Roulette Complete Guide: European, American, French, Odds and House Edge
A probability-led roulette guide covering wheel types, inside and outside bets, payouts, house edge, French rules, live tables and multiplier variants.
A probability-led roulette guide covering wheel types, inside and outside bets, payouts, house edge, French rules, live tables and multiplier variants.
Roulette is one of the easiest casino games to recognise and one of the easiest to misunderstand. The wheel looks random, the betting layout looks busy and dozens of named systems promise to create order from the sequence. The mathematics is much simpler than the mythology: every standard roulette bet can be understood from the number of pockets it covers, the payout offered and the number of pockets on the wheel.
That makes roulette a useful game for learning the difference between probability and prediction. A correct probability tells you how often an outcome should occur over an enormous sample. It does not tell you what the next spin will be.
The first comparison should therefore be the wheel and paytable, not the last twenty results. European single-zero, American double-zero and French rule sets can look similar on screen while creating materially different expected returns.
European roulette uses a wheel with numbers 1 through 36 plus a single zero, for 37 pockets in total. American roulette normally adds a double-zero pocket, creating 38 pockets. French roulette also uses a single-zero wheel but is often paired with French terminology and rules such as La Partage or En Prison on selected even-money bets.
Those structural differences change the house edge before any betting system enters the conversation.
On a single-zero wheel, a straight-up number has a probability of 1 in 37 and normally pays 35 to 1. The player receives 35 units in winnings plus the returned stake when successful. If there were no zero, a 35:1 payout across 36 equally likely numbered outcomes would be mathematically balanced. The extra zero creates the casino advantage.
On a double-zero wheel, the same normal 35:1 straight-up payout is spread across 38 possible pockets. The additional losing pocket increases the house edge.
Some modern American-style wheels add a third zero or another branded symbol. That changes the denominator again. Never infer the mathematics from a red-and-black layout alone; count the actual outcome pockets and read the rules.
European roulette has 37 pockets. Consider a one-unit straight-up bet. There is one winning pocket that produces 35 units of profit and 36 losing pockets that each lose one unit. Expected value is:
(1/37 ร 35) - (36/37 ร 1) = -1/37, approximately -2.70%.
That same approximately 2.70% house edge applies to the normal standard bets on a conventional single-zero table because the paytable scales with the number of covered outcomes.
American double-zero roulette has 38 pockets. For a normal straight-up bet:
(1/38 ร 35) - (37/38 ร 1) = -2/38, approximately -5.26%.
That is why wheel type matters more than a progression system. No stake sequence can remove the extra zero.
One American bet can be worse still: the five-number basket covering 0, 00, 1, 2 and 3 traditionally pays 6:1, creating a house edge of about 7.89%. If that bet appears, its mathematics should be checked separately rather than assuming every wager on the table has the same edge.
The formula also explains why a giant-looking payout does not automatically make a wager attractive. Payout must always be compared with the probability of winning.
La Partage can reduce the cost of even-money bets on a single-zero wheel. When zero wins, an eligible wager such as red/black, odd/even or high/low loses only half rather than the entire stake. Under standard conditions, that reduces the effective house edge on those even-money bets to about 1.35%.
En Prison is a different mechanism with a related purpose. Instead of immediately losing the full eligible even-money bet when zero appears, the stake can remain "in prison" for the next spin under the table's specified settlement rule.
Exact implementation matters. A game labelled French Roulette does not guarantee that every familiar French rule is present. Verify whether La Partage or En Prison is active, which wagers qualify and how a second zero or repeat zero would be handled if the product uses a non-standard mechanism.
French terminology can also appear on a table that has no special zero rule. Words on the layout are not enough; the game help page controls the comparison.
Inside bets are placed on the numbered grid and cover relatively small sets of pockets.
A straight-up bet covers one number. Standard payout: 35:1. This has the highest usual standard payout and the lowest hit frequency among ordinary single-position bets.
A split covers two adjacent numbers sharing a line on the layout. Standard payout: 17:1.
A street covers a horizontal row of three numbers. Standard payout: 11:1.
A corner covers four numbers that meet at one intersection. Standard payout: 8:1.
A six-line bet covers two adjoining streets, six numbers in total. Standard payout: 5:1.
On many tables, special combinations involving zero are also available. Their coverage and payout should be confirmed because the top of an American layout is not identical to a European one.
Outside bets cover broader groups.
Dozens cover 1-12, 13-24 or 25-36 and normally pay 2:1. Columns also cover 12 numbered pockets and normally pay 2:1.
Red/black, odd/even and low/high normally pay 1:1. Each covers 18 numbered pockets, but zero and double zero where present are outside those groups.
That missing zero coverage is the important point. "Even money" describes the payout, not a true 50% win probability on a roulette wheel with zero.
A player can also combine multiple outside bets, but the total stake is then the sum of all positions. Covering more of the wheel can make wins more frequent while reducing net profit on covered outcomes or creating outcomes that merely offset part of the stake.
