Roulette betting systems explained: Martingale, Fibonacci, d'alembert - and the mathematical proof none of them work

A complete, honest guide to the most popular roulette staking systems: how each one works, the real numbers behind why they can't overcome the house edge, and what they actually do achieve

Roulette betting systems have circulated for centuries, and Martingale, Fibonacci, and D'Alembert remain the three most widely discussed today. This guide explains exactly how each system works mechanically, walks through the underlying mathematical proof for why no staking pattern can change a game's built-in house edge, and - honestly - covers what these systems genuinely do accomplish, since it isn't nothing, even though it isn't beating the house.

The mathematical proof, stated plainly first

Before covering any individual system, the single foundational fact that every system guide should lead with: if every individual bet has negative expected value, no combination or sequence of those bets can produce positive expected value. This isn't an opinion, a simplification, or a matter of debate - it's a provable mathematical theorem. A European roulette spin carries a 2.70% house edge whether you bet £1 or £1,000, whether it's your first spin of the session or your hundredth, and regardless of what happened on the previous ten spins.

The real numbers behind even-money bets. Red/black, odd/even, and high/low are commonly described as "50/50" bets, but they aren't genuinely 50/50 - on European roulette, you statistically win these bets only 48.65% of the time and lose 51.35% of the time, with the zero pocket accounting for the gap. This 2.7-percentage-point difference is small enough to feel negligible session to session, but it's precisely what every betting system covered below runs into over a large enough sample of spins.

Why the "law of large numbers" is the mechanism that defeats every system. As the number of spins increases, your actual results converge closer and closer to the expected mathematical result - which for roulette means losing approximately 2.70% (European) or 5.26% (American) of everything wagered, regardless of the specific pattern in which you staked it. Short-term variance is real, and any individual session can produce a profit - the distinction is between winning in a single session (entirely possible, and common) and winning systematically over a long run through a reliable staking method (which the mathematics make effectively impossible).

Martingale: Double after every loss

The mechanic: Bet a fixed unit on an even-money outcome. After every loss, double your stake for the next spin. After a win, reset to your original base unit. The logic: since you eventually have to win, the eventual win recovers all prior losses plus one unit of profit.

A worked sequence: £1 (loss) → £2 (loss) → £4 (loss) → £8 (win) → reset to £1. The £8 win recovers the previous £1+£2+£4 = £7 in losses, plus £1 profit - exactly one base unit ahead, regardless of how long the losing streak lasted before the win arrived.

Why it fails despite the clean logic. The system requires an effectively unlimited bankroll and no table maximum to guarantee eventual recovery - in reality, both constraints exist. A genuinely unlucky losing streak (which, while individually unlikely on any given run, will eventually occur across enough sessions) can require bet sizes that either exceed your bankroll or hit the table's maximum stake before the recovering win arrives - at which point you're stuck, unable to continue the doubling sequence, having lost an amount far larger than the single unit of profit the system was chasing.

The Grand Martingale variant: Follows the same doubling-after-loss principle but adds one additional unit to each increased stake, escalating even faster than standard Martingale - a sequence of £1 (loss) → £3 (loss) → £7 (loss) → £15 (win) → reset to £1, each step doubling the previous stake plus one extra unit. This produces a larger profit per completed cycle but accelerates the bankroll and table-limit risk correspondingly faster.

Fibonacci: the sequence-based progression

The mechanic: Stakes follow the Fibonacci number sequence (1, 1, 2, 3, 5, 8, 13, 21...), where each number is the sum of the two preceding it. After a loss, advance one step forward in the sequence. After a win, retreat two steps back.

Why it's considered gentler than Martingale. Because the Fibonacci sequence grows more slowly than pure doubling, a losing streak requires meaningfully smaller total exposure to reach the same point in the progression - a 10-loss streak under Fibonacci requires staking a cumulative 55 units (following the sequence through its tenth step), compared to the 1,024 units a pure Martingale doubling sequence would demand after the same 10 consecutive losses.

Why it still doesn't overcome the house edge. The slower progression genuinely reduces the speed and severity of catastrophic bankroll depletion relative to Martingale, but it doesn't change the fundamental mathematics - you're still making a sequence of individually negative-expected-value bets, and the Law of Large Numbers still applies over enough spins regardless of how gently the stake size increases.

D'alembert: the slowest, "safest" progression

The mechanic: Choose a base stake. After every loss, increase your stake by exactly one unit for the next spin. After every win, decrease your stake by one unit. The stake never falls below your original base level.

