How we test

The problem this page solves

An enormous proportion of casino reviews are written by people who have never opened an account at the casino they are reviewing. The operator supplies a media kit with the game count, the licence number and a list of payment methods. Someone rewrites it, adds a score out of five arrived at by feel, and publishes.

You cannot tell the difference from the outside. A review assembled that way reads exactly like one built from testing, because both contain the same figures — the media kit figures are usually accurate. What differs is everything the media kit does not mention: how long a withdrawal actually takes, what triggers a verification request, whether the provably fair system survives someone checking it independently, and which claims fall apart when you read the terms in full.

So the only useful thing a review site can do is publish its method and invite you to check it. If a figure on this site is wrong, it should be wrong in a way you can demonstrate.

Where our figures come from

Four sources, in descending order of how strong the evidence is. Every number on this site traces to one of them, and we say which.

Primary documents. The licence number checked against the issuing register rather than taken from the casino's footer. The operating company and jurisdiction of incorporation from the same. Terms of service read in full — which is dull work and is where most of the genuinely useful findings come from, because the geographic restrictions, the circumstances under which verification can be demanded, and what happens to a balance in a dispute are all in there and none of them appear in marketing material.

The product itself. Base RTP read from the info panel inside individual games at this operator, which is authoritative for that title here and frequently differs from a studio's headline figure. Paytables read directly rather than reproduced from another site.

Our own withdrawals. Made on an ordinary account with no arrangement with the operator, on more than one chain, and confirmed end to end. Withdrawals on Bitcoin and on USDT via Tron arrived within one to five minutes, with approval instant rather than queued and no document requested at any size.

Arithmetic. Effective RTP is base return plus the rakeback rate for that category, and the working is shown rather than asserted on the RTP table. Payout curves on the zero-edge Originals are derived from the stated return, which is why every row of the Dice table multiplies to exactly one.

Where a figure is secondary, we say so

Some things on this site come from published sources rather than from us. Competitor terms on the comparison pages are read from those operators' own published material and dated. The restricted country list is compiled from several independent accounts that do not fully agree, which is stated on the restricted countries page along with the specific disagreements.

Where sources conflict we publish the conflict rather than picking the tidier version. The clearest case is on the Rollbit comparison, where independent reviews quote that operator's rakeback anywhere from 5% to 75% — we report the spread rather than choosing a number.

Provably fair verification

The verification method we describe on the provably fair page is one you can run yourself, and the walkthrough is written so you can.

The critical instruction there is not to use the verifier the operator provides. A tool supplied by the party being verified proves only that the tool agrees with itself. The whole point of a provably fair system is that an independent party can reproduce a result with nothing but the published seeds and a standard hash function.

Scoring

Eight categories, each scored independently, and the headline figure follows from them rather than the other way round. We do not decide a site is a 4.9 and then reverse-engineer the components.

Categories are weighted by what actually affects a player's outcome. Rakeback value, withdrawal speed and provably fair mechanics carry more weight than the size of the games library, because a library of five thousand titles at a poor edge is worse than two thousand at a good one.

Licensing is scored honestly and it drags the total down. Duel rates 3.9 there against 5.0 on rakeback, because an offshore permit gives you very little recourse. A site that scored licensing generously to protect the headline number would be worthless.

What changed since publication

Pages are corrected when they are wrong, and corrections are made openly. If a figure changes because the operator changed its terms, we update and the date at the top moves. If it changes because we were wrong, we say what was wrong. The alternative — editing quietly and moving on — works for the writer and not for the reader.

This applies especially to competitor terms. The numbers on the comparison pages come from those operators' published material and are dated. VIP programmes change, offers change, and a comparison table correct in August can be wrong by November.

What we do not test

The list matters as much as the method.

Long-run RTP. Not testable in any practical timeframe — establishing a return-to-player figure empirically takes millions of spins, and anyone claiming to have verified a slot's RTP through a test session is describing variance. We publish configured figures and the arithmetic.

Support under a real dispute. We have not had a balance frozen or a payout refused, so we cannot tell you how the operator behaves when something is genuinely at stake — which is precisely when it matters most. The complaints page collects what other players report, clearly marked as secondhand.

Long-term account behaviour. Whether verification is demanded after a year, or after a large win, is not something a testing window answers.

Game fairness beyond the Originals. Third-party slots are certified by testing houses, not by us. We can verify the Originals ourselves because the seeds are published; we cannot do the same for a Pragmatic Play title.

How to check us

What follows is the most important part of this page, because it means you do not have to take our word for anything.

A methodology page is only useful if it makes the site falsifiable. Four things on this site can be checked by anyone with an account and half an hour.

The RTP figures. Open any slot listed on our RTP table and read the info panel inside the game. If it disagrees with us, the info panel is right. Tell us and we will correct it.

The rakeback rate. Place a bet of a known size on a game whose edge we publish, and check the credit that lands. This is the single most important claim on the site and it either reproduces or it does not — the arithmetic is on the rakeback page.

A provably fair round. Follow the walkthrough on the provably fair page using an independent hash tool. This takes about two minutes and is the single most convincing thing you can do for yourself.

The licence. The number is published on this site and on the operator's footer. Check it against the issuing register rather than taking either of us at our word.

If all four hold, the rest of the site is probably built the same way. If any fails, you have grounds to distrust everything here, and that is the correct response.

Why the score is not a five

Worth addressing directly, because a site that earns commission and publishes a perfect score has told you something about itself.

Duel scores 4.9. The missing tenth is almost entirely licensing, which we rate 3.9. An Anjouan Gaming Authority permit is a real licence and it is not a Malta or UK one, and the practical difference appears at exactly the moment you would need it — a frozen balance with no meaningful regulator to escalate to.

Nothing in our testing suggested that scenario is likely here. Every withdrawal was paid, no document was demanded, and the in-house games verify correctly. But the probability of a bad outcome is not zero and the consequence if it happens is total, and a scoring system that ignores tail risk to protect a headline number is not a scoring system.

The same reasoning produces the other deductions. The sportsbook is thin outside four categories. Support slows overnight. There are no progressive jackpots. Each of those is real and each costs the operator something in our figures.

Conflicts and corrections

This site earns affiliate commission when a reader opens an account through a link here. That is disclosed in the footer of every page and explained on the about page.

No operator pays for placement, sees a page before publication, or has any input into a score. The commercial relationship is real and the appropriate response to it is not to deny it but to publish the unfavourable material anyway — the licensing score, the crypto-only withdrawal limitation, the missing jackpot networks and the thin sportsbook are all on this site because leaving them out would make the rest untrustworthy.

If a figure here is wrong, the contact page reaches us. Corrections are made openly rather than quietly edited, and a page that has been corrected says so.