How crash games work
Every round, a crash game picks a hidden "bust point" — the multiplier at which the round ends. A rising curve or number climbs from 1.00×, and you either cash out before it stops or lose the round when it does. In the reference model used on this page — the same model behind this simulator and the other calculators on this site — the chance that the bust point reaches or exceeds a given multiplier x follows a single formula:
P(reach x) = (1 − e) / x where e is the house edge as a fraction (3% = 0.03 in the examples below). Many real crash games are built on something in this family, but providers rarely publish their exact formula or parameters — treat the numbers below as a description of this site's reference model, not a verified claim about any specific branded game.
| Target | Chance of reaching it | 1 in | Expected return |
|---|---|---|---|
| 1.5× | 64.67% | 1.55 | 97.00% |
| 2× | 48.50% | 2.06 | 97.00% |
| 3× | 32.33% | 3.09 | 97.00% |
| 5× | 19.40% | 5.15 | 97.00% |
| 10× | 9.70% | 10 | 97.00% |
| 100× | 0.97% | 103 | 97.00% |
Figures computed by the engine for a 3% house edge — the same default used by the simulator above.
Why every target has the same expected return
Look at the last column: it does not change. In this reference model, cashing out at a target x pays x times the stake if (and only if) the round reaches x — that is the model's actual payout, not a fair one. A genuinely fair, zero-edge game would instead pay 1 / P(reach x), the reciprocal of the win chance, which exactly cancels the odds and returns 100% on average. This model pays less than that: multiply its real payout x by the win chance P(reach x) = (1 − e) / x and the x cancels out, leaving a constant expected gross return of (1 − e) — 97.00% of the stake — no matter which cash-out target you pick. A 1.5× target wins often and pays little; a 100× target wins rarely and pays a lot. Over many rounds both drain the average balance at the same rate — there is no target that beats this model's house edge, and no "safe" low multiplier that avoids it.
One more gap to be aware of: the formula above is the engine's exact theoretical model — it does not round. The live and bulk simulator on this page pays out in whole cents, flooring every win down to the nearest cent. At ordinary stakes that rounding is negligible, but at a very small stake (for example $0.01) the floor can swallow the entire edge-adjusted win, pulling the simulator's realized return below the 97.00% theoretical figure above. sessionStats and crashOdds describe the un-rounded model; the simulator and the risk of ruin calculator apply that per-round cent floor.
What you’ll see in the simulator
In live rounds, each bust point comes from the engine’s provably fair function, animated the same way a real crash game would show it. Short sequences are noisy: you can watch five or ten wins in a row, or a long losing streak, purely from variance — the formula above describes the long-run average per round, not any individual round.
In bulk simulation, you can play thousands of rounds at once. What actually converges as the round count grows is the average net per round — it drifts toward the expected value of the simulator's cent-rounded payouts, which equals the theoretical house-edge figure when rounding has no effect (e.g. stake × target is a whole number of cents; with tiny stakes the floor can push it lower). A run that stops early because the bankroll ran out is a different, path-dependent result. The total dollar net does not get pinned down the same way: its standard deviation grows roughly with the square root of the number of rounds, so a longer run can still finish noticeably ahead of or behind the theoretical average, and a finite bulk session can perfectly well end in profit even under a fixed house edge. See the FAQ below for a worked example of that square-root growth, and the expected loss calculator for the variance maths in full.
Looking for an "Aviator demo"?
If you landed here searching for an Aviator, JetX or Spaceman demo: this simulator is not a copy of any of those games and does not use their exact algorithms, which are proprietary to their providers. It implements the same general crash-game math described above with its own provably fair seeds, so you can study how the odds and house edge behave without any of it being real money. Anyone claiming to sell a "predictor" or "hack" for a specific crash game is selling a myth — see the question below, and you can check this simulator's own rounds with the provably fair verifier. That verifier recomputes only this site's reference scheme; checking a real game's round needs the verifier and documented scheme that game's provider publishes.
Provably fair, verifiably
Every round on this page is generated from a server seed, a client seed and a round number (nonce). The page shows a SHA-256 hash commitment of the server seed before you play — so a seed changed afterwards would no longer match the displayed commitment — and you can reveal the raw seed at any time. Once you reveal it, each of the last 12 rounds links straight to the provably fair verifier, which recomputes the same bust point from the seeds, nonce and commitment hash alone — it only recomputes this site's own reference scheme, so it cannot verify a real-money round from an actual casino provider.
For the maths behind the numbers above, see expected loss calculator, risk of ruin calculator, and the related Mines and Plinko odds tools. Full derivations are on the methodology page.
