Roulette Odds Explained
What this guide covers
Roulette's house edge is often quoted as a single number — 2.70% for European wheels, 5.26% for American — but understanding where that number actually comes from, how it applies (almost) uniformly across every bet type, and what it means for a real session takes a closer look at the underlying math. This guide walks through true odds versus payout odds, a full bet-by-bet breakdown, the concept of variance and why short sessions can look nothing like the long-run average, and how roulette's math compares to other casino games. For the rules that structure these bets, see roulette betting options explained and roulette rules explained.
True odds versus payout odds
The house edge in roulette comes from a simple, consistent gap: the payout odds on every bet are set slightly worse than the true mathematical odds of winning. Take a straight-up bet on a single number in European roulette: there are 37 pockets, so your true odds of winning are 36-to-1 against. But a winning straight-up bet only pays 35:1. That one-unit gap, present in some form across every bet type on the wheel, is exactly where the casino's edge comes from — not through any hidden manipulation, but through a payout table that's calibrated to pay slightly less than a bet's true, fully fair odds would require.
Deriving the house edge formula
The house edge for any bet can be calculated with a simple formula: house edge = (true odds against winning − payout odds) ÷ (true odds against winning + 1), expressed as a percentage. For a European straight-up bet, true odds against are 36-to-1 and the payout is 35-to-1, giving (36 − 35) ÷ (36 + 1) = 1/37 ≈ 2.70%. This same formula, applied to any other standard European bet, produces the same 2.70% result, which is exactly why the house edge is described as "nearly uniform" across the wheel — the payout table for every standard bet type is calibrated using the same underlying logic relative to that bet's own specific true odds.
Why the edge is (almost) the same on every bet
On a standard European wheel, nearly every bet type — straight up, split, street, corner, six line, dozens, columns, red/black, odd/even, high/low — carries the same 2.70% house edge, because the payout-to-true-odds gap scales consistently with each bet's own probability. This means, contrary to intuition, that no single standard bet type on a European wheel is mathematically "better" than another in the long run — a $10 bet on a single number and a $10 bet on red carry the exact same expected loss over time, despite feeling like completely different kinds of bets.
A full bet-by-bet breakdown
| Bet type | Numbers covered | True odds against | Payout | House edge |
|---|---|---|---|---|
| Straight up | 1 | 36 to 1 | 35 to 1 | 2.70% |
| Split | 2 | 17.5 to 1 | 17 to 1 | 2.70% |
| Street | 3 | 11.33 to 1 | 11 to 1 | 2.70% |
| Corner | 4 | 8.25 to 1 | 8 to 1 | 2.70% |
| Six line | 6 | 5.17 to 1 | 5 to 1 | 2.70% |
| Dozen | 12 | 2.08 to 1 | 2 to 1 | 2.70% |
| Column | 12 | 2.08 to 1 | 2 to 1 | 2.70% |
| Red/Black, Odd/Even, High/Low | 18 | 1.05 to 1 | 1 to 1 | 2.70% |
| American top line | 5 | 6.6 to 1 | 6 to 1 | 7.89% |
Notice the pattern: every payout across the standard European bets is set very slightly below the true odds of winning, in a way that produces exactly the same 2.70% edge regardless of how many numbers the bet covers. The American top-line bet is the sole standard exception, breaking that pattern in the house's favor.
The exception, explained mathematically
The American wheel's five-number "top line" bet (0, 00, 1, 2, 3) is the one common exception to the uniform house edge, and it's worth understanding exactly why. True odds against winning are 33-to-5 (equivalent to 6.6-to-1), but the bet pays only 6-to-1. Running this through the same formula used above — (6.6 − 6) ÷ (6.6 + 1) ≈ 7.89% — shows why this specific bet is calibrated so much less favorably than every other bet on the table: its payout simply doesn't scale correctly against its true odds the way every other bet's does, making it worth avoiding specifically, even on tables where American-format wheels are your only option.
RTP: the other way to express the same number
House edge and RTP (return to player) are two ways of describing the same underlying math, just inverted. A European roulette bet with a 2.70% house edge has a 97.30% RTP — meaning that, over a very large number of spins, the game is mathematically expected to return 97.30% of all money wagered back to players collectively, with the remaining 2.70% representing the house's structural edge. Neither figure describes what happens in any single session (which is governed by variance, covered below) — both are long-run averages that only become statistically meaningful over a very large number of spins.
