How crash games work

In a crash game a multiplier climbs from 1.00x until it stops at a random point. Cash out before the stop and your stake is multiplied; wait too long and the stake is lost. The stopping points are distributed so that every cash-out target returns less than 100% on average. Timing does not change that.

By Nairobi Ultimate Editorial (prepared with AI tools under our editorial policy) · Last updated 2026-10-04

What happens in one round

A round has three stages. Players place stakes during a short betting window, a multiplier starts at 1.00x and rises on screen, and at some point the round stops ("crashes").

A player who cashes out before the stop is paid the stake times the multiplier shown at that moment. A player who has not cashed out when the round stops loses the stake. The rising line is an animation of a result produced by a random number generator; it is not a market, a flight or anything a player can read.

The only decision is when to leave. That makes the game feel like a test of nerve or timing, but the price of the game is set by how the stopping points are distributed, and the player has no influence on that.

How the curve is priced

Start with a version that has no margin. If the chance of the multiplier reaching m were exactly 1/m, then a 2.00x target would be reached half the time and a 10.00x target one time in ten. Stake times payout times probability would equal the stake at every target: a fair game.

A commercial game shaves that probability. In the standard model the chance of reaching m is (1 − h)/m, where h is the house margin. One way to build this is to let a share h of rounds stop at 1.00x, where nobody can cash out, and spread the remaining rounds along the fair curve.

average return at any target m = m × (1 − h)/m = 1 − h

The target cancels out of the formula. Whatever multiplier you aim for, the average return is the same, and it equals the game's RTP.

The same cost at every target

The table applies the model with a margin of 3%. The margin is chosen for the example only; the figure for a real game is the one in its displayed rules.

Illustrative example, standard model with a 3% margin. A 2.00x target looks like a coin flip, yet it loses 51.5% of the time. That small tilt, repeated every round, is the whole business model.

What does change with the target is volatility. Low targets give frequent small wins and a slow, steady drain. High targets give long losing runs broken by rare large wins, with the same average drain.

Why timing, patterns and systems do not help

The history bar that shows recent crash points invites pattern reading. If rounds are independent, as certified randomness requires, a row of low results tells you nothing about the next one. The belief that a big multiplier is "due" is the gambler's fallacy in a new costume.

Staking systems fail for the same reason they fail at roulette. Doubling after each loss at a 2.00x target needs only a short losing run to demand a very large stake: in the model above, five losses in a row happen about once in 28 attempts and ten in a row about once in 760.

Features such as automatic cash-out or placing two stakes in the same round change convenience and the spread of results. Each stake still returns 1 − h on average, so two stakes are simply two purchases of the same priced product.

Software or paid "signals" that claim to forecast the stop point cannot work against a properly tested generator. Paying for them adds a second loss on top of the first.

Speed is the hidden cost

Crash rounds are short, so a lot of money passes through the margin in little time. The cost of a session is the margin multiplied by everything staked, not by the deposit.

Because winnings are usually staked again, a small starting balance can be turned over many times in one sitting. Setting a session limit before the first round is the only control that works on this cost.

"Provably fair" and what it proves

Some crash games publish a cryptographic hash, a kind of digital fingerprint, of the data that fixes a round's result before bets are taken. After the round the data is revealed, and anyone can confirm it matches the fingerprint.

This shows the result was fixed in advance and not adjusted in response to the bets. It says nothing about the margin, which sits in the distribution of results itself. A game can be perfectly verifiable and still cost 3%, 5% or more of every stake.

What Kenyan law requires around such a game

The Gambling Control Act, 2025 defines a casino to include an "approved virtual or online platform" where a person may take part in "a game approved by the Authority" (section 2). The Conduct of Gambling Operations Regulations, 2026 add that a licensee must "seek approval from the Authority to introduce any new mechanics in the operations" (regulation 57(2)(b)).

Regulation 45 requires the game rules to be clearly displayed before any wager, the random number generator to be certified, and the theoretical return to player percentage for each game to be disclosed. Regulation 50(1) requires platforms to be secure from "bots or automated bulk entries that could distort the results". The detail is in the guide to how RNG testing works in Kenya.

Section 71(1) of the Act sets a floor of twenty shillings for an online bet. Regulation 87 obliges licensees to give players deposit, loss and session limits and reality checks, which matter more in a fast game than in a slow one.