The Fibonacci system is the gentle chaser of crash betting: you step up the sequence 1, 1, 2, 3, 5, 8, 13 after a loss and step back two places after a win, so your stakes climb at roughly the golden ratio (about 1.618 per step) instead of Martingale’s brutal doubling. It feels controlled, and it is slower to ruin. It is also, in expectation, a guaranteed loss.
Of the three staking systems people bring to crash games, Fibonacci is the one that feels most reasonable. It does not double your bet after every loss the way Martingale does, and it does not hand your winnings back in one shot the way Anti-Martingale can. It promises a measured recovery, clawing losses back gradually across a couple of wins.
That measured feeling is exactly what sells it, and it is also a distraction. The growth rate is gentler, but the destination is identical: at a 2.00x cash-out target on a 97% RTP game, every round you stake loses 3% in expectation, and no order or size of bets changes that. This article shows the maths, walks a full 20-round example, and explains why even the cleverest progression cannot dent the house edge.
The 30-second version
Fibonacci is a moderate negative progression, bet more after a loss and less after a win, which puts far more turnover through the game than flat staking and so raises your absolute expected loss while the percentage edge stays welded at 3%. It grows roughly as 1.618 to the n versus Martingale’s 2 to the n, so it survives longer streaks before the stakes turn absurd, but the long-run loss is still mathematically certain. It is the gentle chaser, not a winning system.
🎮 How the Fibonacci staking system works
Fibonacci is a negative progression built on the sequence where each term is the sum of the two before it: 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144 and onward. As a staking plan it has two simple rules, and the natural crash-game target is a 2.00x auto-cashout, where a win doubles your stake and a loss costs it.
- After a loss, move one place forward. Advance one term up the sequence and bet that many units. A run of losses walks you up 1, 2, 3, 5, 8 and so on.
- After a win, move two places back. Retreat two terms and bet that smaller amount, floored at the base unit. The idea is that two wins roughly unwind the losses banked on the way up.
The rationale players give is recovery without recklessness. Because each term equals the sum of the previous two, a couple of wins steps you back down the ladder while, in theory, clearing the losses you accumulated climbing it. That is the appeal, and it is genuinely less violent than doubling. What it is not is a way to change the odds.
📝 For the record: Implementations differ at the edges. We floor the progression at the base unit and keep playing. Some variants instead “walk away” the moment a win would carry you back past the start of the sequence, and some begin at the second 1 so an early loss-then-win nets a push. These tweak the session bookkeeping, never the expected value.
🏆 The gentle chaser: Fibonacci against Martingale and Anti-Martingale
The single thing that sets Fibonacci apart from Martingale is how fast the stakes grow, and that gap is enormous. Imagine a 10-loss streak from a £1 base. Under Martingale, your 11th bet is 2 to the 10th, or £1,024, and you have already staked £1,023 across the ten losses. Under Fibonacci, your 11th bet is the 11th term, £89, and you have staked just £143 in total.
That is the headline contrast: a £1,024 next bet versus an £89 next bet, and £1,023 burned versus £143 burned. Fibonacci numbers grow sub-geometrically, multiplying by roughly 1.618 each step rather than doubling, so they buy far more losing rounds before the stakes become unplayable. That is why the system feels controlled.
“Fibonacci is the gentle chaser: it slows the climb but never moves the cliff.”
A gentler climb is still a climb to the same edge. The table below sets the five common staking patterns side by side. The escalation speed differs wildly. The proportional house edge does not move a fraction.
The ratio between consecutive Fibonacci terms converges on the golden ratio quickly: 55 divided by 34 is 1.6176, and 6,765 divided by 4,181 is 1.6180339, accurate to seven decimals. So Fibonacci climbs at a fixed multiple of about 1.618 while Martingale climbs at 2.0. The full Martingale ruin case is covered in our Martingale debunk, and the mirror-image positive progression in the Anti-Martingale debunk; both reach the same edge by a different road.
🔢 Why Fibonacci cannot change the house edge
Fibonacci cannot change your expected value because expected value is fixed per round and simply adds up across rounds. Start with one round at a 2.00x target on a 97% RTP game. The win probability is the RTP divided by the target, 0.97 divided by 2.00, which is 0.485.
The expected value of staking S on that round is therefore:
EV = 0.485 x (+S) + 0.515 x (-S) = -0.03S
Every round, whatever its stake, loses 3% of that stake in expectation. That 3% is the house edge, and it is welded to the game, not to your staking pattern. Total expected loss over a session is just the sum of those per-round losses:
E[total loss] = 0.03 x (sum of all stakes) = house edge x total turnover
💡 Key insight
The order and the size of bets drop out of that sum entirely. Reordering or resizing a series of fixed-percentage losses cannot change their total. That single fact is why Fibonacci, Martingale, Anti-Martingale, D’Alembert and flat betting all carry the identical proportional edge. They differ only in how much turnover they generate.
This is not a CrashEdge opinion, it is settled probability theory, and three results converge on it. Epstein’s first theorem states that over many plays at a constant single-trial probability, any and all betting systems lead to the same mathematical expectation of gain per unit wagered. Ethier’s conservation of fairness names Fibonacci explicitly among six systems and proves none of them, nor any system one might devise, can turn subfair wagers into superfair ones. Doob’s optional stopping theorem closes the door on clever quitting: a bounded strategy cannot beat a fair or sub-fair game, and on a negative-edge game your expectation can only stay the same or get worse. The full general proof lives in our crash gambling maths guide.
