A number about a game is only as strong as the evidence behind it. A provider's stated RTP is a claim. A few hundred rounds say almost nothing. Past 10,000 recorded rounds, a measured return with its 95% interval starts to test the claim, and larger counts can detect smaller gaps.
Why isn't every number equally strong?
Because results in a crash game are noisy. Each round either loses the stake or pays a multiple of it, so a small count of rounds can land far from the game's true return by chance alone. The more rounds you count, the closer the count settles to the truth, and the narrower the range of values it is consistent with.
A statement such as "this game returned 94%" means very different things after 300 rounds and after 300,000. The first is noise. The second is strong evidence. Reading a number well means asking how many rounds sit behind it.
What are the levels of evidence?
It helps to think of three levels. They are a reading aid, not grades: the lab does not award badges or rankings, it publishes each figure with its sample size, its time and its interval, and lets those speak.
Level 1: the claim. The provider's stated RTP, from the game's rules. It tells you what the game is designed to do. On this site it is always labelled as a provider claim. Many game pages are at this level today, because counting has not started or has not reached the minimum.
Level 2: a published measurement. At least 10,000 recorded rounds, with the measured return shown beside the claim, together with its sample size, the time of the count and a 95% interval. At this level a large gap between claim and measurement would show; a small one might not.
Level 3: a narrow measurement. Many more rounds, enough that the interval is tight. At this level even a small gap, a fraction of a percentage point at low cash-out targets, becomes visible.
How wide is the interval at each sample size?
For the return at a 2x cash-out in a game designed to return 97%, the 95% interval is roughly:
| Rounds counted | Interval, either side | What it can show |
|---|---|---|
| 100 | about 20 points | Nothing useful |
| 1,000 | about 6 points | Only a gross mismatch |
| 10,000 | about 2 points | A gap of a few points |
| 100,000 | about 0.6 points | A gap under one point |
Illustrative numbers, not lab measurements.
Higher cash-out targets need more rounds for the same precision, because wins are rarer. At a 10x target the interval at 10,000 rounds is close to 6 points either side. Understanding RTP shows the arithmetic.
What does the lab show at each stage?
Each game page carries a status that tells you which level you are looking at:
- Not started. No rounds have been counted. The page shows the provider's claim and nothing else.
- Counting. Rounds are being recorded but the count is below the minimum. The page shows how many rounds are still needed, never a provisional figure.
- Measured. The count has passed 10,000 rounds. The measured figures appear with their sample size, time and 95% interval, next to the claim.
Pages never show a placeholder number. An empty state is shown as empty. Browse the games list to see where each game stands, or open a single game such as Spaceman or Aviator.
Why not publish early and update later?
Because an early figure is easy to misread and hard to take back. A screenshot of "measured RTP 91%" after 400 rounds would circulate as proof the game is rigged, even though 400 rounds cannot show that. A figure of 103% would circulate as proof that a game "pays". Both would be noise.
Waiting for a minimum sample protects readers from that. It also matches how the lab answers the is it rigged? question: with counted rounds and intervals, game by game, never with impressions.
Does more evidence ever mean a game is "safe"?
No. Strong evidence that a game returns what its provider states tells you the price of play is as advertised. It says nothing about whether playing is a good idea, and nothing about the next round. History describes behaviour. It does not predict the next round.
It also does not cover everything. A count of rounds tests the return and the shape of results. Whether a single round can be checked for tampering is a different question, answered by the game's fairness system. For provably fair games, our hash commitment check confirms a revealed seed; see Provably Fair, Explained.
How should you read a claim of "proof" elsewhere?
Apply the same test to any number you see about a crash game, whether in a video, a signal group or a forum post:
- How many rounds? Under a few thousand, treat it as an anecdote.
- Is there an interval? A figure without one cannot be weighed.
- Can the data be checked? If the rounds cannot be obtained and recounted, or could have been changed after the fact, the figure rests on trust. See Reproducibility.
- Is it a claim or a count? A stated RTP and a measured one are different kinds of statement.
Predictor sellers in particular tend to show a handful of winning rounds and no denominator. The predictor bots check explains why a short winning streak is expected from any guess.
Was this page useful?
Anonymous: no account and no cookie.