Betting Systems Explained: Fibonacci, Paroli, D'Alembert
A betting system is a rule for changing your stake after wins and losses. Negative progressions like the Fibonacci, D'Alembert and Labouchere raise the stake after a loss; positive ones like the Paroli raise it after a win. None of them changes the odds of the game, so none beats the house edge. We ran five systems and plain flat betting on 2,605 real sessions of 200 live roulette spins each, and every one of them returned between 97.0% and 97.7% of what it staked. What the systems changed was how often a session ended ahead, and how often it ended with the bankroll gone.
Key takeaways
- Two families: negative progressions (Martingale, Fibonacci, D'Alembert, Labouchere) raise the stake after losses; positive ones (Paroli) raise it after wins.
- Same cost: on 2,605 real 200-spin roulette sessions, every system returned 97.0% to 97.7% of its stakes, the 2.7% edge of a single-zero wheel.
- Fibonacci: finished ahead in 57.1% of sessions, but hit a stake it couldn't cover in 42.2% of them, losing 72 units on average when it did.
- Labouchere: the most volatile: 82.9% of sessions ended behind with a stake larger than the bankroll left, yet it produced the single biggest win, +373 units.
- Paroli and flat betting: never ran dry on a 100-unit bankroll, and finished ahead in about a third of sessions.
What is a betting system?
A betting system is a rule that decides your next stake from your last results. It doesn’t choose what to bet on, only how much. The appeal is control: a system gives a session structure, and most promise that a string of losses can be recovered or a winning run exploited.
The mathematical catch is that a stake size can’t change the odds of the bet it’s placed on. Every bet on a single-zero roulette wheel keeps 2.7% on average, whatever size it is, so a sequence of those bets keeps 2.7% of the total too.
Negative progressions: Fibonacci, D’Alembert, Labouchere
These raise the stake after a loss, so a win later in the sequence recovers what went before.
- Fibonacci: stake along 1, 1, 2, 3, 5, 8, 13, 21… After a loss, move one step up; after a win, move two steps back. Gentler than doubling, but the stake still climbs quickly in a long losing run.
- D’Alembert: add one unit after a loss, remove one after a win. The slowest progression, sold as the safe one, because the stake grows in a straight line rather than exponentially.
- Labouchere: write a line of numbers, for example 1-2-3-4. Each bet is the first number plus the last. A win crosses both out; a loss adds the lost stake to the end. When the line is empty, you’ve won its total, 10 units here.
The Martingale, which doubles after every loss, is the most aggressive member of this family, and we tested it separately on 4.5 million real rounds.
Positive progressions: the Paroli
The Paroli turns the idea around: double the stake after a win, go back to one unit after a loss, and reset after three wins in a row. Losses stay small because you’re mostly betting one unit; the hope is to catch a streak. It’s sometimes called the reverse Martingale, which usually means doubling after wins with no cap on the streak.
We tested five systems on 2,605 real roulette sessions
We took 521,000 live spins from three single-zero tables in our records and cut them into 2,605 sessions of 200 spins. Each system bet on red, with a 1-unit base stake and a 100-unit bankroll; a session stopped early if the next stake was bigger than the bankroll left.
| System | Sessions finished ahead | Stopped behind: next stake over bankroll | Average result | Return on stakes |
|---|---|---|---|---|
| Flat betting | 31.7% | 0.0% | -5.6 | 97.21% |
| Paroli | 33.6% | 0.0% | -10.1 | 97.02% |
| Fibonacci | 57.1% | 42.2% | -10.3 | 97.31% |
| D’Alembert | 38.0% | 57.0% | -25.3 | 97.12% |
| Labouchere | 17.1% | 82.9% | -23.5 | 97.57% |
| Martingale | 32.0% | 67.9% | -8.8 | 97.74% |
Sessions that stopped early lost 59 to 92 units on average, depending on the system; only the Labouchere ever stopped while ahead, in 0.3% of sessions, when its stake outgrew a bankroll that was still up; those count among its 17.1% ahead. Every other Labouchere session stopped short behind, which is why its two columns add up to 100%. The last column is the one that matters: every system returned about 97% of what it put on the table, which is the wheel’s 2.7% edge doing its job. The systems only moved the other columns around.
