There are three places to learn about a crash game: the provider's claim, the round history the game shows you, and rounds recorded independently over a long period. Each answers a different question. Clash Watchdog shows all three where they exist, labels each one, and never blends them into a single number.
Source 1: the provider's claim
What it is. The figures in the game's rules: the stated return to player, the maximum multiplier, the rules for cash-outs and bonuses. For example, Aviator is stated at 97% and Spaceman at 95.5%.
What it tells you. What the game is designed to do, and the price of play it implies. A stated 97% means a designed house edge of 3%.
What it cannot tell you. Whether the game in front of you behaves that way. A claim is a description of a design, made by the company that built it.
How we show it. Always labelled as a provider claim.
Source 2: the game's own round history
What it is. The strip of recent results every crash game shows, and, in provably fair games, the seed hashes and revealed seeds behind each round.
What it tells you. Exactly what happened in the last few rounds. In provably fair games, it also lets you check that a revealed seed matches the hash shown before the round, which shows the round was not changed after bets were placed. Our hash commitment check does that comparison in your browser.
What it cannot tell you. Anything about the next round, and very little about the game's return. Twenty or fifty results are far too few to test an RTP, and the strip invites the belief that a high result is due after low ones. History describes behaviour. It does not predict the next round. See The Streak Illusion and the patterns and hot hours check.
Source 3: rounds recorded independently
What it is. Rounds recorded one by one, from official provider demos and public live result streams, by a collector that runs separately from this website. The rounds are stored as per-round records, sealed each UTC day with a published SHA-256 fingerprint.
What it tells you. What the game actually did across thousands of rounds: the measured return at each cash-out target, the share of rounds reaching each multiplier, the median crash, the instant-crash rate. Each figure comes with its sample size, its time and a 95% interval. Once a game passes 10,000 rounds, those figures appear on its page beside the claim.
What it cannot tell you. Anything about a single future round. It also has limits worth stating plainly:
- Demo and live streams are what we can see. We record what a provider makes public. If a game behaved differently somewhere we cannot observe, a count of public rounds would not show it.
- Small counts are weak. Below the minimum sample, the lab shows how many rounds are still needed instead of a number. See Evidence Tiers.
- Some games cannot be checked round by round. For certified-RNG games such as Spaceman and High Flyer, and for live dice games, counting rounds is the only independent check there is.
Why keep the three apart?
Because the useful information is in the comparison.
- Claim and measurement agree. The counted rounds are consistent with the stated design, within the interval. The price of play is as advertised.
- Claim and measurement disagree. If the claim sits outside the measured interval, that is a finding about the counted rounds, reported with its sample size and time and re-checked on fresh data.
- History and measurement disagree with what players believe. This is the most common case. A player sees a strip of low results and feels the game has gone cold; the long count shows the same shape it always had. Our myth checks are built on exactly that gap: predictor bots, patterns and hot hours, doubling after a loss and is it rigged?.
Blending the sources would erase all three comparisons.
What sources do we not use?
Some kinds of data are deliberately left out:
- Screenshots and player-submitted histories. They are easy to select or edit, and there is no way to know which sessions were left out.
- Operator data feeds. We take no data or money from the sites that offer these games, and never link to them.
- Personal player tracking. We measure games, not people.
How can you check our work?
Every published figure is meant to be recomputed. A game's page lists the fingerprints of its sealed days, a sealed day's records are released on request to logged-in users through the data page (Telegram login), up to 5 requests in any 24 hours, and the methodology describes how each figure is calculated. Reproducibility explains the steps. If your count of the same records disagrees with ours, that is a finding about our work, and we want to hear it.
Was this page useful?
Anonymous: no account and no cookie.