Start with the bet, not the wheel
Most roulette questions start with the wheel, but expected value starts with a single wager. You pick a bet, the ball lands somewhere, and either you are paid or the chips are taken. The question is what that bet is worth before the ball stops, when you do not yet know the outcome. That worth is the expected value, and it is built entirely from two things: the payout the table promises and the probability that the promise is kept.
Roulette is a clean place to learn this because the probabilities are fixed by the wheel's physical layout. A European wheel has 37 pockets and a single zero. An American wheel has 38 pockets with a single and double zero. Every pocket is equally likely on a fair spin, so the probability of any bet is the number of pockets it covers divided by the total pockets on that wheel. The payout is printed on the table layout. Nothing else enters the calculation.
How a bet's expected value is built
Expected value is the average result per unit wagered if the same bet were repeated an enormous number of times. The formula is simple: multiply each outcome by its probability, then add the results. For a roulette bet there are only two outcomes: you win the payout, or you lose the stake.
Take a single-number bet as an example. Suppose you wager 1 unit on one number on a European wheel. The payout for a straight-up number is printed on the layout; call it P units of profit for a 1-unit stake. The ball lands on your number with probability 1 in 37, and on any other pocket with probability 36 in 37. The expected value is (P × 1/37) plus (−1 × 36/37), where the −1 is the stake you lose on a miss. If P is 35, the arithmetic gives (35/37) − (36/37), which is −1/37 of a unit per spin. That negative result is the house edge expressed as an amount rather than a percentage.
The same method works for every bet on the layout. A split covers two pockets, a street covers three, a dozen covers twelve. You multiply the printed payout by the fraction of the wheel the bet covers, then subtract the fraction it misses. The answer is always a small negative number per unit wagered on that wheel, because the payout is always slightly short of what a fair game would pay for the same coverage.
Where RTP and house edge meet
Return to Player and house edge are two ways of stating the same arithmetic. For the same bet under the same rule set, they sum to 100%. If a bet's expected value works out to a loss of a small fraction of each unit wagered, the RTP is whatever remains. One number is the player's side of the same equation, the other is the operator's side.
What changes between wheels is the size of that fraction. The extra double-zero pocket on an American layout means every bet covers a slightly smaller share of the wheel than it would on a European layout, while the payout stays the same. The expected value therefore moves further from zero on the American wheel. The relationship is visible in the arithmetic: the probability of winning falls, the payout does not rise to compensate, and the gap between the two is the edge.
| Wheel type | Pockets | Zero pockets | Effect on a straight-up bet's probability | Effect on expected value |
|---|---|---|---|---|
| European | 37 | 1 | 1 in 37 chance of a hit | Small negative amount per unit wagered |
| American | 38 | 2 | 1 in 38 chance of a hit | Larger negative amount per unit wagered |
The table does not need percentages to make the point. The American wheel gives each bet one more losing pocket and the same payout, so the expected value per unit wagered is worse. The exact figures are printed nowhere on the layout, but the arithmetic is fully determined by the pocket count and the payout, both of which the player can read at the table.
Why short sessions look nothing like the average
Expected value is a long-run property. It says nothing about what happens on the next spin, the next ten, or the next hundred. A player who hits a straight-up number on the third spin is ahead for the session even though the bet carries a negative expected value. A player who misses for an hour is behind by more than the average predicts. Both outcomes are ordinary variance, and variance dominates any session short enough to remember.
Volatility is the reason. Roulette bets differ sharply in how their outcomes are distributed. A straight-up bet rarely wins but pays a large multiple when it does; an even-money bet wins close to half the time but pays only the stake back as profit. Both carry a negative expected value on the same wheel, but their session-by-session behaviour feels completely different. The straight-up player experiences long losing runs punctuated by occasional large hits. The even-money player sees frequent small wins and losses that drift downward slowly.
| Bet type | Payout structure | Win frequency | Session feel | Expected value |
|---|---|---|---|---|
| Straight-up number | Large multiple of stake | Rare | Long dry spells, occasional big hits | Negative on every wheel |
| Even-money (red/black, odd/even) | Stake returned as profit | Close to half, minus zero pockets | Frequent small swings | Negative on every wheel |
| Dozen or column | Moderate multiple of stake | Roughly one spin in three, minus zero pockets | Between the two extremes | Negative on every wheel |
None of these bet choices changes the expected value from negative to positive. They change the shape of the ride, not the direction of the long-run average. A player choosing a straight-up bet is choosing higher variance, not a better expected value.
