House Edge and Variance
A worked expected-value example showing why the average result of playing is a loss — and why short sessions swing so much.
Two numbers describe almost everything about the money side of a crash game. House edge is the average share of each stake the game keeps over the long run. Variance describes how far individual results scatter around that average. Understanding both helps explain why a session can feel lucky or cursed while the underlying cost stays the same.
Important: these numbers are illustrative
The figures below use a simple, made-up model with a 3% house edge so the arithmetic is easy to follow. They are not the return-to-player (RTP) of Sky Fryer, Flying Fish, Submarine Dive, or any specific game. Always check the rules or information panel of the game you are looking at for its stated RTP.
The model
In a common crash design, the chance a round reaches at least multiplier m before crashing is roughly (1 − edge) ÷ m. With our illustrative 3% edge, the chance of reaching 2.00× is 0.97 ÷ 2 = 48.5%, and the chance of reaching 10.00× is 0.97 ÷ 10 = 9.7%.
Worked example: auto cash-out at 2.00×
Suppose you stake 1.00 unit with auto cash-out at 2.00×.
- With probability 0.485 you cash out and receive 2.00 units: a net gain of +1.00.
- With probability 0.515 the round crashes first: a net loss of −1.00.
Expected value = (0.485 × +1.00) + (0.515 × −1.00) = 0.485 − 0.515 = −0.03 units per round. Put another way, on average you lose 3% of each stake.
Same edge, different target: 10.00×
- With probability 0.097 you receive 10.00 units: a net gain of +9.00.
- With probability 0.903 you lose the stake: −1.00.
Expected value = (0.097 × 9.00) + (0.903 × −1.00) = 0.873 − 0.903 = −0.03 units. The average cost is identical. Changing your target changes how often you win and by how much, but in this model it does not change the edge.
Where variance comes in
Over 100 rounds of 1-unit stakes, the expected total in both cases is about −3 units. What differs is the spread. At 2.00× you win close to half the rounds, so totals usually land within a modest band around −3. At 10.00× you win fewer than one round in ten; a session of 100 rounds can realistically swing from a sizeable gain to a loss of 30 units or more, purely by chance.
That spread is why short-term results say little about whether a choice was "good". A winning evening at 10.00× does not show the setting beats the game, and a losing evening at 2.00× does not show the game is unfair. Only very large numbers of rounds bring the average close to the expected value — the law of large numbers.
What this means in practice
- Treat the expected loss as the price of entertainment, and budget for it.
- Higher targets mean longer losing runs are normal, not a sign something is "due".
- Staking progressions do not change the per-bet edge; see Crash Game Myths.
- Set spending and time limits before you start, as described in Bankroll and Risk.
18+ only. Crash games involve financial risk and there is no guaranteed profit. Help is available via our responsible gaming page.