◆
📊 Worked example: 20 rounds at a £1 base unit
A concrete run makes the turnover problem visible. The table below plays 20 rounds at a 2.00x target with a £1 base, using a plausible mix of 8 wins and 12 losses, including an early six-loss streak to show the escalation. Position is the place in the sequence, the stake is the Fibonacci number at that position, and you move +1 position on a loss and -2 on a win, floored at position 1.
Under Fibonacci this run stakes £58 in total, for a theoretical expected loss of 3% of £58, or £1.74. The actual result on this particular ordering was +£2. Now run the identical 8-win, 12-loss pattern flat at £1 a round: total staked is £20, theoretical expected loss is £0.60, and the actual result is 8 minus 12, or -£4.
The bottom line: Fibonacci staked nearly triple the turnover of flat (£58 versus £20), so its absolute expected loss is nearly triple too (£1.74 versus £0.60), while the percentage edge is identical at 3%. The reason it finished ahead here is pure luck of ordering: the big recovery bets at rounds 7 and 8 (£13 and £5) happened to land on wins. Had that streak run a few rounds longer, or had those recovery bets lost, the session would be deeply negative. A positive session on a higher expected loss is exactly the illusion that sells negative progressions.
🔍 Where Fibonacci comes from, and why the mystique is irrelevant
The sequence has a genuine pedigree, none of which has anything to do with gambling. It entered Western mathematics through a medieval rabbit-breeding puzzle, was named centuries later, and was only bolted onto betting in modern times.
Liber Abaci, by Leonardo of Pisa
The sequence appears in a rabbit-population problem: one pair, each mature pair breeding monthly, giving totals 1, 1, 2, 3, 5, 8, 13. The pattern was already known to Indian mathematicians studying Sanskrit prosody, but this is where Western mathematics first recorded it.
Edouard Lucas names it
The French mathematician Edouard Lucas attaches the name “the Fibonacci sequence” to the rabbit problem. The pattern had been largely forgotten between 1202 and the 19th century.
Bolted onto betting
The staking system is a modern appropriation. Leonardo of Pisa never proposed the sequence as a way to bet, and applying it to roulette, baccarat or crash is a 20th-century invention dressed in 13th-century mathematics.
🔍 Worth noting
The golden ratio (phi, about 1.618) that the term ratios converge on is genuinely real and turns up in sunflower heads, shells and architecture. It has zero bearing on probability. A coin, a roulette wheel and a provably fair crash RNG have no memory of, and no sensitivity to, the ratio between your consecutive stakes. Even in finance, “Fibonacci retracements” are explicitly described as possible zones of interest, not guaranteed reversals. The crash version inherits none of the mathematics and all of the marketing.
◆
⚙️ Running Fibonacci on a real crash game
Here is the crash-specific catch that most guides skip: standard auto-bet panels cannot encode a Fibonacci ladder directly. They are built around percentage adjustments and resets, not an additive number sequence with a memory of the previous two stakes.
Aviator’s native autoplay only repeats a flat stake. It lets you set the bet amount, the number of rounds, an auto-cashout multiplier and stop conditions, but there is no progression field at all. Stake Crash offers “On Win” and “On Loss” triggers, but only as “Reset” or “Increase by X percent”, which makes Martingale trivial and Fibonacci impossible: a fixed percentage produces a geometric sequence, not an additive one. The only mainstream platform that can run true Fibonacci natively is BC.Game Crash, whose advanced tab includes a custom script editor where you can code the ladder, though that means writing or copying a script rather than ticking a preset.
The practical upshot is that most people running Fibonacci adjust the stake by hand every round. At crash speed, where rounds last seconds, that is slow and error-prone. And the deeper a losing streak runs, the more the stakes climb. The table shows how the numbers escalate from a £1 base.
To still be playing at step 15 you must place a single £610 bet from a £1 base, having already churned roughly £1,596. By step 20 the next bet is £6,765 and you have cycled £17,710. Long before that, you hit the same two walls that finish Martingale: your own bankroll, or the game’s maximum bet. The climb is gentler, the destination is the same.
⚠️ Important
Slower ruin is still ruin. Fibonacci buys you more losing rounds before the stakes turn unplayable, which makes it feel safer, but the long-run loss is mathematically certain in every case. The only thing it changes is the shape of the journey, never where the journey ends.
The escalation pull is also where staking systems shade into chasing, which is one of the clearest risk signals in fast-paced play. We cover the research, the warning signs and the support routes in a dedicated guide: crash gambling and player harm.
💡 What to do instead
If you take one thing from this, it is that no staking sequence touches the 3% edge, so the honest choices are about variance and cost, not about beating the game.
- If you want low variance, stake flat. It minimises turnover and therefore minimises absolute expected loss, £0.60 against Fibonacci’s £1.74 on our 20-round example. Our flat betting guide sets out the baseline.
- If you want the “controlled” feel, understand the trade. You are buying frequent small wins and a smoother session in exchange for a rare large loss and higher total expected loss. That is a psychology choice, not an edge.
- Cap it with hard limits. Set a maximum sequence position in advance, a stop-loss and a session budget. Any losing streak that pushes the next stake beyond 1 to 2% of your bankroll is a stop signal.
- Change the one variable that matters: RTP. Moving from a 97% game to a 99% game cuts the edge from 3% to 1%, a real, EV-level change that no staking system can match. Stake Crash and BC.Game Crash both run at 99%.
Fibonacci is the most elegant of the staking systems and one of the most persuasive, which is exactly why it is worth being clear-eyed about it. It is a way to manage how a session feels, nothing more. For the wider picture of why progressions and “systems” cannot work, and where to start instead, see our beginner’s guide to crash game strategy.