Fibonacci betting system: the sequence, the streak and the edge

Leonardo of Pisa’s sequence is a way to grow integers, not a way to grow an edge. The gambling version maps those integers onto even-money units.

Standard rule set:

1. Write the list 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, …
2. Start at the first 1 (some players start at the second 1; it barely matters).
3. After a loss, advance one term. After a win, move two terms back, or to the start if you run out of room.
4. Some rooms say “reset to 1 after a win”. That is a harsher, simpler variant. Know which rule you are actually using.

The Fibonacci betting system is sold as the civilised cousin of doubling. 1 → 1 → 2 → 3 → 5 is less shocking than 1 → 2 → 4 → 8 → 16. After ten losses you are at 89 units, not 1,024. That is still 89 units, and the unpaid losses behind you are the sum of the path, not the last term alone.

This page sits with the other progressions in the Games and odds topic. Read Martingale for the exponential extreme and Paroli for the raise-on-win mirror. PVPspinArena is 18+.

Players reach for Fibonacci because they remember a school-poster spiral and want a climb that looks reasoned. A reasoned climb that still faces 2x on 7/15 is just a prettier Martingale. Write the next three terms before you sit down. If 8, 13 and 21 already look larger than the session cap, the system does not fit the budget, and no later win will make the fit better.

The first twelve terms sum to 376. That sum is the money you have already put on the table if you lose twelve times from a $1 first term before the thirteenth stake (the thirteenth term is 233). People remember “Fibonacci grows slowly” and forget to add the column. Slow relative to 2^n is not slow relative to a $40 session.

Martingale’s win wipes the cycle: one 2x on the last double recovers every chip in the hole plus one unit. Fibonacci’s win does not.

Suppose you are at 8 units after the path 1, 1, 2, 3, 5, 8. Losses so far = 1+1+2+3+5 = 12, and you have just lost the 8 as well if we are about to bet 13 — be careful with off-by-one when you recount. The clean way is to keep a running P/L.

A win at 8 returns 16 (2x), for +8 on that bet. If losses already booked were 12, you are still −4, and the rule then steps you back to 2 (two steps left from 8: 8 → 5 → 3, or 8 → 3 depending on whether 5 is skipped as “two steps”). House write-ups disagree on the exact cursor. The common textbook is: pointer moves two places toward the start. You are not even at base, and you are not automatically in profit.

Why people like the two-step back

It feels like you “keep some pressure on” after a win so the next hits can finish the repair. That is a story about cursor position. It is not a repair guarantee. A win–loss–loss after you thought you were coming home sends the pointer right again.

Unit = $1. Even-money bet. Start at the first 1.

Eight losses and one win, still −$20, sitting on a $13 request. This is a normal ugly path, not a freak 20-loss myth. On Purple, eight losses in a row have probability (8/15)^8 ≈ 0.6%, about 1 in 160 isolated attempts — common across a long night of cycles.

A $1 Martingale would have been trying to bet $256 after eight losses, with a $255 hole. Fibonacci is “kinder” and still leaves you in a hole a single win did not fill. Kinder is not plus-EV.

Run the same W/L tape as flat $1: eight losses and one win is −$7. Fibonacci turned it into −$20 because more dollars sat on the table during the cold stretch. That is the system working as designed.

A second tape: early win, then cold

Start at 1, win (P/L +$1, pointer two back = start), then lose five times: stakes 1, 1, 2, 3, 5. Extra losses $12, net −$11, next term 8. You “reset” after the first win and still walk into a double-digit hole. The two-step-back rule only helps when wins arrive while the pointer is already high. Hoping for that arrival is not a method. It is a dependence on a streak the sequence does not cause.

Let e be the house edge per dollar (0.0667 on PVPspinArena Purple, 0.027 on European red). Expected loss equals e times total wagered. Fibonacci increases total wagered after losses. Expected loss therefore rises relative to flat betting the base unit for the same number of rounds.

There is no hidden identity in 1+1+2+3+5 that equals a fair 50% coin. The sequence is a convenience for choosing integers. The wheel pays 2x on 7 of 15 slots. Those facts do not multiply to a surplus.

A compact EV reminder: whatever term F you bet, EV = −0.0667 F on Purple. The pointer is a function of the past. The past does not appear in −0.0667 F. Sum those terms along any path and you have the expected leak on that path's turnover. The Fibonacci betting system is a path generator. It is not a discount.

House edge is the one-page version. D’Alembert is the linear cousin: same minus sign, different schedule. If a video says Fibonacci “hits a balance point”, ask for the EV on the next dollar. It is still −e.

Independence also kills the folk theory that a miss makes a hit more likely, which is the ghost inside every negative progression. See the gambler’s fallacy if you want that error named cleanly.

Paroli does not belong in the “after 7 losses” row: it would still be betting $1. That is its virtue and its limit. Fibonacci sits in the middle of the negative-progression pack: enough growth to hurt, not enough recovery to mimic Martingale’s all-or-nothing reset.

Choose among them only as entertainment aesthetics. The grown-up stake is a flat unit sized from a budget, not from a medieval sequence.

Roulette

Purple or Silver is the intended canvas. Green is not: you will walk far to the right of 89 before a 1-in-15 hit, and the 14x does not line up with a two-step-back rule designed for 2x. European red is the same system with a smaller e. American red is a larger e. Open Roulette and count 15 slots if you want the p that the sequence is ignoring.

Coinflip

A 0% fee Coinflip is ~0-EV. Fibonacci then manufactures a random walk with an expanding step after losses. Mean near zero, variance higher than flat. Someone must match 13, 21, 34. Many will not.

Jackpot

A Jackpot ticket’s p is stake/pot, not 1/2. Mapping Fibonacci onto pot entries is a way to buy a larger share after you already lost, which is a mood decision, not a sequence identity. Skip it.

Honest results are still checkable on fairness. A verified L is still an L. The next term is still minus-EV on a house colour.

If a written rule helps you stop inventing stakes, write a better one:

• one unit, always;
• stop at a dollar loss cap;
• stop at a clock;
• never advance a sequence after a loss.

The Fibonacci betting system cannot beat a house edge. A slower progression is a slower way to put extra dollars into a leak. If the pointer is already walking toward 34 because you are angry, that is chase. Stop. Use the responsible gambling page. Gambling is optional and 18+ only.

Treat any completed “come back to 1” as variance. Treat the sequence as a museum piece. Price the colour with p × 2, then decide whether the entertainment is worth −6.67% of whatever you are about to shove along the list. If the next term is already 13 and you are explaining the sequence to yourself, the list has started making the decisions. Close it. A named integer series is not a reason to stay. The sequence was built to count rabbits, not to cancel a 6.67% leak. Leave it in the maths book. If you need a sequence at all, use 1, 1, 1, 1 and a clock.