Expected loss per spin, in real terms
A 2.70% house edge translates concretely to an expected loss of $2.70 for every $100 wagered on a European wheel, spread across however many individual bets that $100 covers. This is a mathematical average, not a per-spin guarantee — any individual spin either wins fully according to its payout odds or loses fully, with nothing in between. The 2.70% figure only becomes a reliable predictor of actual results once you're looking at a genuinely large number of spins, which is a distinction worth understanding clearly, since it explains why individual sessions can and regularly do deviate substantially from this average in either direction.
Why this makes roulette different from blackjack
In blackjack, the house edge depends heavily on the decisions a player makes during a hand — correct basic strategy meaningfully reduces it compared to playing by feel, sometimes by more than a full percentage point. In roulette, there are no decisions after a bet is placed that change the odds of that specific bet; the wheel and ball determine the outcome entirely independently of anything the player does. This is the core reason no roulette betting system (see the strategy guide) can lower the house edge the way skillful decision-making can in blackjack — the house edge in roulette is fixed the moment a bet is placed, with no subsequent player input capable of altering it.
Comparing house edges across wheel formats
| Wheel format | House edge (standard bets) | House edge with en prison/la partage |
|---|---|---|
| European | 2.70% | 1.35% (even-money bets only) |
| French | 2.70% | 1.35% (even-money bets only, usually included by default) |
| American | 5.26% | Not applicable (rule doesn't exist on American wheels) |
This comparison makes the practical stakes of format choice concrete: playing American roulette instead of European, all else equal, roughly doubles your expected loss rate for the same total amount wagered. Choosing a European or French table with en prison or la partage active, where available, meaningfully improves on even the standard European figure for even-money bets specifically.
How roulette's house edge compares to other casino games
| Game | Typical house edge |
|---|---|
| European roulette | 2.70% |
| American roulette | 5.26% |
| Blackjack (basic strategy) | ~0.5% |
| Baccarat (banker bet) | ~1.06% |
| Craps (pass line) | ~1.41% |
| Slots (typical range) | 2%–10%+ |
Roulette sits in a middle position among common casino games — worse value than blackjack played with correct basic strategy or baccarat's banker bet, but generally better than the typical house edge on slot machines. Its appeal isn't primarily about having the best available odds in the casino; it's about a simple, fast, purely luck-based format with no decisions to master, which is a genuinely different value proposition than a skill-sensitive game like blackjack.
Understanding variance alongside house edge
House edge describes the long-run average outcome, but variance describes how much any individual session can deviate from that average — and roulette's variance differs substantially by bet type. A straight-up bet, with its rare 35:1 payout, has very high variance: most spins lose the full stake, but an occasional win produces a large payout relative to the bet size. An even-money outside bet has much lower variance: it wins close to half the time, with smaller, more frequent swings in either direction. Both bet types share the same 2.70% house edge on a European wheel, but they produce very different session-to-session experiences purely because of this variance difference.
Why short sessions can look nothing like the long-run average
Because of variance, a short session — even one covering a few dozen spins — can easily produce results well above or well below the mathematically expected outcome purely by chance. A player betting flat on red might win eight of ten spins in one session and lose seven of ten in the next, despite the underlying true win probability being roughly 48.6% in both cases. This isn't evidence that the house edge fluctuates or that a particular session was somehow "rigged" in either direction — it's simply what statistical variance looks like over a small sample, and it's exactly why the house edge only becomes a reliable predictor over a genuinely large number of spins, not a handful.
The law of large numbers in practice
The house edge's long-run reliability rests on a foundational statistical principle called the law of large numbers: as the number of independent trials (spins, in this case) increases, the average result converges more and more closely toward the true expected value. Over ten spins, results can look wildly different from the 2.70% average; over ten thousand spins, results converge much more tightly around it. This is precisely why casinos, which see an enormous volume of spins across all their tables and players combined, can rely on the house edge as a stable, predictable revenue source, even though any individual player's individual session remains fundamentally unpredictable.
