The Wheel Does Not Remember the Last Spin
A player walks up to a roulette table and sees a screen showing the last twenty winning numbers. Three of them were 17. The player reasons that 17 is now less likely, or more likely, depending on their theory. Neither theory changes anything. A roulette wheel does not keep a memory of where the ball landed, and a physical wheel with no bias produces the same probability for every pocket on every spin. The question 'what are the best numbers to put on roulette' is really a question about whether any number can be better than another. The answer sits in how the wheel is built, not in which numbers have appeared recently.
Every pocket on a European wheel is the same size, and the ball has no preference. The same is true on an American wheel, which simply has one extra pocket. The numbers printed on the pockets do not influence the physics. A 7 is not luckier than a 32. What differs between bets is not which number you choose but how many numbers you cover and how the payout is structured against that coverage.
Where the House Edge Lives
Roulette's built-in advantage does not come from any number being worse than another. It comes from the payout for a winning bet being slightly smaller than the true odds of that bet winning. On a European wheel there are 37 pockets. A straight-up bet on a single number pays 35 to 1. If the payouts reflected the true odds, a single-number win would pay 36 to 1, because there are 36 losing numbers and one winning number. The missing unit is the house edge.
The same structure applies to every roulette bet. A split bet covers two numbers and pays 17 to 1, but the true odds against winning are 17.5 to 1 on a European wheel. A street bet covers three numbers and pays 11 to 1, while the true odds are 11.33 to 1. The pattern continues across the layout. The operator keeps the difference between the payout and the true odds, and that difference is the same rate for every bet on the same wheel.
RTP and house edge are complements: they sum to 100% for the same bet under the same rule set. If a bet on a European wheel carries a house edge of roughly 2.For a small fraction of coin-in, the RTP for that same bet is roughly the same.30%. The American wheel adds a double zero, taking the pocket count to 38. The extra pocket increases the house edge on the American layout because the payouts stay the same while the chance of winning any given bet falls slightly.
| Bet type | Numbers covered | Payout | True odds on a European wheel | What the gap means |
|---|---|---|---|---|
| Straight up | 1 | 35 to 1 | 36 to 1 | One unit of the true odds is withheld |
| Split | 2 | 17 to 1 | 17.5 to 1 | Half a unit of the true odds is withheld |
| Street | 3 | 11 to 1 | 11.33 to 1 | A third of a unit of the true odds is withheld |
| Corner | 4 | 8 to 1 | 8.25 to 1 | A quarter of a unit of the true odds is withheld |
| Six line | 6 | 5 to 1 | 5.17 to 1 | A fraction of a unit of the true odds is withheld |
No number on the layout is priced differently from any other. The edge is in the payout schedule, and the payout schedule is the same for every straight-up number, every split, and every street. Choosing 17 instead of 23 changes nothing about the expected rate of loss.
What a Single Session Looks Like Against the Long Run
The long-run average is not a forecast for an evening at the table. RTP is a statistical property that emerges over a very large number of spins. In a short session, a player can be ahead or behind by an amount that looks nothing like the house edge. Suppose, as an example, a player makes 100 straight-up bets of one unit each on a European wheel. The expected loss from the house edge is about 2.7 units over those 100 bets, but the actual result will almost never be exactly that. The player might hit two numbers and finish well ahead, or hit none and lose all 100 units. Neither outcome proves the wheel is fair or rigged, and neither outcome changes the edge on the next spin.
Variance is the term for how results spread out around the average. Roulette is a high-variance game, especially on straight-up bets. A single-number bet wins rarely and pays a large multiple when it does. That creates wide swings in a short session. A player who covers more numbers per spin, such as with an even-money bet, trades the size of the rare win for a higher frequency of smaller wins. The expected loss per unit wagered stays the same on the same wheel, but the shape of the session changes.
