Crypto Crash Game Strategy: When to Cash Out (2026)

Crash is the game people are most confident they have solved, and the one where confidence costs the most. The curve looks like it has momentum, the chat fills with people calling the next round, and almost none of it survives contact with how the result is actually produced. What follows is the maths behind the multiplier, what a cashout target does and does not change, and the bet sizing that decides whether a session lasts twenty rounds or two hundred. None of it is a system for beating the edge, because there isn't one — it is how to spend a fixed number of rounds without handing them away.

How the multiplier is actually generated

A crash round is not a physics simulation. The server picks a crash point before the round starts, commits to it with a hash, and the rising curve is only animation. On a provably fair implementation like the one on this site, the round result comes from HMAC-SHA256 of the server seed, your client seed and a nonce. The resulting hex is converted to a number and mapped onto the crash curve with a formula close to crash = (2^52 / (h + 1)) / 2^52 * (1 - edge), then floored to two decimals. The important consequence: the curve is memoryless. A run of eight 1.00x busts changes the ninth round by exactly nothing.

The maths of a cashout target

With a 1% house edge, the probability of a round reaching multiplier m is roughly 0.99 / m. So 2.00x hits about 49.5% of the time, 3.00x about 33%, 10.00x about 9.9%, and 100x about 0.99%. Multiply probability by payout and you get 0.99 in every case — the expected value is identical at every target. That is the single most useful fact in crash: no cashout multiplier is mathematically better than any other. What changes is variance.

Variance is the thing you are actually choosing

At 1.20x you win about 82% of rounds but each win adds 20% of your stake. At 10x you win about one round in ten and need that win to cover nine losses. Over 500 rounds at a fixed 1% of bankroll per bet, a 1.5x target will usually leave you within about 15% of where you started; a 10x target routinely swings 60% either way. Pick your target from how long you want the session to last, not from a feeling about which number is "due".

Bet sizing beats cashout tuning

Risk of ruin is driven by stake size. Betting 10% of bankroll per round at a 2x target gives you a realistic chance of busting inside 30 rounds purely from a normal losing streak — eight consecutive losses under 2x happens roughly once every 400 rounds, and that alone costs 57% of a 10% bankroll. At 1% per round the same streak costs 8%. If you claim 10,000 free SPUNK a day, a 1% stake is 100 SPUNK, which is also the table minimum here — that is a deliberate design point, not a coincidence.

Why martingale fails specifically in crash

Doubling after each loss at a 2x target needs your bankroll to survive n consecutive misses, costing 2^n - 1 units. Eleven misses in a row — about a 1-in-2,000 event, so a couple of times per long session — needs 2,047 units to recover one. Table maximums and bankroll size both cut you off long before the streak is exhausted. See martingale in crypto gambling for the full loss table.

A cashout routine that survives contact

• Set the auto-cashout before the round starts. Manual cashouts add 150-400 ms of reaction time and that is where most of the "I was at 4x!" losses come from.
• Use two auto-bets if the client supports it: a 1.5x base bet for turnover and a small 20x bet as your lottery ticket.
• Fix the session in rounds, not in profit. 200 rounds, then stop, regardless of result.
• Rotate your client seed at the start of a session and record the previous server seed so you can verify past rounds afterwards.

Verify before you trust

Any crash game that will not show you a hashed server seed in advance is asking you to take the distribution on faith. After you rotate seeds, re-run the HMAC yourself and confirm the historical crash points match. The method is walked through step by step in our provably fair gambling guide.