Calculating the probability of a losing streak
Understanding streak probability helps explain why progression betting systems like Martingale carry real tail risk, covered fully in the strategy guide. On a European wheel, an even-money bet wins with roughly 48.6% probability per spin (18/37). The probability of losing five spins in a row is (1 − 0.486)^5 ≈ 3.8%; ten in a row drops to roughly 0.14%. These probabilities sound small in isolation, but across a long session involving many bets, the odds of encountering at least one such streak at some point rise considerably — which is exactly the mathematical basis for why betting systems that require recovering losses in a single win eventually run into trouble.
The gambler's ruin concept
"Gambler's ruin" is a classical probability result describing what happens when a player with a finite bankroll repeatedly bets against a game with a house edge (or even a perfectly fair, zero-edge game): given enough attempts, the probability of eventually losing the entire bankroll approaches certainty, since the player's finite funds will eventually run out during an unlucky stretch even in a fair game, and a house-edge game only accelerates this. This is a purely mathematical result, not a comment on any specific player's luck or skill — it's simply what happens when a bounded amount of money repeatedly faces any negative (or even neutral) expected-value proposition over a long enough time horizon.
Standard deviation as a practical planning tool
Beyond the house edge itself, standard deviation — a statistical measure of how widely results typically spread around the average — gives a more complete picture of what to expect from a session. A roulette session's standard deviation depends heavily on which bets you're placing: sessions built around high-variance straight-up bets will show much wider swings in results than sessions built around low-variance even-money outside bets, even when both carry an identical 2.70% house edge. Thinking in terms of both numbers together — the average you should expect over the long run, and how widely any given short session might reasonably deviate from it — gives a more realistic picture than the house edge figure alone.
Simulating roulette outcomes
For anyone mathematically curious, running a large-scale simulation — tens of thousands of virtual spins, tracking results against a fixed 2.70% house edge — is one of the clearest ways to see the house edge and variance concepts play out concretely. Early in a simulated run, results often swing well above or below the expected average; as the simulation extends into the thousands of spins, the running average reliably converges toward the theoretical 2.70% expected loss rate. This kind of simulation is a far more statistically reliable way to understand roulette's real mathematical behavior than any individually played real session, regardless of how that particular session happened to turn out.
Why the house edge doesn't mean you'll always lose
It's worth being explicit about something the math above implies but doesn't always feel intuitive: the house edge describes a long-run average, not a guarantee that any individual session results in a loss. Plenty of individual sessions — even a large share of them, depending on bet type and session length — end in a genuine profit purely due to ordinary variance, exactly as the math above would predict. The house edge guarantees the casino's long-run advantage across its enormous overall volume of play; it says nothing certain about any single player's single session, which remains genuinely unpredictable in the short run even though the long-run average is mathematically fixed.
Putting the math to practical use
None of this math requires memorization to enjoy roulette, but it's worth internalizing three practical takeaways: European (or French, ideally with en prison or la partage) wheels are meaningfully better value than American wheels; no bet type on a European wheel offers genuinely better odds than another, so bet-type choice is really about variance preference rather than value-hunting; and short-session results, whether good or bad, say very little about the underlying math, which only reveals itself reliably over a much larger number of spins than most individual sessions actually involve.
Working through the expected value formula directly
Beyond the house-edge shortcut formula covered earlier, it's worth seeing expected value calculated directly for a specific bet, since it's the same underlying concept applied more explicitly. For a $1 straight-up bet on a European wheel: there's a 1/37 probability of winning $35 (the payout, not counting your returned stake) and a 36/37 probability of losing your $1 stake. Expected value = (1/37 × $35) + (36/37 × −$1) = $0.9459 − $0.9730 = −$0.0270. That −$0.027 result is exactly the 2.70% house edge, expressed in dollars per dollar wagered rather than as a percentage — the same number, just arrived at through the raw probability-weighted calculation rather than the shortcut formula.
Why 37 (or 38) pockets specifically matters
The house edge's exact size traces directly back to pocket count. A hypothetical roulette wheel with exactly 36 pockets and no zero at all would produce a perfectly fair, zero-house-edge game for every standard bet — a straight-up bet's true odds (35-to-1 against) would exactly match its 35:1 payout. The zero pocket (and the second zero on American wheels) is what creates the gap between true odds and payout across every bet type simultaneously, without requiring the payout table itself to be adjusted bet by bet. This is why the house edge shifts so predictably between formats: adding one more non-paying pocket (European to American) mechanically worsens the odds on every single bet by a consistent, calculable amount.
