A 96% RTP figure is a long-run average computed over millions of simulated spins. A busy evening on mobile might be eight hundred. The distance between those two sample sizes is where every real gambling experience lives, and it is why two people can play the same game on the same night and come away with completely different accounts of it.
Volatility — the industry usually calls it variance — describes how widely results scatter around that average. It is a separate property from RTP, and games are deliberately designed across the whole range of it.
Two games, same RTP, nothing else in common
Imagine two slots, both built to 96%.
The first pays small amounts often: a return on roughly a third of spins, nothing ever exceeding forty times the stake. Sessions on it drift. You lose slowly, occasionally finish ahead, and rarely feel much of anything. Low volatility.
The second pays on perhaps one spin in five, with the great majority of its 96% concentrated in a bonus round that arrives every few hundred spins and pays anything from twenty times the stake to several thousand. Most sessions on this game are worse than the average — considerably worse — and the average is held up entirely by the rare sessions that catch the feature. High volatility.
Both return 96% over a lifetime of play. Only one of them will ever hand somebody a story. The median session on the high-volatility game is much worse than the mean, because the mean is dragged upward by outcomes most players never see. When people say a game "feels tight", they are usually describing volatility rather than RTP, and they are usually right about the feeling and wrong about the cause.
How long a dry run is normal?
This is the question the design of high-volatility games makes unanswerable by intuition, so here is the arithmetic. If a game returns something on a proportion p of spins, and those spins are independent, then the chance of a run of n spins with nothing at all is (1 − p) multiplied by itself n times.
Try 20% hit frequency over 20 spins: about a 1.15% chance of nothing, or one attempt in eighty-seven. That sounds rare until you notice that a 600-spin session contains hundreds of overlapping twenty-spin windows. Rare events are guaranteed at volume. The correct response to a long blank run is not that the game has changed but that you are sampling a distribution with a long tail, and you have been sampling it for a while.
One honest caveat about the tool: "hit frequency" counts spins that return anything at all, including the very common case of a return smaller than the stake. On many games a large share of hits are net losses dressed as wins. Great Britain's rules now prohibit operators from celebrating a return equal to or below the stake with win sounds and animations, precisely because that presentation made losing spins register as wins. If a game shows you a hit frequency figure, read it as "spins where something happened", not "spins where you profited".
The gambler's fallacy has a respectable cousin
Everyone knows not to expect a red after ten blacks. Fewer people notice the more sophisticated version: believing that because RTP is 96%, a session that is down 60% is "due" a correction back toward the average.
It is not. Regression to the mean works by dilution, not repayment. If you are £120 down after a thousand spins and play a million more, your percentage return converges toward 96% — but it does so because the £120 becomes a rounding error against an enormous turnover, not because the game returns it. In pounds, expected loss only ever grows with turnover. The average improves; the balance does not.
This is the single most expensive misunderstanding in the category, because it converts a bad session into a reason to keep playing. The expected loss calculator shows the direction of travel plainly: every additional hour adds to the expected loss, and none of them recovers a previous one.
What variance is worth knowing for
Not for beating anything — volatility does not change the price, only the ride. It is worth knowing for two practical reasons.
It tells you how much money a game needs to be played at all. A high-volatility slot at a stake that gives you two hundred spins will, most of the time, simply end. The game's return is stored in events you have bought too few tickets to reach.
And it explains why other people's accounts of a game are useless as evidence. Somebody who won on a high-volatility slot is not describing a generous game, and somebody who lost heavily on the same title is not describing a rigged one. They are both describing single draws from a wide distribution, which is exactly why this site does not publish, aggregate or repeat that kind of testimony.
18+. If chasing losses is something you recognise in your own play rather than in the abstract, that is worth acting on. Deposit limits, time-outs and self-exclusion sets out what is available, and the National Gambling Helpline is free on 0808 8020 133.