TLDR
Advertised RTP and real RTP are not the same measurement. The advertised number is a certified long-run figure a testing lab confirmed over millions of simulated spins. Real (or observed) RTP is what actually comes back across a finite, live sample of real bets. Over a small sample the two almost never line up, because variance dominates. And there is a sneakier gap: some slots ship with multiple RTP versions, so the certified number you read on one site may not be the version a given casino deployed. That is the difference RTPspy is built to watch.
What is advertised RTP, really?
Advertised RTP is the game's designed, certified return. The studio builds a paytable, symbol weights, and bonus frequency that settle on a target payback, and an independent lab verifies it. Labs like GLI, eCOGRA, and iTech Labs do this either by calculating the theoretical return from the game's math, or by simulating an enormous number of spins and measuring what comes back.
The key word is long-run. When a slot lists 96% RTP, that means across millions of spins the game is designed to return about 96 cents of every dollar wagered and keep about 4. That 4% is the house edge, baked into the math on purpose. It is a real, verified property of the game, measured over its entire lifetime across every player who ever spins it. It is not a promise about tonight, which is where almost all the confusion lives.
What is real (observed) RTP?
Real RTP is the number you get when you stop trusting the label and start counting actual money. Take a batch of real bets on a slot, add up everything wagered and everything paid back, and divide. That ratio is the observed RTP for that sample.
This is exactly what a live tracker does. RTPspy pulls real bet-feed data across many crypto casinos and watches what those slots actually return in the wild, not what a spec sheet claims.
The catch is sample size. Observed RTP over a few hundred spins is basically noise. Over tens of thousands it starts to mean something. Over millions it converges on the certified number, assuming the deployed game really is the version on the label. That is not always a safe assumption.
Do real RTP and advertised RTP actually match?
Over a large enough honest sample, yes, they converge. That is the whole point of the math, and it is why the certified figure is still the most useful number on a slot. Over a small sample, no, and you should not expect them to. Here is the honest version of why:
- Variance is huge in the short run. A high-volatility slot can pay nothing for hundreds of spins, then return more than 100% in one hit. Any short window reads wildly above or below the advertised number.
- Big wins carry the average. The certified RTP only shows up because a tiny number of massive payouts drag the whole thing up. Miss those in your sample and the observed number just looks broken.
- Samples are finite. A tracker never sees "millions of spins" on a fresh or quiet slot, so it should flag when a sample is too thin to trust.
So a short-term gap between real and advertised RTP is normal. It is variance, not a scandal. The number to watch is what the observed return does as the sample grows large. If it keeps drifting below the advertised figure on a big sample, that is when it gets interesting, because of the next problem.
Why the same slot can have two different RTPs
This is the part the glossy explainers skip. Many studios do not ship one version of a game. They ship several, at different RTPs, and let the operator choose which to deploy.
The same slot might be available at, say, a 96% build and a 94% build (and sometimes lower). The math, art, and feel are identical. The only difference is the dial. A casino running the lower version offers a worse deal on a game that looks, to the player, exactly like the better one. This is a documented industry practice, and lists of which providers offer reduced-RTP configurations circulate openly in player communities.
Now stack that on the first problem. The RTP you read on a review site is usually the headline or default version, but the version a specific casino actually loaded can be different, and the casino is rarely loud about it. So "advertised RTP" itself is slippery: advertised by whom, and is it the build you are actually playing?
You cannot settle this from a spec sheet. The only way to know what a slot is really returning at a casino is to measure the real bets flowing through it. That is the gap observed data closes.