What is the best cash-out multiplier in Crash?

In expectation, none: the chance of the round reaching x is 0.99 ÷ x, so every auto cash-out returns 99% of the stake on average. 1.5× hits 66% of rounds, 2× hits 49.5%, 10× hits 9.9%. Lower targets win more often and pay less; the product does not move.

Does "cash out at 2× and martingale" work?

It is martingale on a 49.5% coin, exactly as on Dice. With a 1,000 bankroll and 10 base the doublings cover 6 losses; the 7th miss in a row appears in 57% of 200-round sessions and 99% of 1,000-round sessions. Each completed cycle wins 10; the wall loses the bankroll. The realised return on turnover still sits near −1%.

Why does Crash sometimes end at 1.00×?

Because 0.99 ÷ x is capped at 99%: 2% of rounds crash before reaching 1.01×, and nobody cashes out in time. That one percent is the entire house edge. Every multiplier above it is priced fairly relative to the others, which is why the expectation is the same at every target.

Crash guidePublished 2026-09-27

Crash strategy: what cashing out actually does

Crash strategy is a choice of threshold, and every threshold is worth the same. The crash point is max(1, floor(99 ÷ (1 − X)) ÷ 100) for a uniform X, which makes P(reach x) = 0.99 ÷ x exactly, so cashing out at x returns x × 0.99 ÷ x = 0.99 of the stake in expectation, whether x is 1.1 or 1,000. This guide computes the chance of reaching each target and the expected value at every one, then works through what “cash out at 2×” with martingale does to a bankroll, all on the same formula the free Crash game runs on its browser-side hash chain.

How Crash pays

Each round's hash is keyed with a public client seed through HMAC-SHA256, the first 52 bits become X in [0, 1), and the crash point follows. The floor and the two-decimal rounding do not change the headline: for any two-decimal target x, the point reaches x exactly when 1 − X ≤ 0.99 ÷ x, which happens with probability 0.99 ÷ x. Cashing out manually just picks x with your thumb instead of a box.

What the cash-out target does

Crash: chance of reaching each target, the wait between hits, and the return
Cash out at	P(reach)	Reach, one in	P(miss 10 in a row)	Expected rounds per hit	Return
1.01×	98.0%	1 in 1.0	9.3e-16%	1	99%
1.10×	90.0%	1 in 1.1	1.0e-8%	1.1	99%
1.25×	79.2%	1 in 1.3	0.000015%	1.3	99%
1.50×	66.0%	1 in 1.5	0.002064%	1.5	99%
2.00×	49.5%	1 in 2.0	0.11%	2	99%
3.00×	33.0%	1 in 3.0	1.82%	3	99%
5.00×	19.8%	1 in 5.1	11.01%	5.1	99%
10.00×	9.9%	1 in 10	35.26%	10.1	99%
25.00×	4.0%	1 in 25	66.76%	25.3	99%
100.00×	0.990%	1 in 101	90.53%	101	99%
1,000.00×	0.099%	1 in 1,010	99.01%	1,010.1	99%

Return is P(reach) × target, 99% on every line. At 1.01× a single miss (2% of rounds) costs a hundred wins. At 10× you wait about 10 rounds per hit and ten straight misses happen 35% of the time.

What “cash out at 2×” with martingale does

Auto cash-out at 2× wins 49.5% of rounds and pays 2×: a 49.5% coin, identical to the classic Dice roll. Martingale doubles after a miss so that the next hit recovers everything plus one base bet. From a 1,000 bankroll at a 10 base the doublings cover 6 misses; the 7th stake would be 640 against 370 remaining, so that is the wall.

Martingale at 2×, 1,000 bankroll, 10 base: the wall and how often a session hits it
Rounds in session	P(a run of 7 misses)	Win per cycle	Loss at the wall	Cycles one wall undoes
200	56.6%	10	1,000	100
1,000	98.6%	10	1,000	100

The probability is exact, from a Markov chain on the current losing streak with a miss at 50.5%. 57% of 200-round sessions and 99% of 1,000-round sessions meet the wall, and one wall costs 100 winning cycles.

Reproduce it: martingale, auto cash-out 2×, 200 rounds, 2,000 sessions. The median session looks fine; the ruin rate and the worst drawdown tell the rest, and the return on turnover reads about −1%. Compare flat betting at 2× (a steady drift, almost no ruins), martingale at 1.5× (wins 66% of rounds, but a double no longer recovers a loss: the exact factor is 3×, so the wall of 7 misses, hit in 7% of sessions, is reached with the losses only partly recovered), flat betting at 10× over 1,000 rounds, and reverse martingale at 2×, which trades the other way.

Cashing out manually does not escape any of this: the point was fixed when the hash was committed, and your thumb picks a threshold just like the box does. Waiting for a run of low crashes assumes the chain has memory; each round's hash is a separate SHA-256 link and the multipliers are independent. The Dice page shows the same wall from the chance side.

The curve is decoration on a Dice roll. Cashing out at x is a bet that wins with chance 0.99 ÷ x and pays x; every such bet returns 99% of its stake on average. A staking plan sets the stake before the hash is used, so a session's expected loss is 1% of its turnover under any plan. Martingale at 2× makes many small wins and one large loss; flat betting at 10× makes long droughts and rare paydays. Neither moves the −1%.

Understood the maths? Craze runs the real thing.

Same published odds, real stakes. PlinkoLab is sponsored by Craze; whether to play for money is entirely your call.

Play at Craze

Questions

What is the best cash-out multiplier in Crash?: In expectation, none: the chance of the round reaching x is 0.99 ÷ x, so every auto cash-out returns 99% of the stake on average. 1.5× hits 66% of rounds, 2× hits 49.5%, 10× hits 9.9%. Lower targets win more often and pay less; the product does not move.
Does "cash out at 2× and martingale" work?: It is martingale on a 49.5% coin, exactly as on Dice. With a 1,000 bankroll and 10 base the doublings cover 6 losses; the 7th miss in a row appears in 57% of 200-round sessions and 99% of 1,000-round sessions. Each completed cycle wins 10; the wall loses the bankroll. The realised return on turnover still sits near −1%.
Why does Crash sometimes end at 1.00×?: Because 0.99 ÷ x is capped at 99%: 2% of rounds crash before reaching 1.01×, and nobody cashes out in time. That one percent is the entire house edge. Every multiplier above it is priced fairly relative to the others, which is why the expectation is the same at every target.

Keep going

Last reviewed 2026-09-27