Why You Cannot Measure a Game RTP in a Short Session
RTP is a long-run average, not a session forecast
Return to player (RTP) is the share of everything wagered on a game that the game is mathematically expected to pay back across its full lifetime. A 96% RTP means that for every $100 staked — across every player, every session and every spin the game will ever see — the design returns $96 on average. The remaining 4% is the house edge, and it is permanent: no pattern of play changes it.
The word doing the heavy lifting is average. RTP is a property of a game’s mathematical model, not a description of what any individual session will look like. Treating it as a forecast — “this is a 96% game, so I should get roughly 96% back tonight” — is the single most common misreading of the number, and this guide is about why that reading fails and how badly it fails.
If you want the underlying definition rather than the measurement problem, Casino Capybara’s guide to how slot RTP works covers what the figure is and where it is published. This page assumes you have that and asks a narrower question: could you ever check it yourself?
Does 96% RTP mean I get 96% of my money back?
No, and not approximately either. Over any realistic number of rounds, your personal return can sit an enormous distance from the published figure in both directions without anything being wrong with the game.
The reason is that a slot’s payouts are not clustered around the average. Most spins return nothing. A minority return a small multiple of the stake. A very small number return a very large multiple, and those rare outcomes carry a large share of the total returned. The average is real, but it is assembled out of outcomes that individually look nothing like it — so the average only emerges once you have seen enough of the rare outcomes for them to appear in their correct proportion.
That is a statement about sample size, and sample size is something we can put numbers on.
How many spins does it take to see a slot’s RTP?
If each round returns some fraction of your stake, and those rounds are independent, then the accuracy of your measured return improves with the square root of the number of rounds. The standard error of your estimate is:
standard error = σ ÷ √n
where n is the number of rounds played and σ is the standard deviation of the return on a single round, expressed in units of the stake. To be roughly 95% confident that your measured return sits within a given distance of the true RTP, you need about n = (1.96σ ÷ margin)² rounds.
Everything turns on σ. A low-volatility game paying frequent small wins might have a per-spin standard deviation around 5× the stake; a high-volatility game with a large top prize can sit at 12× or considerably more. Those two figures are illustrative rather than universal, but they bracket a lot of real games. Here is what each demands:
| To pin RTP down to within | Lower-volatility game (σ = 5) | High-volatility game (σ = 12) |
|---|---|---|
| ±10 percentage points | ~9,600 spins | ~55,000 spins |
| ±5 percentage points | ~38,000 spins | ~221,000 spins |
| ±2 percentage points | ~240,000 spins | ~1,383,000 spins |
| ±1 percentage point | ~960,000 spins | ~5,532,000 spins |
Read the bottom row carefully. Distinguishing a 96% game from a 95% game — a difference that genuinely matters over a lifetime of play — takes somewhere between roughly one million and five and a half million spins on a single game. At a brisk five seconds per spin, five and a half million spins is around 320 days — well over ten months of play, running day and night without a break.
What a single session can actually tell you
Turning the same formula around: given a session length, how wide is the range of returns that is completely unremarkable? The figures below are the approximate 95% margins around the game’s true RTP.
| Rounds played | Lower-volatility game (σ = 5) | High-volatility game (σ = 12) |
|---|---|---|
| 100 | ±98 percentage points | ±235 percentage points |
| 500 | ±44 percentage points | ±105 percentage points |
| 5,000 | ±14 percentage points | ±33 percentage points |
| 50,000 | ±4.4 percentage points | ±10.5 percentage points |
| 500,000 | ±1.4 percentage points | ±3.3 percentage points |
A 500-spin session on a high-volatility game carries a margin of roughly ±105 percentage points. Around a 96% expectation, that means almost any result from returning nothing at all to roughly doubling your money is statistically ordinary. The session contains essentially no information about the game’s RTP. It is not a weak signal — it is not a signal.
Even 50,000 spins on that game leaves a margin of about ±10 percentage points, which still cannot separate a 94% game from a 96% one. This is why no amount of personal record-keeping, and no streamer’s session, can establish what a game’s RTP is.
Why most short sessions land below the RTP
There is a further asymmetry that the margins above understate. Because slot payouts are heavily skewed to the right — a long thin tail of rare large wins pulling the average up — the typical session outcome sits below the average, not on it.
Put plainly: the mean is dragged upward by outcomes most sessions never contain. So the median player, over a short run, gets back less than the RTP. That is the mathematically expected result of a correctly working game, not evidence of a problem. It is also why “I never get anywhere near the advertised RTP” is such a common observation and such a poor diagnostic.
The same skew means the standard margins above are themselves an approximation that behaves worst exactly where people most want to use it — at small sample sizes. The real short-run picture is less symmetric and, if anything, less favourable than the tables suggest.
Why you usually can’t even run this calculation
Everything above depends on σ, the per-round standard deviation. Published RTP figures are common; published per-round standard deviations are not. Providers may describe a game’s volatility qualitatively — low, medium, high — or give a proprietary volatility index on their own scale, but the actual number you would need is rarely disclosed.
So in practice you cannot compute how long you would need to play to verify a specific game, because one of the two required inputs is not made public. You are left knowing only that the answer is large. Casino Capybara’s guide to slot volatility covers what the published volatility bands do and do not tell you.
