the single most useful fact about roulette and the one most consistently ignored: there is no such thing as a good roulette bet or a bad one on a given wheel. There is only a good wheel and a bad one.
What your bet choice actually changes
If expected return is fixed, the bet you choose still decides something real: how your results are distributed. A straight-up number wins about once in 37 spins and pays 36 times your stake when it does. An even-money bet wins a little under half the time and pays double. Both hand back 97.30 percent of everything staked on a European wheel over the long run, but the paths there could hardly be less alike.
That is a genuine choice, and it is worth making deliberately rather than by habit. If you want a long session from a fixed number of credits, even-money bets will deliver it. If you want the possibility of a large swing from a small stake, straight-up numbers will. Neither is smarter.
Why systems cannot help
Roulette attracts more betting systems than any other casino game, and the reason they all fail is contained in the table above. A system is a rule for choosing which bets to place and how much to stake. But every bet already returns the same fraction of its stake, and each spin is independent of the one before it. Combining identical-edge bets in any sequence, at any stake size, produces a combination that carries the same edge.
The martingale — doubling after each loss — illustrates this clearly. It produces many small wins and occasional catastrophic losses, and people who try it briefly usually report that it works. It does not: it has simply moved the loss into the tail, where it happens rarely and all at once. The average is untouched.
Watching previous results fails for a related reason. A wheel that has landed red eight times running is exactly as likely to land red on the ninth spin. The pockets do not know what they did last time, and a scoreboard showing recent results is a record, not a forecast.
The rules that would change the maths
Two real-world variations genuinely improve the even-money bets, and it is worth knowing them so you can recognise a better table if you ever see one.
- La partage
- When zero lands, half of a losing even-money stake is returned. This halves the edge on those bets to roughly 1.35 percent on a single-zero wheel and makes them, uniquely, better than everything else on the layout.
- En prison
- When zero lands, an even-money stake is held for the next spin rather than lost. If that spin wins, the stake is returned. The effect is broadly similar to la partage.
Neither rule is implemented on this practice table, which uses the plain schedule shown above. Where one of them is available on a real table, it is the only circumstance in which one roulette bet is genuinely better than another.
How we calculated this
Every figure in this guide is derived from the payout schedule the practice game implements, using the script committed atapps/games/scripts/derivations/roulette-bet-edges.mjs. Run it to reproduce the table above.
The method is one line of arithmetic. For a bet covering c pockets on a wheel ofn pockets, paying p to one, the expected return per unit staked is(c ÷ n) × (p + 1), and the house edge is one minus that. A corner bet on a European wheel is therefore (4 ÷ 37) × 9 = 0.9730, an edge of 2.70 percent. An even-money bet is (18 ÷ 37) × 2 = 0.9730 — the same figure, which is the result the table is built to demonstrate.
Both wheel layouts use their real pocket sequences in the game rather than numbers in order, and zero and double zero are correctly excluded from every outside category. The ratio between the two wheels' edges is 5.26 ÷ 2.70 = 1.947, which is the precise sense in which the American wheel is “almost twice as bad”. The la partage and en prison figures are standard published results for those rule variants and are not implemented in this game; they are included for completeness and marked as such.
Common questions
Which roulette bet has the best odds?
None of them — every bet on a given wheel carries an identical house edge. On a single-zero wheel that is 2.70 percent whether you back one number at 35 to 1 or eighteen numbers at even money. The payouts are scaled so that each bet gives back exactly the same fraction of what you stake, so choosing between them is a choice about volatility, not value.
How much worse is the American wheel?
Almost twice as bad. Adding the double zero takes the edge on every bet from 2.70 percent to 5.26 percent — a ratio of 1.947. Nothing else about the layout changes and every payout stays the same, so a straight-up number still pays 35 to 1 while being measurably less likely to land.
Do roulette betting systems work?
No system changes expected return. Doubling after a loss, waiting for a colour to repeat, or covering most of the layout all leave the same 2.70 or 5.26 percent in place, because spins are independent and every bet already shares one edge. Systems reshape the distribution of results — more frequent small wins traded for rarer large losses, or the reverse — without moving the average.
Why do outside bets lose when zero lands?
Zero and double zero belong to none of the outside categories: not red, not black, not odd, not even, not high, not low. That exclusion is the mechanism that creates the house edge on those bets. Some real-world tables soften it with la partage or en prison rules, returning half a losing even-money stake; this practice table does not implement either.
Is this guide describing real-money roulette?
The mathematics is the same as any single- or double-zero wheel, but the table it derives from uses practice credits with no cash value. There is nothing to deposit, nothing to withdraw, and no link to a real-money operator anywhere on this site.