How Rixer scores casinos
Every number on this site is computed from stored, versioned inputs. This page is the whole model — the factors, their weights, what each one measures, where the data comes from, and what it can actually do to a score. Commercial relationships carry a weight of exactly zero, and there is no code path from partner status into scoring.
Current model — v3
Weights sum to 100. Weight is not the same as impact: a factor can cap the entire score regardless of its arithmetic share, so both are shown.
| Factor | Weight | Impact | Available now? | What it measures | Where it comes from |
|---|---|---|---|---|---|
| Safety & licensing | 30 |
Can cap the score
Can cap the entire score. A revoked licence or an unverifiable one limits the total no matter how strong everything else is.
|
Measurable now
Licence and company are on the operator's own site.
|
Which regulator licenses the brand, whether the licence number is published and verifiable, and the operating company behind it. | Operator's own footer and terms, regulator registers |
| Responsible gambling | 15 |
Standard weighting
Tools must be present and reachable, not merely mentioned in a footer. Absence is scored as absence.
|
Measurable now
The tools are present on the site or they are not.
|
Deposit, loss and session limits, self-exclusion, cool-off, reality checks and links to national help services. | Operator's own RG pages, observed in-account tools |
| Review reputation | 12 |
Standard weighting
Weighted by volume and recency. A handful of five-star reviews moves nothing; a sustained low score with no replies does.
|
Partly measurable
Only for brands that already have review pages.
|
Independent review-platform standing: score, volume, recency and whether the operator replies to complaints. | Trustpilot today; Sitejabber and others to follow |
| Transparency & disclosure | 11 |
Strong day-one signal
Cheap for an honest operator and awkward for a dishonest one, so it is one of the few meaningful signals available on day one.
|
Measurable now
Either they publish who they are, or they do not.
|
Whether the operator publishes who they are: company name, registration number, address, licence number and contactable support. | Operator terms and about pages |
| Product quality | 8 |
Light touch
A good product is not the same as a safe one, so this stays small on purpose.
|
Partly measurable
Suppliers and support are visible; stability is not.
|
Game range and named suppliers, platform stability, mobile experience and support channels that actually answer. | Operator site, observed provider names |
| Social reputation | 8 |
Light touch
Deliberately light. Social volume is easy to buy and easy to brigade; it informs, it does not decide.
|
Partly measurable
Most brands have little or no public discussion.
|
What players say unprompted in public communities, and the balance of complaint threads to ordinary discussion. | Reddit, YouTube, X |
| Company reputation | 8 |
Can escalate to a cap
Usually small, but a confirmed regulatory action escalates into a safety cap. Searching the company is how you find enforcement that never mentions the casino's brand name.
|
Partly measurable
Depends on the operating company being named.
|
News about the OPERATING COMPANY, not just the brand: regulatory action, fines, licence suspensions, ownership changes, insolvency. | Google News on brand and company name |
| Longevity & footprint | 8 |
Cuts both ways
Works in BOTH directions. Long steady presence is a mild positive; a profile dominated by spam and PBN links is a negative. Buying links can only ever lower this score.
|
Measurable now
Domain history and link profile are measurable now.
|
How long the brand has demonstrably operated, and the health of its link profile — sustained presence versus a sudden spike. | DataForSEO link profile, domain history |
What we are not scoring yet
Three factors, worth 35 of the original 100 points, are not in the model above. Each needs evidence about real players that Rixer does not yet hold, and every one of them is something we could have guessed at from an operator's own marketing.
We would rather publish a score that says what it is missing than one that quietly fills the gap. A casino's own page about how fast it pays is an advertisement, not a measurement, and scoring it would make the number worse than useless — it would make it wrong in the operator's favour.
| Factor | Weight when live | What it measures | What it needs first |
|---|---|---|---|
| Complaint history not yet measured | 15 | Volume, outcome and response time of disputes, relative to the brand's size. Unresolved and ignored cases weigh heaviest. | Needs real cases. Shown as not yet measured, never estimated. |
| Withdrawals & payments not yet measured | 15 | Stated and observed payout times, withdrawal limits, fees, verification friction and reverse-withdrawal design. | Stated terms now; observed payout times need real players. |
| Bonus fairness not yet measured | 5 | Wagering multipliers, max-bet and max-cashout caps, expiry windows and game contribution — scored by its own model. | Bonus terms are published and can be read today. |
The remaining 100 points are the eight factors above, renormalised in the same proportions they held in the full model — no weight was re-decided to suit the data we happen to have.