What each system actually changed
The Fibonacci looked like a winner. It finished ahead in 57.1% of sessions, more than any other system, because its gentle progression recovers most short losing runs. But in 42.2% of sessions a longer run pushed the stake past what was left of the bankroll, those sessions lost 72 units on average, and that wiped out the gains of the rest.
The D’Alembert wasn’t the safe option it’s sold as. Adding one unit after every loss sounds tame, but red loses more often than it wins, so the stake drifts upward over a session. It stopped short in 57.0% of sessions, 92 units down on average, and had the worst average result.
Labouchere players get the widest range of all. It stopped short in 82.9% of sessions, the most of any system, and also produced the biggest single session, +373 units. That’s the profile of a lottery ticket, not a strategy.
With the Paroli, a session looks like flat betting with more swing. It never ran dry on 100 units, because it only raises stakes with money it has just won, and its best session reached +103. Its average loss was larger than flat betting’s, simply because it put more money through the wheel.
Is there a better strategy than the Martingale?
If better means less likely to empty your bankroll, yes: flat betting and the Paroli never stopped short in our test, against 67.9% of sessions for the Martingale. If better means more profitable, no: all of them returned about 97% of stakes, and our full test of the Martingale on 4.5 million real rounds shows the same thing at every bankroll size.
Is roulette 100% luck?
For the player, yes. Every spin is chance, and every bet on a single-zero wheel carries the same 2.7% edge. The only real choices are the table and the bet type: a double-zero wheel keeps 5.26% instead, and a single number is the most volatile bet on the table while red or black is the least, at the same price.
Why no system can beat roulette
Each spin is independent, as our test of the gambler’s fallacy shows on the same spins: a run of reds doesn’t make black more likely. A betting system is a bet that the recent past says something about the next spin, and it doesn’t. With no pattern to exploit and the same edge on every bet, the only variable a system controls is how much you stake in total, and a bigger total means a bigger expected loss.
If you like a structure to your play, the one that works is a budget: a fixed stake, a session length and a loss limit set before you sit down. Our house edge guide shows which tables cost the least to play.
If gambling stops being fun, GamCare and Gambling Therapy offer free, confidential help.
Frequently asked questions
What are the different types of betting systems?
There are two main families. Negative progressions raise your stake after a loss, hoping a win recovers it: the Martingale doubles, the Fibonacci follows the Fibonacci sequence, the D'Alembert adds one unit, and the Labouchere uses a written line of numbers. Positive progressions raise the stake after a win to ride a streak, like the Paroli and the reverse Martingale. Flat betting, the same stake every time, is the baseline they're all compared to.
What is the most successful betting strategy?
No staking strategy beats the house edge, so success depends on what you measure. In our test on 2,605 real roulette sessions, the Fibonacci finished ahead most often, 57.1% of sessions, but in 42.2% of them it reached a stake the bankroll couldn't cover and stopped, 72 units down on average. Flat betting never ran dry and lost the least on average per session. Every system returned about 97% of what it staked, so none of them is a winning strategy.
How does the Fibonacci strategy work in betting?
You stake along the Fibonacci sequence, 1, 1, 2, 3, 5, 8, 13 and so on, where each number is the sum of the two before it. After a loss you move one step up the sequence; after a win you move two steps back. Because the steps grow more slowly than the Martingale's doubling, busts come later, but they still come: in our test, 42.2% of 200-spin sessions hit a stake a 100-unit bankroll couldn't cover.
Is there a mathematical way to win at roulette?
No. Every standard bet on a single-zero wheel returns 97.3% of the stake on average, and a staking pattern only rearranges when you win and lose. Changing bet sizes can't change the edge of any single bet, so it can't change the edge of the whole session. Our test of five betting systems on real spins confirmed it: all of them returned between 97.0% and 97.7%.
Is roulette beatable?
Not with betting systems. A roulette result is independent of every spin before it, so there's no pattern for a system to exploit. The historical exceptions involved faulty or biased wheels, or tracking the ball physically, not staking methods, and a staking method exploits neither. For a player, the realistic goal is to lose slowly: bet on a single-zero wheel and keep stakes small.
Set a budget before you play. Free, confidential help is available from GamCare and Gambling Therapy.