Decisions per hour change total exposure, not the edge
Expected value is per unit wagered, so the total expected loss over a session is the per-unit figure multiplied by the total amount staked. Total amount staked is not the same as the bankroll brought to the table. A player with 100 units who bets 1 unit per spin and sees 60 spins in an hour has staked 60 units, not 100. The expected loss for that hour is the per-unit expected value times 60.
This is where game speed enters. Roulette moves at the pace of the wheel and the table, and the number of spins per hour depends on how full the table is, how fast the dealer works, and whether the player bets every spin. A player who sits out spins stakes less; a player who bets every number that comes up stakes more. Neither decision changes the expected value of any individual bet. What changes is how many times the negative expectation is applied.
Bet sizing works the same way. Suppose a player wagers 5 units per spin instead of 1. The expected value per unit is unchanged, but the expected loss per spin is five times larger because the stake is five times larger. The arithmetic is linear: double the stake, double the expected loss, and leave the edge exactly where it was. Bankroll sizing is a method for controlling how long a session can last and how much total exposure accumulates, not a method for improving the expected value of the bets themselves.
Progression systems redistribute outcomes
Progression systems are betting patterns that change the stake after wins or losses. A common one doubles the stake after a loss, aiming to recover the lost amount and gain a small profit on the next win. The pattern feels systematic, and in the short run it often produces frequent small wins. What it does not do is change the expected value of the underlying bets.
The reason is that each spin is independent, and the expected value of a bet does not depend on what happened on the previous spin. Doubling the stake after a loss means the next bet is larger, so its expected loss is larger by exactly the same multiple. The progression redistributes when losses occur: instead of many small losses, the player experiences occasional large losses that wipe out a run of small wins. The total expected loss over any fixed number of spins is the per-unit edge times the total amount staked, and the progression does not reduce the total amount staked. It concentrates it.
What a progression does change is the shape of the session. A doubling progression produces many small wins and rare, severe losses. A player who quits before the severe loss occurs will report that the system worked. A player who stays until it hits will report the opposite. Both are describing the same negative-expectation game seen through different slices of variance. The system has not removed the edge; it has moved the loss to the tail of the distribution.
How to read the numbers for yourself
The expected value of any roulette bet can be worked out at the table from three pieces of information: the number of pockets on the wheel, the number of pockets the bet covers, and the payout printed on the layout. The probability of winning is coverage divided by total pockets. The expected value is the payout times that probability, minus the stake times the probability of losing. The result is always a small negative number per unit wagered.
If the player wants the operator's published figures rather than their own arithmetic, the place to look is the game's information panel or rules page, where the return-to-player percentage for that specific game and rule set is typically stated. That figure is the complement of the house edge for the same game. It is a long-run statistical average, not a promise about the next session, and it applies to the game as configured, not to any individual spin or any particular player's evening.
For the player who wants to know what a session is likely to cost, the calculation is expected value per unit times total units staked. Total units staked is the only part of that equation the player controls directly, through bet size and the number of spins played. The edge itself is fixed by the wheel and the payout table, and no betting pattern changes it.
FAQ
Does expected value change depending on which roulette bet I choose?
On the same wheel, most standard bets carry a similar negative expected value per unit wagered, because the payout is scaled to the coverage. The exceptions are special bets or rule variants that pay differently for the same coverage, and those are printed on the layout or in the game rules. The bet choice changes variance far more than it changes expected value.
Can I calculate expected value without knowing the payout?
No. The payout is one of the two inputs, along with the probability of winning. The payout for every bet is printed on the table layout or shown in the game's paytable. Without it, you can describe the probability of winning but not the expected value, because expected value is the payout weighted by that probability minus the stake weighted by the probability of losing.
Why does my actual result never match the expected value?
Expected value is an average over an enormous number of repetitions. Any session short enough to play in one sitting is dominated by variance, which is the random spread of outcomes around that average. The expected value tells you the direction and rate of the long-run drift; it does not predict the next spin or the next hour.
Gamble responsibly. If you or someone you know has a gambling problem, call the helpline printed in the footer below for support.
Play responsibly
Gambling should be treated as paid entertainment, never as a way to earn income or recover losses. 1-800-MY-RESET (1-800-697-3738)
18+ or 21+ depending on where you are; follow the minimum age that applies to you.
This article is general information about how these mechanics work. It is not legal advice and not a recommendation to gamble or to use any particular operator. Availability and legality differ by jurisdiction — check the rules that apply where you are.