Not all licences are equal
A licence is only worth what the regulator will actually do for a player who is not getting paid. We score the recourse a licence provides, not its existence. These tiers describe the regulator, never the operator's conduct — a strict licence is not a promise of good behaviour, and a weak one is not proof of bad.
Binding dispute resolution, segregated player funds, published enforcement actions and a regulator that can suspend the licence.
For example UKGC, Malta (MGA), Denmark, Sweden, Ontario (AGCO/iGO), Netherlands (KSA), Germany (GGL), Spain (DGOJ), Italy (ADM)
Real supervision and a complaints route, but narrower enforcement reach and less public reporting than tier 1.
For example Gibraltar, Isle of Man, Alderney, Estonia, Romania, Greece
A named regulator exists and licences can be checked, but dispute adjudication is limited and enforcement is rarely visible.
For example Curaçao (CGA, post-2024 LOK regime), Kahnawake
Registration rather than supervision. Little or no player-dispute mechanism, so a player's practical recourse is close to none.
For example Anjouan, Tobique, Costa Rica registration, self-declared seals
We could not establish any regulator. Scored as no recourse, not as an accusation.
For example No licence published, or the stated licence cannot be verified
What we deliberately do not score
Omissions should be as auditable as inclusions, so here is what we leave out and why.
-
excluded
Commercial relationship
Whether a brand pays us has a weight of exactly zero and no code path into scoring. This is the one rule the whole platform rests on.
-
excluded
Affiliate and partner volume
A large affiliate footprint measures marketing spend, not player safety, and rewarding it would make the score purchasable. Our own data shows the opposite correlation: link volume across this cohort is dominated by PBN spam, and the brand with the most backlinks is a closed casino.
-
excluded
Raw domain authority
Authority mostly measures budget. Only the LONGEVITY and spam-health of a link profile feed the score, never the headline authority number.
-
excluded
Server location
Almost every casino sits behind a CDN, so a lookup returns the CDN's edge rather than the operator. It is recorded as information where we can establish it, but it is too unreliable to score.
-
excluded
Bonus size
A bigger bonus is a marketing decision, not a trust signal. Only the fairness of its terms is scored.
Rules that override the arithmetic
- capSafety caps beat good averages. A severe finding — a revoked licence, proven non-payment, confirmed regulatory action — caps the total at 30/100 however strong the rest looks. Capped scores say so on the page.
- honestyInsufficient evidence beats fake certainty. If less than half the factor weight has verified inputs, we publish “insufficient evidence” instead of a number. An unscored casino is not a criticism of it.
- confidenceConfidence is shown next to the score. It reflects how much of the model actually has real inputs, so a thin score cannot masquerade as a thorough one.
- separatePlayer ratings are never blended in. The community score sits apart from the editorial score, and small samples are shrunk toward the average so a handful of reviews cannot manufacture a consensus.
- auditableEvery input and every change is logged. Each casino page shows its score change history; model versions are permanent records. A score you cannot reconstruct is an opinion, not a measurement.
- independentMoney cannot buy points. Partner status is disclosed on every page and carries zero weight.
Scoring a casino we have only just started tracking
Complaints, withdrawal tests and player reviews are the strongest evidence we have — and a brand we have just added has none of them. Rather than guess, the model leans on the signals that exist from day one: who the operator says they are, which regulator stands behind them, how long they have demonstrably operated, and what independent platforms and public communities already report.
Those are weaker evidence and the model treats them that way: their combined weight is deliberately a minority of the total, and a score built mostly from them will show low confidence. As real complaint and payout data arrives, it displaces them. A new casino is not scored as a suspicious one — it is scored as one we do not know well yet, and the page says which.
Bonus Fairness Score
Bonuses carry their own 0–100 fairness score, separate from the casino score. It starts at 100 and loses points for high wagering multipliers, wagering applied to deposit plus bonus, caps on withdrawable winnings, maximum-bet restrictions, short completion windows and reduced game contribution. Unknown terms lose points too — an offer never gets the benefit of the doubt.