Decisions Per Hour Change Total Exposure
The number a player chooses does not change the edge, but the speed of play changes the total amount wagered, and therefore the total expected loss. A player who places one unit on a single number each spin and plays 40 spins per hour has wagered 40 units in that hour. A player at a faster table, or a player who covers several numbers with separate bets, wagers more in the same period. The edge applies to each unit wagered, so total expected loss scales with total action.
As an example, suppose a player bets one unit per spin on a European wheel for 200 spins. The expected loss is about 5.4 units. If the same player bets five units per spin across five straight-up numbers, the expected loss is about 27 units over the same 200 spins, because the total wagered is five times larger. Nothing about the numbers chosen changed the rate. The amount at risk changed the total.
This is the practical answer to the question about the best numbers. The decision that matters is not which pocket to choose but how much total money goes across the table, how long the session runs, and what the payout schedule says about each bet's true odds.
Why Progression Systems Do Not Remove the Edge
A common response to the house edge is to change the size of the bets rather than the numbers chosen. The most familiar version is the martingale: double the stake after every loss, so that the first win recovers all previous losses plus one unit. The logic sounds complete, but it only works if the player has an unlimited bankroll and the table has no maximum bet. Neither condition holds in practice.
Progression systems redistribute outcomes. They turn many small losses into a pattern of mostly small wins punctuated by occasional large losses. The total expected loss across all bets remains the product of the house edge and the total amount wagered. A doubling progression increases the total amount wagered faster than flat betting does, so the expected loss grows faster, not slower.
As an example, suppose a player starts with a one-unit bet and doubles after each loss: 1, 2, 4, 8, 16, 32, 64, 128. After seven consecutive losses the next stake would be 128 units, and the player would have already lost 127 units. A table maximum or a depleted bankroll stops the progression before the recovery bet can be placed. The wheel does not know the player is due for a win, and no number is more likely to appear because the player has lost seven times in a row.
No betting pattern changes the built-in edge of a negative-expectation game. The edge is in the payout schedule, and the payout schedule does not change when the stake changes.
What a Player Can Choose
The choices that genuinely vary under a player's control are the wheel type, the bet type, the speed of play, the session length, and the bankroll sizing. A European wheel has 37 pockets and a single zero. An American wheel has 38 pockets with a single and double zero. The extra pocket increases the house edge on the American layout, so choosing a table with fewer zero pockets, where available, changes the expected rate of loss.
Bet type also matters. All bets on the same wheel carry the same house edge under the standard payout schedule, but some bets are slower to resolve and some produce more variance. A player who wants a longer session for the same bankroll can choose bets that cover more numbers and pay smaller multiples. A player who wants the chance of a large single win can choose a straight-up bet and accept the wider swings. Neither choice makes a number better; it changes the shape of the session.
Bankroll sizing is a method, not a strategy for beating the wheel. Setting a loss limit before play and treating the bankroll as the cost of the session keeps the total exposure within a chosen range. Stopping rules, such as ending the session after a set number of spins or after a set loss, do not change the edge but do change how much total money is wagered.
FAQ
Are some roulette numbers statistically more likely to win?
On a fair wheel with correctly sized pockets, every number has the same probability on every spin. A physical wheel can develop a bias from wear or damage, but that is a defect in a specific wheel, not a property of a number. Casinos monitor and maintain wheels, and a player cannot reliably identify a biased wheel from a short sequence of results.
Do hot and cold number displays mean anything?
Displays of recent winning numbers are records of what has already happened. They do not change the probability of the next spin. A number that has appeared several times recently is not more likely to appear again, and a number that has not appeared for a long time is not due. Each spin is an independent event on a fair wheel.
Does covering more numbers reduce the house edge?
Covering more numbers changes the frequency of wins and the size of the payout, but the house edge on each unit wagered stays the same on the same wheel under the standard payout schedule. A player who covers 18 numbers with even-money bets wins more often but loses the same rate per unit wagered as a player who covers one number with a straight-up bet.
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.