RTP and volatility describe different parts of a game

RTP describes a game's theoretical long-run average return. Volatility describes how widely outcomes can be spread around a model's average. Two games can have the same RTP and produce very different patterns of results. Neither measure can promise what happens in one session.

The Gambling Commission's RTP explanations distinguish a long-run average from an individual play period, and its technical guidance discusses volatility and sample size when assessing observed performance. The examples below are original simplified models designed to make the distinction visible; they are not casino game specifications.

Two invented games with the same average

Imagine two independent one-unit games. In Game A, a round returns two units with probability 48% and zero otherwise. In Game B, a round returns 96 units with probability 1% and zero otherwise. In both, the average gross return per one-unit stake is 0.96 units.

Game A spreads awards across many more winning rounds. Game B concentrates awards into a rare large result. Their 96% expected return is identical, but the shape of their outcomes is clearly different. A session can therefore feel very different even when the theoretical average matches.

The models leave out real-game features and assume independent identical rounds. Their purpose is to separate average return from distribution. They do not identify which game a visitor should play or how much to stake.

Gross awards are different from net profit

A one-unit stake followed by a two-unit gross award produces one unit of net profit for that round. A zero award produces a one-unit net loss. Keep the stake and award in the same calculation rather than counting every credited award as profit.

In a real game, an award smaller than the stake can still reduce the account balance. An animation or sound effect accompanying it does not change that arithmetic. The relevant figure for the account is the net result after the stake has been deducted.

This distinction also matters when reviewing a session's turnover. Reusing credited awards can increase the sum of stakes, so a deposit amount alone does not describe all the money wagered. Read game information and RTP for that relationship.

A losing run does not create a debt

Under an independent fixed-probability model, previous losses do not make the next round more likely to win. In the invented Game B, the probability of a 96-unit award remains 1% after a loss, just as it was before that loss. There is no repayment obligation created by the earlier results.

The probability of no such award in 100 rounds is 0.99 to the power of 100, approximately 36.6%. That is a consequence of the model, not evidence that the next round is due. After those rounds, the assumed next-round probability remains 1%.

This is an illustration of independent events, not a claim about every game feature. Some games have stateful mechanics described in their rules. A visible progress meter or special feature must be read under those rules rather than used to justify a general due-to-pay belief.

Session return is an observation

You can calculate an observed gross return by dividing the sum of awards by the sum of stakes for a defined period. If 500 units are staked and 430 awarded, the observed return is 86%. That does not establish that the game's theoretical RTP is 86%.

The observed number comes from one sample. Its relationship to the model depends on the game's distribution, sample size and other assumptions. A short period can produce a high or low observation without proving the model has changed.

A meaningful record names the period, game, version, mode, total stakes and total awards. Combining different games or modes into one percentage can make a comparison harder to interpret. A screenshot of the final balance lacks several of those details.

Volatility labels are not a universal scale

Low, medium and high are often provider descriptions rather than a common numerical standard that every studio uses identically. Treat a label as an indication that needs context, not an exact measurement directly comparable across every game.

Look for published rules, feature probabilities or more detailed technical information when available. If a provider supplies only a broad label, a review should not invent a variance figure or a likely winning frequency. The absence of detailed data is an evidence limit.

Likewise, a large maximum award is not the same as its probability. A game can display an impressive maximum without telling you that it is likely in a session. The size of a possible award and the chance of receiving it are different quantities.

More play is not a solution to uncertainty

A long-run statistical statement does not tell an individual to keep gambling until an observed return approaches the advertised value. The player can run out of money, exceed personal limits or experience harm long before any large-sample behaviour becomes informative.

Do not chase a target return or increase stakes after losses. A negative expected-cost model remains a cost model even when a winning round is possible. No RTP percentage turns ordinary casino gambling into reliable income.

If the urge to continue is driven by recovering a loss or proving that a game must pay, step away. Support and account controls are available, and the exclusion-scope guide explains why the reach of a restriction matters.

Fairness and suitability are separate questions

A game can follow its published mathematical model and still carry a substantial risk of losses. Fairness checks address whether the game operates as specified; they do not establish that the product is suitable for a particular person's circumstances.

Similarly, provably fair verification can address a specific result-generation process without proving a favourable RTP or reliable payment handling. Separate those layers when reading operator claims.

Use both measures without overclaiming

The average helps explain a model's expected cost over turnover. The distribution helps explain why outcomes can vary so much around that average. A session record tells you what happened in a defined sample. Keeping those three concepts distinct is more useful than treating one percentage as a promise.

Read the game's actual information, keep the limits of the statistics visible and make decisions within your own spending and time boundaries. You do not need to gamble further to validate a theoretical average.