RTP is an expectation over a defined model, not a promise for a session.
Hit frequency and volatility answer different questions from RTP.
Keep the tested mathematics, configuration and deployed version connected.
A fictional probability model explains the arithmetic. It is not a playable-game specification, certification, prediction of a player’s results or a strategy for winning. British standards are identified as jurisdiction-specific.
A fictional unit-stake game
| Outcome probability | Total payout | Contribution to expected payout |
|---|---|---|
| 70% | 0× stake | 0.00 |
| 20% | 2× stake | 0.40 |
| 10% | 5× stake | 0.50 |
| Total | 30% positive-payout frequency | 0.90 = 90% theoretical RTP |
This distribution is a teaching example, not a recommendation for a commercial game. An individual session can differ substantially from the expectation.
Calculate the expectation before interpreting a result
Let each outcome have a probability p and a total payout multiplier m for a unit stake. The theoretical payout expectation is the sum of p × m. In the invented distribution below, 70% of outcomes return nothing, 20% return twice the stake and 10% return five times the stake: 0 + 0.40 + 0.50 = 0.90, or 90% theoretical RTP. The table is deliberately simple enough to check by hand; production games have more states and dependencies.
Define a hit before reporting hit frequency
In this example, 30% of outcomes pay more than zero. All paying outcomes also exceed the unit stake, so positive-payout frequency and profitable-outcome frequency happen to match. They need not match in a real game: a payout below the stake may count as a hit while leaving a net loss. Document the denominator, treatment of feature rounds and whether the metric counts any payout or a net win. A headline percentage without that definition is ambiguous.
Separate observed RTP from the theoretical model
The Gambling Commission calculates observed RTP from winnings divided by turnover. Its monitoring guidance explains that interpretation also depends on volatility and the amount of play: smaller samples allow wider variation. A result below the theoretical percentage does not by itself establish a defect. Use the supplier’s documented model, sample definition and suitable statistical tolerance rather than inventing a universal spin threshold.
Evaluate distribution, not just one percentage
Two games can have the same expected payout but different frequencies and sizes of awards. For procurement, request a distribution summary, feature contribution, maximum payout and the assumptions behind any exposure calculation. Separate base-game, purchased-feature and other modes where their rules differ. Ask the maths owner to explain how changing a stake, configuration or feature price changes the model. Neither a high RTP nor a high hit rate establishes that a product is suitable for a given audience.
Keep randomness and presentation aligned
For relevant British remote products, RTS 7 addresses acceptably random outcomes and prohibits adaptive behaviour that changes outcome probabilities during play in response to player behaviour. Its detailed requirements apply within that regulatory scope. A product team should identify which rules and testing process apply to its actual deployment rather than treating a source link as evidence that the game is compliant.
Make release acceptance a versioned decision
Our recommended handover pairs the rules, paytable, model version, configuration identifier and test evidence with the exact deployed build. Include rounding rules, cancelled-round treatment and replay examples. Independently reconcile expected payouts against representative settlement messages, including interrupted or retried rounds. Keep visual celebration rules consistent with whether an outcome is a net loss or gain. A later configuration change should trigger an explicit review of affected evidence, with an owner who can stop release if those records diverge.
Sources and scope
A fictional probability model explains the arithmetic. It is not a playable-game specification, certification, prediction of a player’s results or a strategy for winning. British standards are identified as jurisdiction-specific. Sources reviewed 2026-10-18; check the linked originals for later changes.
Gambling Commission — Calculating and monitoring RTPGambling Commission — RTS 7: random outcomesCompare the same requirements across suppliers and record evidence, tests and unresolved gaps. The editable workbook starts with no supplier results or approvals.
Get the free supplier evaluation worksheetThis educational guide is independent, free to access and designed as a practical starting point. Regulatory and legal requirements should always be confirmed with the relevant authority and qualified advisers.
Our evidence methodology