Methodology: the ten parameters, threshold by threshold
The methodology below is the whole of it. Ten parameters, each computed by a printed rule from what an operator publishes, then averaged into one score out of 10. No editor moves a casino up or down, because there is no place in the pipeline where an editor could. If you disagree with a position, the argument you want is with a threshold on this page, and that is a far more useful argument to have.
The guide
What the licence line pays for
Regulators are worth different amounts because recourse is worth different amounts. A UK Gambling Commission permission scores 10 and a Malta Gaming Authority one scores 9. Neither appears in this roster, and the scale keeps them anyway so it never needs rewriting when a brand moves up. Below that sit the licences that do appear: a direct Curacao Gaming Authority B2C licence is worth 7, an Anjouan Gaming Authority licence 5, and a Tobique Gaming Commission licence 4. A regulator the record cannot identify at all scores 3.
One point is then added when the operator publishes a licence number, because a number is the only part of a licence claim you can take to a register and check. The total is capped at 10. That single point is the difference between a claim and a checkable fact, and it is why Winna scores 4 here: the Tobique permission is real and the regulator's register confirms the licensee, but no number is printed for a reader to carry across.
How a cashier becomes a single number
Three cryptocurrencies earn one point. The count is deliberately coarse, because the gap between a site with thirty coins and one with forty is invisible to a player who deposits in one of them. Crypto alone tops out at 9, and a working fiat cashier adds the last point, capped at 10, since card and bank rails reach people that crypto does not.
Rollbit runs the widest cashier in the roster at 25 coins and still stops at 8, because it takes no fiat. Weiss sits at the other end with 3 coins, which is the one case where this parameter reads almost like a warning label.
Four verification modes and their scores
Verification is scored by friction as a player meets it, not by how strict the compliance department is on paper. No documents requested for ordinary play scores 10. Documents only above a withdrawal threshold scores 7. Documents on a risk-based check, meaning the operator reserves the right but does not ask by default, scores 6. Documents before the first withdrawal, no exceptions, scores 4. A policy that is not published clearly enough to place scores 5, which sits deliberately above the strictest published rule and below every relaxed one.
Gamdom is the only entry in the roster that scores the full 10, and the reason is a sentence in its help centre rather than an absence of policy. Silence is not the same as permission, and the scale treats them differently on purpose.
What payout speed is worth on this scale
Withdrawals processed instantly score 10. Inside an hour, 9. Inside a day, 7. Longer than a day, 5. An operator that publishes no processing time anywhere also scores 5, which is the same as being slow, because a figure you cannot plan around is worth what a slow one is worth.
A further point goes to operators with no monthly ceiling, capped at 10. A monthly cap costs nothing to a casual player and everything to someone with a real win to move, and this is the parameter where that difference shows up. Stake carries both halves of it and scores 10.
Bonus value is the size divided by the rollover
A welcome offer is worth its percentage divided by its wagering multiplier, and nothing else. That ratio gets added to a base of 3, rounded, and clamped between 3 and 10. A hundred percent at forty times rollover produces a ratio of 2.5; three hundred percent at the same rollover produces 7.5, and the scores follow. The headline number alone never moves this parameter.
Three special cases carry their own rules. An operator with no welcome bonus by design scores 4 rather than zero, because running rakeback instead of a deposit match is a choice about the business model rather than a failure. An offer whose wagering is not published anywhere scores 6, since an unstated rollover could be anything. A wager-free bonus scores the full 10, because winnings you can withdraw without clearing a rollover need no division at all. The arithmetic behind those ratios is worked through properly in its own guide.
Rakeback decides the loyalty score
A published rakeback percentage is the only loyalty number that means anything before you play, so it drives this parameter directly: half the percentage added to a base of 5, floored at 5 and capped at 10. A ladder of tiers with no published rate scores 6, which says a programme exists without pretending to know what it pays. Nothing published at all scores 4.
One more point goes to operators that will match a VIP status earned somewhere else, capped at 10. Winna and Duel both reach 10 on this line by very different routes, and what those rates cost to reach is a guide of its own.
Support, mobile and what makes a site distinctive
Live chat running around the clock scores 8. Live chat with limited hours scores 6. Neither scores 4. A published email address adds a point and a Telegram channel adds another, capped at 10. A chat window is where an urgent problem gets solved; a written channel is where a slow one leaves a trail.
Mobile is three states rather than a spectrum. A native app from the operator scores 9, an installable web app scores 7, and anything unverified scores 5. The distinctive features parameter starts at 5 and adds: two points for in-house original games, two for provably fair verification, one for the operator's own token, and one for a sportsbook on the same balance, capped at 10. Those four are the things a site cannot buy from a provider catalogue, which is exactly why they are the ones counted.
Reputation counts years rather than review stars
Half a point for every year of operation, added to a base of 4 and capped at 9. Age is a weak signal, and it is used here because it is the only trust signal that cannot be bought, farmed or written by an affiliate. A confirmed incident subtracts two points with a floor of 1, and the incident is named in the reason column rather than hidden inside the number.
Nothing here reads user reviews. Aggregated ratings get gamed in both directions, by operators buying praise and by players writing revenge after a lost bet, and folding that into a computed score would undo what the other nine parameters are for.
Why the same casino moves between lists
The ten scores are computed once. A list then multiplies the parameters its question cares about and averages the result, and that is the only difference between one ranking here and another. A parameter with no multiplier counts once. The verification list triples verification and nearly cancels the bonus; the payout list triples speed. Multipliers are printed at the top of every list, so the weighting is something you can reject rather than guess at.
How the head to head record is counted
Every casino plays every other one, which comes to 190 pairs. Inside a pair, each of the ten parameters awards a point to whichever side is higher, ties award nothing, and the side with more points takes the pair. Equal points make it a draw. The win rate treats a draw as half a win, which is why a site that draws often lands above one with the same number of outright wins.
Trust that record over the overall average, because it survives a change of weighting. A casino that takes the pair against Betfury or Winz beat them on parameters, not on a mean one strong line was carrying.
Where to take these rules next
These rules read better against live examples. The Stake scorecard prints all ten with a reason beside each; the Gamdom scorecard puts a maximum verification score next to a weak licence; the Betfury scorecard puts a top bonus score next to a bottom payout score. The verification-weighted list is where the multipliers do their most visible work.
What people ask about these scores
Can an operator pay to change its score?
No, and the structure is what prevents it rather than a promise. Scores are computed by script from a facts file, the reason for each number is printed beside it, and the affiliate link lives in a separate field that no rule reads.
Why does an unpublished figure score better than a bad one?
It usually does not. An unpublished processing time scores the same as a slow one, and unpublished wagering scores below a generous ratio. The one place silence scores mid-table is verification, where no published policy genuinely does mean less friction in ordinary play than a documented demand for papers.
How often do the numbers change?
Whenever the facts are re-read and the collector runs again. Terms move, licences lapse into assessment, cashiers gain and lose coins, so a score reads a document at a point in time rather than settling anything permanently.
Next step
Rules read? See what they produce
The leaderboard is the running total of every pair scored by these thresholds, and each review shows the rule next to every number it produced.
18+. Casino links here are affiliate links. Scores are computed, never sold. How that works.
Read next