Updated

Roulette odds, house edge and expected loss: a worked guide

Compare hit probability with net profit, calculate expected loss from total wagers, and understand why a losing streak does not make a win due.

A roulette bet can win often and still have a negative expected value. To check a claim about a betting system, separate the chance of a hit, the net payment for each outcome and the total money wagered. None of those calculations predicts the next spin.

The examples below assume fair, independent spins and standard payouts. A single-zero wheel has 37 pockets; a double-zero wheel has 38. Check the actual wheel and posted rules: special rules and different bets can change the calculation.

Calculate one bet from its possible outcomes

For a one-unit straight-up bet paying 35:1, a hit earns 35 units of net profit, and a miss loses one unit. The returned original stake is not another unit of profit.

On a single-zero wheel:

  • Hit probability: 1/37, about 2.70%.
  • Miss probability: 36/37.
  • Expected net result: (35 × 1/37) − (1 × 36/37) = −1/37 unit.

On a double-zero wheel, the same wager has expected net result (35 − 37)/38 = −2/38 unit, about −5.26% of the stake. The house edge is the positive percentage representing the expected loss. A positive house edge corresponds to a negative player expected value, not a “negative house edge.”

These are averages over the outcome distribution. A single one-unit bet does not literally lose 0.027 unit; it wins or loses according to the result.

A frequent hit is not the same as a profitable session

A standard red bet on a single-zero wheel wins on 18 pockets and loses on 19, including zero. With an even-money payout, its expected net result is (18 − 19)/37 = −1/37 unit per unit wagered. Its hit probability is about 48.65%, much higher than a straight-up bet, yet the expected loss per unit is the same under these rules.

The outcomes have different variability. A straight-up wager has infrequent large wins; an even-money wager has smaller individual swings. That does not guarantee a smoother or profitable session for a particular player. For combinations of wagers, calculate the net result of the whole combination on each pocket; one winning component can coexist with an overall loss.

Do not extend the usual 5.26% double-zero figure to every possible wager. The standard five-number basket covering 0, 00, 1, 2 and 3 pays 6:1. Its expected net result is (5 × 6 − 33)/38 = −3/38, about a 7.89% house edge. Actual posted rules control.

Use total wagers, not starting funds, for expected loss

Suppose a fictional paper exercise places twenty separate five-unit standard single-zero red wagers. Total stakes placed are 100 units. Expected loss under the assumptions is 100/37, about 2.70 units. This does not promise a 2.70-unit actual loss; the observed result could be substantially better or worse.

The starting amount is a different quantity. Reusing returned stakes and winnings can make total wagers much larger than the starting funds. A 100-unit starting balance does not cap expected-loss arithmetic at 100 units of turnover if the exercise repeatedly wagers the same money. Conversely, the calculation is not a prediction that a particular starting balance will last a certain time.

For a paper comparison, record each stake, the outcome’s net change and cumulative stakes placed. Do not count a returned stake as profit or use only the number of winning spins to assess the result.

At least one hit is not the probability of finishing ahead

Consider a hypothetical paper exercise with forty one-unit bets on the same single number, using a fair single-zero wheel and the standard 35:1 payout. Assume all forty bets are placed regardless of earlier outcomes; this is an arithmetic exercise, not a recommendation to keep playing.

The probability of at least one hit is one minus the probability of forty misses: 1 − (36/37)^40, approximately 66.6%. That is not a 66.6% chance of profit. One hit earns 35 units of net profit on that wager, but the other 39 misses lose 39 units. The session still ends four units down.

With H hits across these forty wagers, net result is 35H − (40 − H) = 36H − 40 units. No hits means minus 40; one hit means minus 4; two hits means plus 32. Finishing ahead therefore requires at least two hits under these exact assumptions. Do not substitute the probability of “something winning” for the probability that the whole session ends positive.

You can check any proposed system the same way on paper: state the wheel and payout, list every stake, count returned stakes separately from profit, and define the outcome being measured. A simulation reporting a high hit rate may be answering a different question from the one you care about. A winning example does not remove the negative expected value, and stopping or changing stakes would require a different session calculation.

What a doubling sequence leaves out

Doubling after a loss does not change the next spin’s probability. On a standard single-zero even-money bet, the chance of six specified consecutive losses is (19/37)^6, about 1.83%. That is the chance for one specified sequence, not the chance of encountering such a run somewhere in a longer session.

Starting at one unit, six losses cost 1 + 2 + 4 + 8 + 16 + 32 = 63 units. The next proposed stake is 64 units. Funding both those losses and that next wager requires 127 units, and the table must permit the next wager. These figures illustrate exposure; they are not a recommended bankroll or a safe loss allowance.

A later win is not guaranteed before funds or limits prevent another wager. A fixed run of previous losses does not make red, black or any number due. A stop rule can limit the amount you choose to risk; it does not remove the house edge.

Keep the purpose clear

Tink’s Toolchest offers the casino companion and session features described on its product page. It cannot predict independent outcomes or guarantee recovery of losses. You can audit every example here on paper without gambling.

If you choose to play, use only money you can afford to lose, keep a time and loss limit, and do not borrow or increase your allocation to chase losses. Understanding expected value is a way to recognize risk, not a method for earning reliable income.

Sources and revision

Updated October 28, 2026: corrected the sign of house edge, added the five-number exception, distinguished returned stake from profit and turnover from starting funds, and replaced reassuring progression language with finite-funds arithmetic. Original publication date retained.

Updated November 4, 2026: added a forty-wager example distinguishing at least one hit from a profitable session. Hypothetical arithmetic independently checked; original publication date retained.

← Back to the strategy lab