Last updated:
The Math Behind the Multiplier
Every crash round at a provably fair casino is a deterministic function of three inputs: a server seed, a client seed, and a round counter called the nonce. They are hashed with HMAC-SHA256, the first 52 bits are converted to a uniform float h between 0 and 1, and the game applies one line of arithmetic to produce the crash point. No animation, no live tick, no streaming decision. The multiplier is locked the moment the bet is placed. The rising curve is just a graph of a number that was already decided.
The canonical formula, descended from the 2014 Bustabit source, is:
crashPoint = floor(100 × (1 − houseEdge) / h) / 100
Take h = 0.5 with a 3% edge and the round crashes at 1.94x. Take h = 0.001 and the round survives to 970x. The distribution that falls out of this single equation is the only thing a player needs to understand the rest of the game.
What That Formula Means in Practice
For a house edge of 3%, the probability the multiplier exceeds m is exactly (1 − 0.03) / m, capped at m = 1. Roll the math forward:
- 48.5% of rounds survive past 2x
- 19.4% past 5x
- 9.7% past 10x
- 0.97% past 100x
That last number is the reason the 1000x screenshot exists at all. Out of 10,000 rounds, roughly 97 reach triple digits. The casino pays out 1,000x to one player; the casino keeps the long tail from the other 9,903.
Auto-Cashout Is the Only Strategy That Matters
Set auto-cashout at 2x with a 3% edge. Each round resolves to either 2.00 or 0. Expected value per unit staked is 0.485 × 2 + 0.515 × 0 − 1 = −0.03. The 3% is the house edge, recovered in pure arithmetic.
Push auto-cashout to 5x. Win probability drops to 0.194, payout rises to 5. EV = 0.194 × 5 − 1 = −0.03. Same number. Push to 10x: EV = 0.097 × 10 − 1 = −0.03. Push to 100x: EV = 0.0097 × 100 − 1 = −0.03. The expected return is identical at every cashout point, because the formula guarantees it. The only thing auto-cashout actually changes is variance.
This is the data that kills most “strategy” videos. There is no multiplier sweet spot, no optimal target, no margin of safety at 1.5x versus 1.6x. There is one negative-EV bet with a tunable variance knob.
Why Betting Systems Always Fail
Martingale at 2x looks like a 50/50 coin flip with deep pockets. It is not. Each round is independent: the SHA256 hash of round 482,917 has zero statistical relationship to round 482,918. A 10-red streak in crash means nothing. A 10-red streak is exactly as likely as red-after-red-after-red for the next ten flips of a fair coin. The streak a player remembers and the streak they do not are the same streak in expectation.
Then the variance. At 2x with a 3% edge, a single $1 bet has standard deviation around $1.39. A Martingale sequence that survives 8 consecutive losses requires a bankroll of 255 base units. After 8 rounds, expected loss is 8 × 0.03 = $0.24 per base unit, or $61.20 total across the 255-unit stake. The 9th loss is a $256 wipeout. Gambler’s Ruin probability for that sequence at 51.5/48.5 odds is 0.515^9 = roughly 1 in 2,000 per attempt, but the 1 in 2,000 erases an entire 8-round profit. The math does not negotiate.
Streak-betting on “10 reds then black” is the same fallacy wearing a different hat. The generator has no memory. Round 4,829 is independent of round 4,839, no matter what happened in between. Live-dealer crash variants that show recent history do not change the underlying RNG; they just show a recent history.
Variance, Standard Deviation, and Bankroll Sizing
For a target multiplier m at 3% edge, single-bet variance is m × 0.97 − 0.0009, since E[X²] = p × m² and p = 0.97/m. At m = 2, variance is 1.94 and standard deviation 1.39. At m = 10, variance is 9.70 and standard deviation 3.11. Higher target, higher variance, identical expected loss.
For 2x target, 1,000 rounds, and a bet size of 1% of bankroll, total staked is 10x bankroll, expected loss is 30% of bankroll, and the standard deviation of final PnL is roughly 44% of bankroll. The 95% confidence interval runs from −116% to +56% of bankroll. The lower bound exceeds total bankroll, which is the math behind every recorded crash ruin.
Cut the bet to 0.25% of bankroll and the same 1,000 rounds produces a 7.5% expected loss with 11% SD. Survivable, and the 95% drawdown band lands at roughly −29% of bankroll. A quarter-Kelly approach (full Kelly = edge / odds = 0.03, then take 1/4 of that) lands on 0.75% of bankroll per bet, which is the empirically sane middle ground for a 1% to 4% edge game.
There is no number of rounds at which this turns positive. Expected value per round is negative at every target, every edge, every bet size. The only question is whether variance finishes the bankroll before the session cap does.
Crash Games and Bonus Wagering
For players running crash through a deposit bonus, the same logic runs in reverse. Of the 1,094 casinos reviewed at Betting-Center, 861 carry a wagering requirement with a median of 35x. Crash contributes 100% at most of them. That sounds generous. In practice, a 35x WR played on a 3% edge game requires a multiplier of roughly 0.97^35 = 0.34 to clear in expected value, meaning about 34 cents of every bonus dollar survives the playthrough, before the edge grinds the rest. The bonus value calculator shows the breakdown for any specific offer; most crash-friendly promos look generous in headline form and bad in the bottom line.
Some operators exclude crash from bonus play entirely. Of the 1,094 reviewed sites, 873 are crypto-friendly and 415 are instant-withdrawal, but bonus eligibility on crash is a property that must be checked per casino. Reading the terms matters more than picking the casino.
What to Actually Do
Pick a multiplier. Pick a bet size under 1% of bankroll, ideally around 0.25% to 0.75%. Set auto-cashout. Cap the session at 50 to 200 rounds depending on variance tolerance. Do not raise the bet after a loss. Do not chase a 50x. The expected value of every round is between 0.96 and 0.99 of the stake no matter what target is selected; the only variable in player control is whether variance wipes the bankroll before the session cap does.
For anyone comparing bonus offers before playing, the wagering database with 1,094 entries sorts by contribution, edge, and effective bonus value. Most crash-friendly promotions are not as friendly as they look once the 1% to 4% house edge compounds across a 35x playthrough.