A worked sequence, base stake £6: £6 (loss) → £7 (loss) → £8 (win) → £7 (loss) → £8 (win) → £7 (win). Notice the much smaller swings compared to Martingale or Fibonacci - this is D'Alembert's core appeal, and it's genuinely accurate that it's the least volatile of the three major systems.

Why "balancing wins and losses" is a logical trap. D'Alembert's underlying premise - that wins and losses will eventually balance out, making the gradually-increasing-then-decreasing stakes profitable - rests on an assumption called the gambler's fallacy: the belief that a loss makes a subsequent win more likely, or that outcomes need to "balance." Roulette spins are statistically independent events. A loss does not make a win more likely on the next spin, no matter how many consecutive losses have occurred, and the 48.65%/51.35% true win/loss ratio on even-money bets doesn't shift based on recent history.

The honest verdict on D'Alembert. It's genuinely the most conservative and lowest-variance of the three major systems - smaller bankroll requirements, lower risk of hitting table limits, more comfortable session-to-session experience. But over a long enough run, D'Alembert converges to the same house-edge-determined loss rate as flat, unchanging bet sizing - it's a slower, more comfortable way to lose at the same underlying rate, not a genuinely different mathematical outcome.

The James Bond system: a flat bet that covers most of the wheel

Unlike every system covered so far, James Bond isn't a progression system at all - your stake never changes based on previous results. It's better described as a fixed hedge: a specific distribution of chips across the table, placed identically every single spin, designed to cover a large majority of the wheel's numbers at once.

The classic distribution, using a $200 total stake as the traditional reference point: $140 on the high numbers (19-36, an even-money bet paying 1:1), $50 on the six-line covering 13-18 (a six-number bet paying 5:1), and $10 placed straight-up on zero (paying 35:1 on European roulette specifically). Together, this covers 25 of the wheel's 37 numbers (on a European single-zero wheel) - a genuinely large 67.5% of all possible outcomes on any given spin.

Why it's designed for European roulette specifically. The James Bond system's straight-up zero bet only makes mathematical sense on a single-zero European wheel - attempting the same distribution on an American double-zero wheel would leave one of the two zero pockets completely uncovered, undermining the system's core "cover most of the wheel" logic.

The profit tiers, and the genuine flaw. A winning spin on 19-36 nets a modest profit; hitting the 13-18 six-line pays considerably more; hitting zero itself delivers the largest single-spin return of all, at 35:1 on just a $10 stake. But the numbers 1 through 12 aren't covered by this distribution at all - a spin landing anywhere in that range (a 32.5% chance on European roulette) loses the entire $200 stake outright, with nothing recovered.

Why James Bond doesn't escape the same mathematical limitation as every other system covered in this guide. Despite covering a genuinely large share of the wheel, the underlying house edge on every individual bet within the distribution remains exactly what it always is on European roulette - 2.70% on the even-money portion, and the same underlying edge on the six-line and straight-up bets. Spreading your stake across more of the wheel changes how often you win (more frequently, in smaller amounts, with the rare large zero hit) without changing your long-run expected loss rate in any way. It's a genuinely different playing experience from a progression system - more frequent small wins, an occasional large zero-hit, and complete stake loss roughly a third of the time - but not a different underlying mathematical outcome.

Oscar's grind: the slowest, most conservative system of all

Introduced by Allan N. Wilson in his 1965 book The Casino Gambler's Guide, where he credited the method to an unknown dice player named Oscar, Oscar's Grind is a positive progression system - the increase happens after a win, not a loss, distinguishing it structurally from Martingale, Fibonacci, and D'Alembert, all of which escalate stakes in response to losses.

The mechanic: Start with a one-unit bet. After a loss, repeat the exact same stake on the next spin - no escalation at all. After a win, increase your next bet by exactly one unit. The specific goal of each cycle is to walk away with a net profit of precisely one unit, at which point the cycle resets and begins again from a one-unit base bet.

Why this is considered the most conservative system covered in this guide. Because Oscar's Grind never increases your stake following a loss - the single behaviour responsible for Martingale's catastrophic bankroll risk - a losing streak under this system simply means repeating the same modest bet indefinitely, rather than escalating exposure. This makes Oscar's Grind dramatically gentler on a bankroll during a losing run than any negative-progression system covered above.

The honest limitation. Because increases only happen after wins, and the target is a modest one-unit profit per completed cycle, a genuinely extended losing streak can drag a single Oscar's Grind cycle out over many, many spins before it finally completes - tying up your attention and bankroll in a single slow-moving cycle for a comparatively small eventual profit target. And, exactly like every other system covered in this guide, it doesn't change the underlying house edge - it simply reshapes a session into a long series of small, cautious bets working toward a modest, defined goal.