Combinatorics behind the split bet's odds
A split bet covers two adjacent numbers with a single wager, so on a 37-pocket European wheel, your true odds of winning are (37 − 2)/2 = 17.5-to-1 against. The bet pays 17:1. Running the house-edge formula: (17.5 − 17) ÷ (17.5 + 1) ≈ 2.70%. The same consistent gap-to-true-odds ratio shows up here as it did for the straight-up bet, just scaled to a bet covering twice as many numbers — this consistency is what makes the "nearly uniform house edge" property of European roulette a mathematical certainty rather than a coincidence of the specific numbers involved.
Combinatorics behind the street and corner bets
A street bet covers three numbers in a row on the layout, giving true odds of (37 − 3)/3 = 11.33-to-1 against, versus an 11:1 payout — again working out to almost exactly 2.70% once run through the formula. A corner bet covers four numbers meeting at a single point on the grid, with true odds of (37 − 4)/4 = 8.25-to-1 against, against an 8:1 payout, producing the same result once more. Each of these bets covers a different fraction of the wheel, and each has a correspondingly different payout — but the ratio between true odds and payout, and therefore the resulting house edge, stays essentially fixed across all of them.
Combinatorics behind dozens and columns
Dozens and columns each cover twelve numbers — a full third of the non-zero layout — with true odds of (37 − 12)/12 ≈ 2.08-to-1 against, against a 2:1 payout. This produces the familiar 2.70% figure once more, and it's a useful example for seeing how the house edge formula scales smoothly from bets covering a single number all the way up to bets covering a third of the entire wheel, without ever meaningfully favoring one bet size over another in terms of long-run value.
Why betting on multiple dozens simultaneously doesn't create an edge
A common intuition among newer players is that betting on two of the three dozens simultaneously — covering 24 of the 36 non-zero numbers — should meaningfully improve your odds of winning that round. It does increase your probability of winning a given spin (up to roughly 64.9%, accounting for the zero), but it doesn't create a positive expected value, because you're now risking two units to win one unit's net profit rather than one unit to win two. Working through the math: you win $1 net (getting $2 back on your $2 total stake, having wagered $1 on each of two dozens) with 24/37 probability, and lose your full $2 stake with 13/37 probability. Expected value = (24/37 × $1) + (13/37 × −$2) = $0.6486 − $0.7027 = −$0.0541 per $2 wagered, which works out to the same 2.70% house edge applied to your total stake — covering more of the wheel changes your win frequency and typical bet size, not your underlying expected return rate.
Common misconceptions about odds and probability
A handful of odds-related misconceptions recur often enough among roulette players to be worth naming directly. Believing a number is "due" after not appearing for many spins (the gambler's fallacy) misunderstands that each spin is fully independent of previous results. Believing a "hot" number that's hit recently is more likely to hit again similarly misreads independence in the opposite direction. And believing that combining many small bets across the layout somehow changes the underlying house edge, rather than simply changing which specific numbers you're covering, is another common misread of how probability aggregates across multiple simultaneous bets — as the multi-dozens example above demonstrates concretely.
Why the Kelly criterion doesn't apply to roulette
Kelly criterion is a well-known bankroll-sizing formula used in genuinely positive-expected-value betting scenarios (some forms of professional sports betting, or blackjack card counting under favorable conditions) to calculate the mathematically optimal fraction of a bankroll to wager on each opportunity. It has no meaningful application to standard roulette, because Kelly's formula is built around exploiting a genuine positive edge — and roulette, played through any standard bet, never offers one. Applying Kelly-style bankroll math to a negative-expected-value game like roulette simply confirms what's already known: the mathematically "optimal" fraction to wager on a guaranteed-negative-edge proposition is zero, which isn't a useful practical betting strategy so much as a restatement of the house edge itself.
How odds intersect with betting system marketing
Betting system marketing — covered more fully in the strategy guide — frequently glosses over the odds math covered in this guide, presenting a system's short-term win rate without acknowledging that the underlying probability of any individual bet never changes. Understanding the true-odds-versus-payout-odds gap directly is a useful defense against this kind of marketing, since it makes clear that no rearrangement of bet sizes across a sequence of spins can touch a figure (the house edge) that's fixed the moment a bet's payout table is set, entirely independent of staking pattern.
