Roulette odds are calculated by comparing the pockets covered by a bet with every pocket on the wheel. Payouts look generous because less likely bets pay more, but they are set slightly below fair odds. That difference is the house edge.
Three terms that should not be confused
Probability is the chance that a bet wins. Payout is the amount returned after a win. Expected value combines all possible wins and losses into a long-run average per wager. A bet can win frequently and still have negative expected value.
European roulette probabilities
| Bet | Pockets | Probability | Profit payout |
|---|---|---|---|
| Straight up | 1 | 2.70% | 35:1 |
| Split | 2 | 5.41% | 17:1 |
| Street | 3 | 8.11% | 11:1 |
| Corner | 4 | 10.81% | 8:1 |
| Six line | 6 | 16.22% | 5:1 |
| Dozen/column | 12 | 32.43% | 2:1 |
| Even money | 18 | 48.65% | 1:1 |
Divide covered pockets by 37 for European roulette or 38 for American roulette. Multiply by 100 to express the result as a percentage.
Calculating expected value
For a one-unit straight-up European bet, one outcome earns 35 units and 36 outcomes lose one. Expected value is (1/37 × 35) + (36/37 × −1) = −1/37, or about −0.027 units. The expected loss is 2.70% of the amount wagered.
A dozen wins on 12 pockets for two units and loses on 25. Its expectation is (12/37 × 2) − (25/37 × 1) = −1/37 again. Different standard bets change volatility and hit frequency, but not the European house edge.
House edge is not a session prediction
A 2.70% edge does not mean every 100-unit session loses exactly 2.70 units. It means the mathematical average over a very large amount wagered approaches that proportion under the assumed rules. A player can win 50 units, lose 100 or finish level in a short session.
The expected cost scales with total action. One unit wagered on 100 spins creates 100 units of action even if the same chips circulate through wins. At 2.70%, theoretical expected loss is 2.70 units. Increasing stake size or speed increases expected cost proportionally.
House edge and return to player
Return to player (RTP) is the complement of house edge under standard definitions. European roulette with a 2.70% edge has theoretical RTP of 97.30%. American roulette with a 5.26% edge has RTP of 94.74%. These are long-run theoretical figures, not a guarantee that a session returns that percentage.
Why zero matters
Even-money groups divide numbers 1–36 evenly, but zero sits outside both sides. It is neither red nor black, odd nor even, low nor high. That imbalance gives the casino one additional winning outcome on a European wheel. Double zero gives it two on an American wheel.
Variance and volatility
A straight-up wager has high variance: most decisions lose, while a rare win produces 35 units of profit. Red/black has lower variance because almost half of decisions win one unit. If both have the same edge, their average expected loss per unit is equal, but the distribution of session outcomes differs considerably.
Progression systems manipulate that distribution further. Martingale produces many small targets and occasional large drawdowns. Paroli risks accumulated winnings during a positive sequence. Neither changes the probability or payout of the underlying roulette bet.
Conditional and streak probabilities
On a fair European wheel, the chance that red loses once is 19/37. The chance it loses eight specified times in a row is (19/37)^8, about 0.48%. That number applies to a particular block of eight future spins. Across a long session with many overlapping blocks, the chance of seeing at least one eight-loss run is higher.
After seven losses, the eighth spin is not more likely to win. Its probability remains 18/37 for red because independent spins have no memory. The probability of completing a streak and the probability of the next result answer different questions.
Can bonuses change the calculation?
A genuine rebate, loss-back offer or promotion can change expected value, but only after wagering requirements, maximum bets, excluded games, withdrawal restrictions and counterparty risk are included. Advertising language is not part of the formula. Evaluate the written rules and applicable law.
Use the numbers in practice
- Confirm wheel type and special rules.
- Count the pockets covered by the wager.
- Calculate win probability.
- Check profit payout and returned stake.
- Calculate expected value.
- Multiply the edge by total planned action.
- Choose limits based on affordable loss, not a target profit.
Use our roulette odds calculator to verify standard bets, then compare theoretical probability with observed results in the free simulator.
Further calculations worth understanding
Implied fair payout
Fair total return is the reciprocal of win probability. A European straight-up bet has probability 1/37, so its fair total return is 37 units and fair profit would be 36. The actual total return is 36, leaving one unit of expected shortfall across the full 37-outcome model.
Confidence versus certainty
Observed frequency moves toward theoretical probability as samples grow, but not in a smooth line and never by obligation. A 10,000-spin sample can still deviate. Statistical intervals quantify plausible sampling variation; they do not certify that a future block must correct the past.
Risk of ruin
Risk of ruin asks whether ordinary variance can exhaust a finite bankroll before a goal or time boundary. It depends on bankroll, stake, wager distribution and stopping rules. House edge alone does not describe it. Two strategies with equal expectation can have very different ruin probabilities.
Why covering more numbers is not a loophole
Covering 30 numbers raises hit frequency but requires multiple chips and produces a small or negative net result on many hits. Always calculate return after all simultaneous losing chips. “Winning spin” and “profitable spin” are not synonyms.
Audit any roulette claim in five lines
- Write total pockets and covered pockets.
- Calculate the exact win probability.
- Write total stake and net profit after a hit.
- Calculate expected value across win and loss outcomes.
- Multiply edge by total amount wagered, not initial bankroll.
If a claim cannot supply those inputs, a high win percentage has little analytical value. Where software skips spins, define whether skipped outcomes affect configuration and whether they were selected before the result.
Frequently asked question: can expectation ever be positive?
A defective physical wheel, exploitable information, incorrectly priced promotion or operational error could change the model, but each requires specific evidence and may raise legal or contractual issues. A generic staking progression does not. The responsible default for a regulated standard wheel is the published negative expectation.
How to use this research
Treat the worked examples as explanations, not forecasts. Reproduce the assumptions with a free tool, change one variable at a time and keep losing as well as winning trials. A short favourable run cannot validate a strategy, while a short unfavourable run cannot measure its full distribution. For any real-money decision, verify local law and operator rules, set an affordable limit before play and never use a progression to recover money that has already been lost.