Other systems worth knowing

The Labouchère (cancellation) system: Write out a sequence of numbers (for example, 1-2-3-4). Each bet equals the sum of the first and last numbers in your current sequence (1+4=5 for the example above). After a win, cross out both numbers used. After a loss, add the lost amount as a new number at the end of the sequence. The sequence is considered "complete" (profitable) once every number has been crossed out. Like every other progression system, it changes the shape and pacing of a session without altering the underlying negative expected value.

Paroli (positive progression): The opposite structural approach to Martingale - rather than increasing stakes after a loss, Paroli increases stakes after a win, resetting to base stake after any loss. The theory is to press your advantage during a winning streak rather than chase losses during a losing one. It reduces the risk of catastrophic single-session losses relative to negative-progression systems, but carries exactly the same underlying mathematical limitation - it cannot change the game's house edge, only reshape when and how quickly you're likely to experience it.

What betting systems actually do achieve

This is worth stating honestly rather than dismissing systems entirely: they don't overcome the house edge, but they do genuinely reshape the variance and psychological experience of a session, which has real value for some players even without changing the underlying mathematics.

They add structure and a sense of control to a fundamentally random activity. Part of the appeal of any system is psychological - playing according to a defined method feels more purposeful than betting haphazardly, even when the underlying expected outcome is identical.

They genuinely affect the distribution of outcomes across a session. A conservative system like D'Alembert produces more frequent, smaller wins and losses, with a lower chance of a single catastrophic loss in any given session - at the cost of the same long-run house-edge-determined result. An aggressive system like Martingale produces the opposite profile: frequent small wins punctuated by the rare, occasionally severe loss when a losing streak outruns your bankroll or the table limit.

Choosing a system, if you want one, is really a choice about session shape, not about beating the house. If you want a calmer, lower-variance session with smaller swings, D'Alembert delivers that. If you're comfortable with rarer but larger losses in exchange for more frequent, satisfying small wins, Martingale delivers a different but equally mathematically fair (in the sense of "equally losing over time") experience.

The one genuine lever you have: which roulette variant you play

While no staking system changes the house edge, your choice of roulette variant absolutely does - and this is the single most impactful, genuinely effective decision available to any roulette player, system or no system.

Why this matters more than any staking system. Playing any system - Martingale, Fibonacci, D'Alembert, or none at all - on American roulette starts you at nearly double the house edge of the same system played on European roulette. Choosing the lowest-edge variant available is not a "strategy" in the system sense - it's a prerequisite that matters more than any staking pattern you layer on top of it. Where French Roulette with La Partage is available (returning half your stake on a losing even-money bet if the ball lands on zero), it offers the best mathematical odds of any roulette variant covered here.

Frequently asked questions

Does the Martingale system work at roulette?

No - while the logic (double after a loss to recover previous losses plus profit) is mathematically clean in isolated examples, it requires an unlimited bankroll and no table maximum to guarantee eventual recovery. In practice, a sufficiently long losing streak can force bet sizes beyond your bankroll or the table limit before the recovering win arrives, at which point the system fails.

Is d'alembert a safer roulette system?

It's genuinely lower-variance than Martingale or Fibonacci, with smaller bet-size swings and lower risk of catastrophic single-session losses. However, it converges to the same house-edge-determined long-run loss rate as any other staking pattern - it's a slower, calmer way to lose at the same underlying mathematical rate, not a way to overcome the house edge.

Can any betting system beat the house edge?

No. This is a mathematical certainty, not a matter of opinion: if every individual bet has negative expected value, no sequence or combination of those bets can produce positive expected value. Every system reshapes variance and session experience, but none change the underlying house edge.

What's the real win rate on red/black bets in European roulette?

48.65%, not the commonly assumed 50% - the house's edge comes from the single zero pocket, which belongs to neither red nor black, accounting for the gap between the "50/50" perception and the actual 48.65%/51.35% split.

What actually reduces the house edge in roulette?

Your choice of variant, not your staking system. European roulette (2.70% house edge on even-money bets) is meaningfully better than American roulette (5.26%), and French Roulette with La Partage (1.35% on even-money bets specifically) is better still, where available.

This guide is for educational purposes. No betting system changes the mathematical house edge of roulette, and all roulette play involves risk of loss. Gambling problem? In the UK: BeGambleAware.org | 0808 8020 133. In the US: 1-800-GAMBLER. Information correct as of September 2026.