Observed vs Stated RTP: What a Real Session Shows
Every review on this site carries two return figures: the RTP the provider states, and the return our own logged session produced. They almost never match, and that mismatch is not a scandal. It is the whole story of how slots behave over a normal night, told in two numbers.
Two numbers, two questions
Stated RTP is the game's certified long-run average, measured by the studio over millions of simulated spins. It answers one question: what share of all stakes does this game return over its full cycle? For a primer on that figure, see our RTP guide.
Observed RTP is arithmetic on a real session: total wins divided by total stakes over the spins I logged. It answers a different question: what did one honest sitting of this game actually pay? It is a sample, and we always publish the spin count next to it so you can judge the sample for yourself.
Why they almost never match
Sample size. A stated figure needs millions of spins to settle. A session is a few thousand at best, and at that scale the rare, heavy outcomes that carry a big share of the return, top jackpots and maximum features, usually have not happened yet. Remove the rare events from a small sample and the observed number lands below the stated one. Catch one of them, and it lands far above.
Both directions are normal. The same game, at the same stake, gave 90.0% in one logged session and 99.63% in another. Neither run was broken, and neither run was the truth. They were two slices of one distribution.
A worked example from our logs
Sweet Rush Bonanza by Pragmatic Play states an RTP of 96.50%. Here is what the three logged demo sessions behind our review returned:
| Session | Spins | Observed RTP |
|---|---|---|
| Session A | 1,170 | 90.0% |
| Session B | 1,629 | 99.63% |
| Session C | 2,626 | 93.6% |
| Combined | 5,425 | 94.63% vs stated 96.50% |
Even 5,425 spins sit short of the stated figure, and the review explains why: no complete bonus round in the pool beat 150.05x against a 5,000x cap, and the heaviest single hit, a 341.4x base-game chain, landed exactly once. The top of the distribution is exactly the part a small sample tends to miss. The full breakdown, session by session, is in the Sweet Rush Bonanza review.
One full session log from this portal is now public, checkpoint by checkpoint and round by round, as the session report for LAVB-20260730-01. The page shows the arithmetic a single observed figure has to close before it is printed anywhere.
What a gap means, and what it does not
A low observed figure over a few thousand spins does not mean the game is rigged, and a high one does not mean it is generous. One session cannot convict or acquit a math model. What the gap does show is character: where the return hides. A game that pays most of its way through the base game and parks the rest in a cap you never saw plays very differently from one that drips its full return in small change.
How we publish the pair
We never swap one number for the other. The stated figure describes the math; the observed figure describes a sitting. Each review shows both, with the spin count, the stake, the feature log and the balance path behind the session. The process and its limits are documented on the methodology page.
Quick answers
How many spins would it take for observed to match stated?
Far more than any human session. Depending on the game's volatility, convergence to within a fraction of a percent can take hundreds of thousands to millions of spins. This is why we treat a session as a portrait of behaviour, not a measurement of RTP.
Is a session with observed RTP above 100% possible?
Yes, and it proves the same point in the other direction. Catch one heavy feature in a modest sample and the session ends in profit. The long-run average still stands below 100%; you just saw a favourable slice of it.
Why publish observed RTP at all if it cannot match the sheet?
Because it is the one number that closes a real sitting: wins over stakes, from the first spin to the last. It says how the session ended, and it says nothing about how it ran. A sitting that paid once and a sitting that refunded every spin can both close at 100%. The share of spins that paid, the length of the dry stretches and the depth of the balance dip are separate measures. The session report carries them in its checkpoint table and bonus register. The stated figure tells you the price of the game; the observed one tells you what a sitting cost, and the report tells you how.
Former studio-side game mathematician: paytables and RTP models, now checked from the player's side. Full profile and method →