A common mistake is to focus only on hit frequency. Red wins far more often than a single number, but it pays much less. A single number wins rarely but pays 35:1. Under standard rules on the same wheel, those bets can carry the same house edge even though the session feels completely different.
Variance is therefore separate from house edge. Straight-up number betting produces long losing sequences punctuated by larger wins. Broad outside bets produce more frequent wins and losses of smaller relative size. The expected percentage cost can be identical while the path is not.
This distinction matters for budgeting. A player who dislikes long losing runs may prefer broader coverage, but should not call it mathematically better unless a rule such as La Partage actually changes the expected value.
Volatility also changes when several bets are combined. Five single numbers are not one bet; they are five simultaneous straight-up wagers. The chance of any one of them winning rises, but so does the total amount staked each spin.
The Martingale doubles the stake after a loss. Fibonacci and Labouchere use other progressions. D'Alembert changes stakes more gradually. Reverse progressions increase stakes after wins.
These systems change the distribution of stake sizes. They do not change pocket probability or payout.
A Martingale can generate many small winning sequences followed by an occasional very large loss. Real tables also have maximum bets and real bankrolls are finite. A sufficiently long losing run can make the next required stake impossible, exceed the table maximum or exceed the player's chosen budget.
The important question is not whether a progression can win the next five spins. Any betting pattern can have a winning short run. The question is whether it changes the expected value of a bet with a built-in negative expectation. It does not.
Progressions can also create a misleading psychological record: many sessions may end with one-unit profit while a single failed sequence erases numerous earlier gains. Session count is not the same metric as money won and lost.
If black has appeared eight times, red is not automatically due. On independent RNG spins, the probability of the next colour is governed by the wheel model, not by a balancing obligation to recent results.
The same warning applies to hot and cold numbers. Random sequences naturally create clusters. A number appearing several times in a short sample does not prove that it has become more likely on the next fair spin.
Physical roulette uses mechanical objects, but licensed games are designed and monitored so the outcome mechanism is not supposed to become predictably biased. Historical boards are primarily records, not forecasting devices.
A useful test is to imagine hiding the spin history. If the rule and wheel are unchanged, the probability of each fair next-spin outcome is unchanged merely because the display is hidden.
RNG roulette uses software to determine outcomes. The interface may animate a wheel after the result is generated or combine random outcome generation with a visual simulation. What matters from a fairness standpoint is the applicable game rule and random-number control.
Great Britain's RTS 7, for example, requires random outcomes to be acceptably random, distributed according to the theoretical probabilities and not adaptively changed in response to play.
A useful RNG roulette comparison checks the wheel type, paytable, special rules, game speed, minimum and maximum stake and whether the rules disclose relevant likelihood or RTP information.
A beautiful animation is not evidence about the mathematics. Likewise, a fast instant result is not evidence that the game is less random. Presentation speed and outcome generation are separate concepts.
Live roulette streams a physical wheel from a studio. The presenter opens betting, spins the wheel and ball, closes betting, and the system resolves wagers after the result is detected and confirmed.
The live experience adds operational factors:
A player should still start with wheel type and rules. Live video does not neutralise a double zero.
Roulette also differs from seated live blackjack because one wheel can often serve many remote bettors simultaneously. That makes table availability less about an empty seat and more about limits, language, variant and market availability.
Auto Roulette uses an automated wheel rather than a presenter-led table. Some versions use multiple wheels or very rapid round timing.
Speed changes the number of betting opportunities per hour, which can change how quickly money is put at risk even when the house edge per unit staked is unchanged.
This is a crucial distinction. House edge is commonly expressed per amount wagered, not per hour. If a format allows twice as much turnover in the same period, the expected monetary loss can rise even if the percentage edge is identical.
When comparing speed, instant or auto roulette, include pace in the risk assessment rather than focusing only on the published RTP.
Games such as Lightning Roulette and other multiplier formats modify standard roulette by selecting numbers for enhanced payouts or adding bonus mechanics. They are not simply ordinary single-zero roulette with free extra prizes.
A multiplier variant can reduce the normal straight-up payout, add a fee, select enhanced numbers by RNG, or change other parts of the return distribution. The headline maximum multiplier is therefore insufficient for comparison.
Read the exact paytable. Ask what a non-multiplied straight-up number pays, how special numbers are selected, whether the selection mechanism is random, and what stated RTP or house edge applies to the wager you intend to use.
The same principle applies to other branded or game-show hybrids. A wheel, bonus round or multiplier can create a genuinely different product even if a familiar roulette grid remains visible.
European and French interfaces often include a racetrack display representing the physical order of numbers on the wheel. Named sectors can include Voisins du Zรฉro, Tiers du Cylindre and Orphelins.
These are bundles of individual chip placements, not new probability laws. The total amount staked and exact covered numbers should be visible before confirmation.
A racetrack shortcut can place several chips with one tap. That convenience creates a practical risk: the player may look only at the denomination of one chip instead of the total stake across all underlying positions.