Odds as a lens for comparing table offers
Two tables offering "European roulette" can differ meaningfully in expected value once en prison or la partage enters the picture, even though both are technically the same wheel format. Applying the odds framework covered in this guide — checking true odds, payout, and any special rule adjustments before playing — turns table selection from a purely aesthetic choice (interface design, dealer presentation) into one that can be evaluated on genuine expected-value grounds, which is a meaningfully more useful lens than simply picking whichever table loads first or looks the most appealing.
Frequently asked questions
Is any single roulette bet type actually better value than another on a European wheel? No, with the sole common exception of the American top-line bet — otherwise the house edge is essentially uniform (2.70%) across the whole standard betting layout.
Why does the house edge stay the same even though payouts differ so much between bet types? Because the payout for each bet type is calibrated against that specific bet's own true odds — a bet with worse true odds (like a straight-up number) gets a correspondingly bigger payout, and the proportional gap between true odds and payout stays consistent across most bet types.
Does understanding the odds table help me win more? It won't change your long-run expected outcome, but it does make clear why no bet type or combination of bets on a standard European wheel offers a genuine edge over another — useful context for deciding how you want to bet, even though it doesn't tilt the math in your favor.
What's the difference between house edge and RTP? They're two ways of describing the same underlying math — RTP (return to player) is simply 100% minus the house edge, expressed as the long-run expected percentage of wagered money returned to players collectively.
Does a 2.70% house edge mean I'll lose 2.70% of my money every session? No — that figure is a long-run average across a very large number of spins. Any individual session is subject to variance and can land well above or below that average purely by chance.
How does roulette's house edge compare to slots? Roulette (2.70% European, 5.26% American) generally offers better odds than the typical range for slot machines, which commonly runs from around 2% up to 10% or higher depending on the specific game.
Why do some bets feel riskier than others if they share the same house edge? This is a variance difference, not a house-edge difference — a straight-up bet wins rarely but pays a lot when it does, while an even-money bet wins often for a smaller payout, producing very different session experiences despite an identical long-run expected loss rate.
What is the gambler's ruin concept, and does it apply to roulette? It's a classical probability result showing that a player with a finite bankroll repeatedly betting against a house-edge game will, given enough attempts, eventually approach a near-certain probability of losing that bankroll entirely — it applies to roulette exactly as it does to any other house-edge game.
Can a losing streak really run to ten spins or more? Yes — while any single ten-loss streak on an even-money bet is individually unlikely (roughly 0.14% probability), across a long session involving many bets, encountering at least one such streak at some point becomes considerably more likely than the per-attempt figure alone suggests.
Is there a mathematically "safest" way to bet on roulette? Choosing low-variance even-money outside bets produces smaller, more frequent swings in results than high-variance inside bets, but both carry an identical house edge on a European wheel — "safer" here refers to variance, not to a better long-run expected outcome.
Does betting on more numbers at once improve my odds of profit? It increases your probability of winning a given spin, but proportionally increases your total stake and reduces your net profit per win, leaving the underlying expected value — and house edge — unchanged.
Why doesn't the Kelly criterion apply to roulette betting? Kelly criterion calculates optimal bet sizing for genuinely positive-expected-value opportunities. Since no standard roulette bet offers a positive edge, the formula's own logic simply confirms that the mathematically "optimal" wager on the game is zero — not a usable staking strategy.
Is the house edge calculated differently for inside bets versus outside bets? The same underlying formula (comparing true odds against payout odds) applies to every bet type — it just produces the same 2.70% result for nearly every standard European bet, regardless of how many numbers that specific bet covers.
Does the house edge apply per spin or per session? It's a long-run average rate applied to total amount wagered over time, not a guaranteed outcome for any single spin or session — an individual spin either wins fully or loses fully, with the 2.70% figure only emerging reliably across a large number of spins.
Can two tables both labeled "European roulette" have different actual house edges? Yes — if one offers en prison or la partage on even-money bets and the other doesn't, their effective house edges differ meaningfully (1.35% versus 2.70% on those specific bets) despite sharing the same underlying 37-pocket wheel format.
Is roulette's math the same across RNG and live dealer tables? Yes — the odds and house edge come from the wheel's pocket count and the payout table, not from whether the wheel is a physical object filmed for a live stream or a certified algorithm generating outcomes digitally; both formats produce mathematically identical results for the same wheel type.
