How to Play Crash Gambling: Targets, Odds and Variance
A crash round from stake to cash out, and why the published formula takes the same 1% edge at every auto cash out target. Only the variance changes.
Photo: Dirk Schuneman on Pexels
Learning how to play crash gambling takes one round: you place a stake before the round starts, watch a multiplier climb from 1x, and cash out (by hand or with a preset auto cash out target) before the round crashes. Cash out in time and your stake is multiplied by the figure at that moment; miss the crash and the stake is gone. The part most guides skip: under the published crash formula, the house keeps the same share whatever target you pick, so your choice of target changes how bumpy the results are, not what you can expect to keep.
How does a crash round work from bet to cash out?
A crash round is a single hidden number, the crash point, revealed slowly. Everything you see on screen is the multiplier counting up toward a value that was already fixed before you clicked bet.
The sequence runs the same way on every crash table:
- Stake. You choose an amount during the short betting window between rounds.
- Set a target, or don't. An auto cash out target closes your bet the instant the multiplier reaches it. Without one, you cash out manually.
- The climb. The multiplier rises from 1x. Each player still in the round watches the same number.
- Cash out or bust. If the multiplier reaches your target (or you click first), you are paid stake times that multiplier. If the round crashes first, the stake is lost.
- Next round. A new crash point is drawn and the cycle repeats.
The only decision that matters is step 2. Timing a manual click feels like skill, but the crash point does not react to anyone's play, so a manual cash out is just an auto target chosen late. Our explainer on how crash gambling games generate the multiplier covers the hashing chain behind that hidden number in more detail.
Where does the crash point come from?
The crash point comes from a published formula applied to an integer derived from the game's seeds. In the seeding post that the open-source fair verifier quotes, the crash point is defined as the larger of 1 and 2^32 divided by (int + 1), multiplied by (1 - 0.01).
Two parts of that line do all the work:
| Part of the formula | What it does |
|---|---|
| 2^32 / (int + 1) | Turns the seed-derived integer into a raw multiplier. Small integers give huge multipliers, large ones give values near 1. |
| × (1 - 0.01) | Shaves the same proportion off every raw multiplier. This is the house edge. |
| max(1, ...) | Floors the result, so the lowest possible crash point is 1x. |
The verifier's comments state that a 0.01 edge parameter means a 1% house edge with a lowest crash point of 1. The floor matters: any round where the shaved value would fall below 1 simply crashes at 1x, before anyone can cash out at a profit.
Before launch, that operator also tied the seed to Bitcoin block 584,500, a block hash nobody could know in advance, so the sequence of crash points could not be chosen to suit the house.
Why does every auto cash out target return the same?
Every target returns the same on average because the edge is a constant multiplier on the crash point, not something that depends on where you stop. This is derived from the formula above, not a claim any source makes in these words.
Take any target T. Your bet wins if the crash point reaches T, which means 2^32 / (int + 1) × (1 - 0.01) has to be at least T. Rearranged, the raw value 2^32 / (int + 1) must be at least T divided by (1 - 0.01). If the integer is spread evenly across its range, as a hash-derived value is designed to be, the chance of the raw value clearing any level is roughly one divided by that level. So:
- chance of reaching T is roughly (1 - 0.01) / T
- payout when you do is T times your stake
- expected return is the product: (1 - 0.01) of your stake, whatever T is
The T cancels. A target twice as high is reached about half as often and pays twice as much, so the average lands in the same place. The 1% comes off the top on every target, from a modest one to a moonshot.
What does change is variance. A low target wins most rounds and loses rarely, so a session moves in small steps. A high target loses most rounds and occasionally pays a large multiple, so a session swings hard and can run a long losing streak before a hit. Same expected cost, very different ride. If you want to see that spread rather than take the algebra on trust, our crash gambling simulator piece on variance runs the same idea across many simulated rounds.
Does adjusting your strategy between rounds help?
No. Changing the target between rounds, or changing the stake, moves money between rounds but cannot change the proportion the formula removes from each one.
This is where "play again and adjust your strategy" advice falls apart. Each round's crash point comes from its own integer, and the formula has no memory of the last bust. Systems that double the stake after a loss or raise the target after a win only rearrange when you win and lose. Every individual bet still carries the same (1 - 0.01) expected return, so the sum of many such bets carries it too, and larger stakes simply mean a larger expected loss in absolute terms.
The honest uses of the target are practical ones: pick the variance you are comfortable with, and decide in advance how much you are prepared to lose in a session. The rest of our Crash Gambling coverage applies the same arithmetic to other questions about the game.
How do you check that a crash round was fair?
You check a crash round by recomputing its crash point from the revealed seed and comparing it with what the game showed. The formula is public, so anyone can run it.
Two details from the verifier sources are worth knowing before you rely on a site's own page:
- Proofs can cover more than independence. The bustabit crate notes that its scheme also proves the bust coefficient, meaning the number a round busted at, not only that rounds are independent of each other.
- Published formulas can be wrong. For Stake's related Limbo game, the author of the
fairverifier reports that the published formula did not reproduce results, and that the working client code floors before dividing: it floors float × 1e8 and adds 1. That correction was found by reverse engineering the client-side JavaScript. The lesson carries over to crash: verify against code that actually reproduces past rounds, not against a formula in a help article.
The provably fair calculator walkthrough shows how to reproduce the HMAC step by hand, which is the input to the crash formula above.
Questions on this
How does the crash gambling game work?
You stake before the round, a multiplier rises from 1x, and you cash out at any point before it crashes. The crash point is fixed in advance by a published formula with a 1% edge, so the climb only reveals a number already set.
What is the best strategy to win a crash bet?
No target or betting system raises the expected return. Under the formula every target keeps (1 - 0.01) of the stake on average; choosing a target only sets how often you win and how large each win is.
Does auto cash out change the odds?
Auto cash out changes the chance of winning a single round, not the expected return. A higher target wins less often in proportion to its larger payout, so the average stays the same.
Can you check a crash result yourself?
Yes. Take the revealed seed, derive the integer, and apply the published formula. Use a verifier that reproduces past rounds, since an independent author found a published Stake formula that did not.
Sources and on-chain links
- docs.rs (open-source fair crate): fair 0.0.13 crash.rs (verifier source quoting the Stake Crash seeding post), accessed November 2026
- docs.rs (bustabit crate): bustabit 0.2.1 Rust crate, accessed November 2026
- docs.rs (open-source fair crate): fair 0.0.13 limbo.rs, accessed November 2026