Neighbours bets similarly cover a selected number plus nearby pockets on the physical wheel. The number of neighbours and total stake can vary by interface.
A roulette table can advertise a low minimum while imposing different maximums by bet type. Straight-up numbers may have one maximum; outside bets another. Progression systems can collide with those limits quickly.
Minimums affect budgeting too. A table with a 5-unit minimum is not equivalent to one with a 0.20-unit minimum for a player using the same entertainment budget. The house-edge percentage may be identical while the absolute amount exposed per spin is very different.
VIP tables may raise both minimums and maximums without changing the core wheel rules. A high minimum is a capacity and positioning feature, not evidence of a better return.
Some tables also enforce aggregate maximums across related bets. The maximum shown beside one position may not be the only limit that matters.
Casino bonus rules often assign contribution percentages. Roulette may contribute less than slots, contribute only on particular bets or be excluded from wagering entirely. That is a promotion rule, not a change in roulette's mathematical house edge.
Do not raise roulette stakes simply to complete a wagering requirement faster without checking maximum-bet restrictions. Promotions can also restrict low-risk betting patterns or simultaneous coverage.
The correct workflow is to read the game rules for the wheel mathematics and the bonus terms for promotional eligibility. Mixing those two documents leads to bad comparisons.
House edge becomes easier to understand when translated into turnover.
Suppose a player makes 100 one-unit red bets on standard single-zero European roulette. The total amount staked is 100 units. A 2.70% theoretical house edge implies an expected loss of about 2.70 units across an enormous number of equivalent 100-unit samples.
That does not mean a particular 100-spin session will lose exactly 2.70 units. It can finish far above or below that number. Expected value is an average, not a schedule.
If total turnover becomes 1,000 units instead of 100, the theoretical expected loss scales to roughly 27 units under the same rules. This is why round speed, repeated bets and multiple simultaneous chip positions matter.
La Partage changes that calculation for qualifying even-money bets because the expected cost of a zero result is reduced.
Roulette is a negative-expectation casino game under normal rules. Budgeting should therefore be based on entertainment spend rather than an income target.
Decide the maximum session loss before play. Then choose a realistic number of spins and average total stake per spin. If the intended progression requires repeatedly increasing bets after losses, calculate how many steps fit the budget and table limits before starting.
Remember that covering more numbers does not necessarily reduce risk. Placing chips on twenty individual numbers can expose far more money than one even-money bet.
Time matters. Fast roulette can cycle rapidly, and repeated small wagers can create more total turnover than an occasional larger wager on a slow live table.
Responsible-gambling tools such as deposit limits, time reminders, time-outs and self-exclusion are more useful when set before a stressful session rather than after a loss-chasing pattern begins.
Historical roulette literature includes attempts to identify biased physical wheels. In a modern regulated online live environment, relying on such a strategy is a very different proposition.
A meaningful physical bias would require a persistent mechanical deviation, a large and reliable sample, a stable wheel and enough access before the equipment is serviced or replaced. Studio operators also monitor equipment.
The existence of historical wheel-bias stories therefore does not justify treating a current live table as predictably biased. A few hundred spins showing an uneven histogram can occur through ordinary variance.
First verify market availability. A casino may list a provider globally while enabling different tables in different jurisdictions.
Second identify the exact game: European single zero, American double zero, French, auto, speed, multiplier or another format.
Third open the rules. Confirm payouts, zero rules, table limits, round pace and stated RTP or house edge.
Fourth assess the experience: mobile layout, chip controls, favourite-bet tools, racetrack access, video quality, language, statistics display and interruption handling.
Fifth separate entertainment features from mathematical value. A famous presenter, polished studio or enormous multiplier can make a format more engaging without making the base wager more favourable.
Finally record the verification date. Provider catalogues and casino lobbies change. A roulette title visible in one market today may not be available in another market or at a later date.
Under standard rules, European single-zero roulette has a lower house edge than conventional American double-zero roulette. That is a statement about long-run expected value, not a prediction for one session.
No, not on a fair independent RNG wheel. Previous outcomes do not create a debt that the next spin must repay.
On a conventional European single-zero table, standard inside and outside bets generally have the same house edge, although their variance differs. Special bets and non-standard paytables can be exceptions.
No staking progression removes the built-in edge. The more defensible approach is to select favourable rules, understand the paytable, limit total turnover and keep stakes within a fixed entertainment budget.
A properly operated live wheel is not intended to be predictably exploitable. Historical statistics can be interesting, but they do not make the next fair result knowable.
Not by itself. Standard bets on the same conventional wheel are usually priced to the same house edge. Covering more numbers mainly changes hit frequency, variance and total stake.
Roulette becomes easier to evaluate when every claim is brought back to the wheel, coverage and payout. Wheel type sets the baseline. The paytable prices the bet. Special rules can improve or alter expected return. Everything else, from streak charts to progression names, should be judged against those facts.