The games where RTP doesn’t need measuring
Not every game has this problem. Where outcomes come from a known, fixed structure rather than a proprietary reel design, the return is derivable exactly — no measurement required.
European roulette has 37 pockets and pays 35 to 1 on a single number. The expected return per unit staked is therefore 36 ÷ 37 = 97.297%, giving a house edge of 1 ÷ 37 = 2.703%. American roulette adds a second zero for 38 pockets, so the same bet returns 36 ÷ 38 = 94.737%, a house edge of 5.263%. Those figures are arithmetic. Nobody needs to spin a wheel to establish them, and no session result can revise them.
Blackjack works similarly: the house edge follows from the rule set and the player’s decisions, and falls well under 1% under common rules with accurate basic strategy. Casino Capybara’s guide to the house edge sets out how these compare across games.
The practical consequence: for table games you can know the edge exactly in advance. For slots you are relying on the provider’s published model and the testing that verifies it, because independent field measurement is not realistically available to you.
Things that change the RTP you’re actually playing
Even the published figure may not be the one in front of you.
- Multiple configurations of the same game. Many slots ship in several RTP versions, and which one is deployed is an operator choice. The headline figure quoted in a review may not be the version you are playing — check the game’s own information panel, not a third-party listing.
- Feature and bonus buys. Buying into a bonus round frequently carries a different RTP from base play, sometimes higher and sometimes lower.
- Side bets. On table games, side bets typically carry a substantially worse edge than the main game, and they are not included in the headline figure.
- Bonus terms. Wagering requirements, game weighting and maximum-bet rules change the effective return on money played with a bonus attached, independently of the game’s own RTP.
Each of these is a reason two players on nominally the same game can be playing different mathematics.
A game is never “due”
Each round is an independent event. A random number generator retains no memory of previous outcomes, so a long losing run does not make a win more likely, and a large win does not make the game “tighten up” afterwards. RTP is not a quota the game works towards and then corrects for.
This is the gambler’s fallacy, and it is worth naming because it is the point where a misunderstanding of RTP turns into a costly decision — chasing losses on the belief that a return is owed. Nothing is owed. If your play has stopped feeling like a choice, Casino Capybara’s guide to responsible gambling and self-exclusion sets out the tools available.
Where published RTP figures actually come from
Since RTP cannot be measured from live play in any reasonable timeframe, published figures are not measurements. They are calculated from the game’s mathematical model — the full outcome space and the probability of each result — and then verified by independent test houses before the game is certified for a regulated market.
Testing laboratories such as eCOGRA and Gaming Laboratories International examine the game’s maths and its random number generator, and run simulations over very large numbers of rounds to confirm the implementation matches the stated model. Regulators set the standards those tests work to; in Great Britain, for instance, the Gambling Commission’s remote gambling and software technical standards govern how games must behave and what must be disclosed.
This is the honest answer to “how do we know the RTP is right?” — not because anyone observed it in the wild, but because the model was audited and the software was tested against it. That is also why checking a game is properly licensed and tested matters more than any personal results log. Casino Capybara’s guide to checking a casino’s licence covers how to verify that.
FAQ
How many spins does it take to reach a slot’s RTP?
Far more than any individual will play. Pinning a game’s RTP down to within one percentage point takes roughly one million spins on a lower-volatility game and over five million on a high-volatility one. Even 50,000 spins leaves a margin of several percentage points, which is not enough to separate a 94% game from a 96% one. There is no session length at which the published figure becomes personally verifiable.
Does a 96% RTP mean I get 96% of my money back?
No. It means the game is designed to return 96% of everything ever staked on it, pooled across all players and all sessions. Your own return over any realistic number of spins can sit tens of percentage points either side of that without anything being wrong. RTP describes the game, not your session.
Why did I get much less than the advertised RTP?
Because that is the ordinary outcome. Slot payouts are skewed — a small number of rare large wins carry much of the total returned — so the average is pulled above what a typical short session contains. The median short-run result sits below the RTP. Getting less than the headline figure over a few hundred spins is expected behaviour from a correctly functioning game.
Can you tell if a slot is rigged from your own results?
No. The variance over any personally achievable number of spins is far larger than the difference between a fair game and an unfair one, so your results cannot distinguish the two. Assurance comes from the game being certified by an independent test house and the operator holding a licence you can verify with the regulator — not from a results log.
Is RTP calculated per player or across everyone?
Across everyone. RTP is defined over the total amount wagered on the game by all players over its lifetime. Your own play is a single small sample from that pool, which is precisely why your personal return and the published figure need not resemble each other.
Does a game become due for a payout after a losing streak?
No. Each round is independent and the random number generator has no memory of what came before. A losing run does not make a win more likely, and the game does not correct itself to hit its RTP. Believing otherwise is the gambler’s fallacy, and it is the point at which a misreading of RTP starts costing money.
Sources
- UK Gambling Commission — Remote gambling and software technical standards: gamblingcommission.gov.uk
- eCOGRA — independent testing and certification of gaming software: ecogra.org
- Gaming Laboratories International — game mathematics and RNG testing: gaminglabs.com
- Roulette and blackjack figures in this guide are derived arithmetically from the games’ fixed structures and are shown in full above, so they can be checked directly rather than taken on trust.