Image: Nolimit City
A payout ceiling is not an outcome forecast
A slot's maximum payout tells you the upper award specified by its rules. It does not tell you the chance of that award, the chance of any smaller win, or how much a session will return. A large multiple and a favourable-sounding RTP remain different statistics, even when displayed next to each other.
The first comparison question is therefore 'what does this number measure?' A 99,999x ceiling is measured against a base stake. A hit-frequency percentage counts defined events. An RTP percentage compares theoretical returns with stakes under the model's conditions. None can substitute for an unpublished probability. If a source only supplies a maximum, our comparison should leave the odds unknown rather than produce an impressive-looking estimate.
Five measures and their denominators
| Measure | Question answered | Does not tell you |
|---|---|---|
| Maximum payout | What award ceiling do the rules specify? | Its probability |
| Maximum-win probability | How likely is that event in the stated model? | A guaranteed waiting time |
| Hit frequency | How often does the defined paying event occur? | Whether the return exceeds the stake |
| Bonus frequency | How often does the defined trigger occur? | The return from the feature |
| RTP | What theoretical share of stakes returns under the model? | A personal session result |
A frequency needs an event definition. 'Hit' can be misleading if readers assume it means a net gain while the source counts every positive award. A return of 10p after a £1 stake is a paying event but a 90p loss on that round. Keep the supplier's label when its precise definition is not available.
A probability also needs the mode and configuration. A base-game model and a direct feature purchase have different conditions. If the publication does not specify the relationship, do not assume that one figure covers every menu option.
Two provider examples, with their scope left visible
| Title | Maximum payout | Published max-win probability | Published hit frequency | Headline RTP |
|---|---|---|---|---|
| Mental 2 | 99,999x | 1 in 16 million | 31.41% | 96.06% |
| Tanked 3: First Blood 2 | 25,584x | 1 in 21 million | 20.24% | 95.99% |
These are the provider's public headline figures. We have not established their applicability to each purchased mode or every operator's configuration. They belong in a source-labelled reference table, not a league table of which game is most likely to make a reader money.
The example also shows why the largest maximum need not have the smallest published probability. Neither ranking gives a complete account of the distribution of smaller outcomes, costs or volatility. A source can describe several aspects of a model while still leaving important comparison questions unanswered. Link the source and retain those limits instead of filling the gaps with assumptions.
A million-to-one figure is not a countdown
A hypothetical event with a fixed probability of one in one million does not become due on the millionth attempt. Under an independent-trial assumption, its probability on the next attempt stays one in one million. The assumption matters: this illustration is general probability arithmetic, not a verified reconstruction of any named slot's implementation.
For 1,000 such hypothetical independent trials, the chance of at least one event is 1 minus (1 minus 1/1,000,000) raised to 1,000, approximately 0.09995%. That is close to one tenth of one per cent, not a promise that one in every thousand players will necessarily encounter it in a fixed batch.
Do not use this illustration to recommend a number of spins. It explains why rare-event figures should not be translated into a personal timetable. A long sequence without the event is compatible with a small fixed probability, and money already lost does not become a credit towards a future maximum.
Expected occurrences and probability of at least one differ
In an illustrative set of one million independent trials with a one-in-one-million event probability, the expected count is one. That does not mean every such set contains exactly one event. Some contain none; some contain more. Under those assumptions, the probability of at least one is approximately 63.21%, leaving approximately 36.79% with none.
The distinction between expected count and a guaranteed count is the reason 'one in N' should not be read as 'one after N'. It is also why a player's observed history cannot, by itself, establish that a model has failed. Verifying a supplied game needs its configuration, test evidence and appropriate analysis, not a comparison of one session with a headline.
A provider figure rounded to '16m' also lacks the precision needed for an exact comparative calculation. Preserve the published rounding. Do not manufacture extra decimal places or suggest a statistically certain advantage between modes whose probabilities have not been published.
Hit frequency can include a net loss
An invented five-round example makes the denominator visible. Suppose each round costs £1 and the returns are £0, £0.20, £0, £1 and £3. There are three positive returns out of five rounds, a 60% positive-return frequency in this tiny example. Total stakes are £5 and total returned is £4.20.
Only the £3 round returns more than its own stake. The £1 round breaks even, and the 20p round loses 80p despite paying something. The observed return ratio is 84%, but this is not a theoretical RTP estimate for a real game. It is a deliberately invented ledger showing that frequency, profit frequency and returned amount answer different questions.
If a review says 'wins often', ask which of those meanings it intends. An entertaining sequence of small awards can still cost money. A frequency statistic should never erase the amounts charged, and a clip should not call every positive display a profit.
RTP comparisons need a mode and a version
A hypothetical 96% model returns £960 in expectation per £1,000 wagered under its conditions; a 94% model returns £940. That two-percentage-point difference corresponds to £20 of model expectation on that turnover. It is not £20 guaranteed extra money in a session.
Use our RTP inspection workflow to record the figure in the actual rules before assigning a setting to a casino. Turnover is cumulative wagering, not simply the deposit. A separately priced purchase also needs its own denominator and published return figure if you want to compare it with ordinary play.
If the mode's return is unavailable, label it unavailable. Do not infer a purchase-mode RTP from the main product page or infer max-win probability from RTP. Many outcome distributions can share the same average while having very different frequencies and extremes.
Do not mistake missing data for zero probability
An unpublished max-win probability does not mean the probability is zero. It means this reference cannot establish it from the sources checked. Keep that distinction when comparing a documented title with an undocumented one: the second is not automatically more or less likely to reach its ceiling. Similarly, a rounded statistic cannot justify a precise calculation to several decimal places. Unknown, rounded and configuration-specific are useful labels that preserve the limits of the available evidence.
A useful comparison can contain unknowns
Keep the source URL, checked date, event definition, mode and configuration beside each figure. Leave an empty field when the supplier has not supplied the required information. An honest two-title table with limits is more useful than a larger table that mixes natural triggers, bought rounds and unlabelled casino settings.
Past highlights also need selection context. A stream compilation may show unusually large results while omitting losses, costs and the full number of rounds. Its largest award is evidence of that recorded outcome, not an estimate of how likely the outcome is. Our Mental 2 explanation and Tanked 3 guide keep mechanics and availability separate from promises about future results.
FAQs
Does a bigger slot maximum mean a better chance of winning?
No. Maximum payout describes an award ceiling, not its probability or the chance of making a net gain. A comparison needs separate figures for the same mode and configuration. If the provider does not publish those probabilities, leave them unknown rather than calculating them from the maximum or RTP.
Does one in a million mean I win after a million spins?
No. In a hypothetical independent model it describes an event probability, not a countdown. One million trials with a one-in-one-million probability have an expected count of one, but not a guaranteed count of one. This general arithmetic does not recommend playing that many rounds or predict a named game.
Is hit frequency the same as profitable-spin frequency?
Not necessarily. A positive return can be smaller than the stake and therefore still produce a net loss. Check what the source defines as a hit. In our invented £1-round example, a 20p return is a paying event but loses 80p on that round.
Can RTP reveal a slot's max-win probability?
No. RTP describes a theoretical average return under stated conditions. Different distributions of outcomes can have the same average, so the percentage does not uniquely determine how frequently the maximum occurs. Use a supplier's explicit, mode-labelled probability if available, and keep its published precision and limits.
18+ only. Gambling can be addictive; set limits and never chase losses. Free, confidential support: BeGambleAware.
