Crash Gambling

Crash Gambling Simulator: Variance, Not a Winning Strategy

Simulated crash rounds show every cash-out target loses 3% at 97% RTP. What changes is the swing, from 48.5% wins at 2x to 9.7% at 10x.

By StakeCasino24 MonitorOct 26, 20266 min read

A long exposure light trail of a rocket launch arcs high across a starry night sky.

Photo: SpaceX on Unsplash

A crash gambling simulator replays thousands of rounds drawn from the game's published bust-point distribution so you can watch what a given cash-out target does over time. The result is consistent: every target loses the same share of each stake in the long run, a share fixed by the game's RTP, and the target only changes how violently your balance swings on the way there. A simulator measures variance. It does not find a winning strategy.

How a crash gambling simulator generates bust points

A simulator draws each round's crash multiplier from one rule: for a game with RTP r, the chance that a round reaches at least multiplier m is r divided by m. Everything else in the output follows from that single relation.

Philipp Dubach's analysis of provably fair games (January 2026) lays this out and then checks it against real data. On a 97% RTP game, reaching 2x should happen 48.5% of the time and reaching 10x 9.7% of the time, and those are close to the rates he observed. A live game produces its bust point from seeded hashing rather than a random draw inside a spreadsheet; the mechanics are covered in our explainer on how crash gambling games generate multipliers. A simulator skips the hashing and samples the resulting distribution directly, which is legitimate as long as the distribution matches.

The shape of that distribution matters more than most players expect. It is heavy-tailed: big multipliers are rare but not vanishingly so. Dubach fitted the tail exponent of the crash density at about 1.98 against a theoretical 2.0, and the largest multiplier in his sample was 10,000x. A simulator that caps multipliers low, or uses a thinner tail, will understate both the rare big wins and the long droughts between them.

Why every cash-out target has the same expected loss

The expected return per unit staked is r at every target, so the expected profit is r minus one stake, a 3% loss at 97% RTP whether you cash out early or late. The win probability falls exactly as fast as the payout rises.

At target m you win m times your stake with probability r/m. Multiply the two and the m cancels, leaving r. No target escapes that arithmetic, which is why no amount of simulated testing turns up a target with a positive edge.

Cash-out target	Share of rounds reaching it (97% RTP sample)	Expected return per unit staked
2x	about 48.5%	97% (3% loss)
10x	about 9.7%	97% (3% loss)

The two rows look like very different bets. One wins roughly every other round, the other fails more than nine times in ten. Their long-run cost is identical.

What a crash simulator actually tests: variance

Running many sessions at different targets shows the same average outcome with very different spreads. Low targets cluster results tightly; high targets scatter them widely in both directions.

Dubach ran a Monte Carlo of 10,000 sessions at 1.5x, 2x, 3x and 5x targets. All four had the same negative expected return per unit wagered. What changed was the range: lower cash-out targets reduced variance, higher targets widened the range of possible results.

That is the useful output of a simulator. It answers questions such as:

How deep a drawdown should a bankroll expect at a given target before the average reasserts itself?
How often does a session at a high target end ahead purely by luck?
How long can a losing run last at a target that succeeds less than one round in ten?

A player who sees a high-target session finish well ahead in a simulator is looking at the wide end of that spread, not at evidence the target works. Rerun the same settings and the winner becomes a loser; average enough sessions and every target converges on the same loss rate.

How fast the house edge adds up per hour

Pace converts a percentage edge into a concrete expected loss. At a fixed stake, the hourly cost is the number of rounds played times the stake times the edge.

Dubach measured the observed pace of rounds at 179 rounds per hour, with a 16-second median interval between rounds. At a 3% edge, 179 fixed-stake bets imply an expected loss of 5.37 stake units. Nothing about the cash-out target changes that figure. Only stake size, the RTP and how many rounds you play move it, which makes round count the most honest input to put into any simulator.

What a simulator cannot tell you

A simulator cannot predict the next round, cannot find a target that beats the edge and cannot confirm that a real operator's game follows the distribution it claims. Those limits are built into what it is.

Each simulated round is independent, as each real round is meant to be, so past results in the simulation carry no information about the next one. The simulator also inherits whatever RTP you type in. If an operator's stated RTP is wrong, the simulation is wrong in the same way.

Fairness is a separate check. Provably fair crash games typically derive outcomes from a server seed with HMAC-SHA-256, and RFC 4231 publishes test vectors for that construction (December 2005), so a verification tool's hashing can be checked against fixed known answers before you trust it on real rounds. Verifiers often read only the leading bits of the digest; RFC 2104's guidance on truncated output recommends keeping at least half the hash output length and at least 80 bits, which is why reading a leading slice of an HMAC-SHA-256 digest is sound practice rather than a shortcut.

Questions on this

Can a crash gambling simulator find a winning cash-out target?

No. Because the chance of reaching a multiplier m is the RTP divided by m, every target returns the same share of each stake over the long run. On a 97% RTP game that is a 3% expected loss per bet whether you cash out at 2x or 10x.

Is cashing out at 2x safer than cashing out at 10x?

It is less volatile, not cheaper. In a sample of about 20,000 rounds at 97% RTP, a 2x target succeeded about 48.5% of the time and a 10x target about 9.7% of the time. Both carry the same expected loss; the 10x player simply sees longer losing runs and rarer, larger wins.

How much does a crash player lose per hour on average?

It depends on pace and stake. One observed game ran at 179 rounds per hour, and 179 fixed-stake bets at a 3% edge imply an expected loss of 5.37 stake units. Faster play or a bigger stake scales that figure directly.

Does a simulator show whether a crash game is fair?

No. A simulator assumes the published distribution is honest. Checking a real game means recomputing its outcomes from the revealed server seed with HMAC-SHA-256, a construction you can test against the fixed answers in RFC 4231.