Two scores kept apart, and the one average we do publish

Almost every casino review site publishes a single number. It is the obvious design, and it destroys the most useful information the site holds.

A single score is a weighted average of things that are not commensurable: what the paperwork says, and what happened to people. Averaging them means an operator with an excellent licence and terrible support comes out looking identical to one that is mediocre at both. The disagreement is the signal, and averaging is the operation that deletes it.

So we publish two, side by side, and never let one stand in for the other. There is one average on a review page, and it is a summary rather than a substitute — the last section explains why.

The TrustAudit Index — what we can check

The editorial score, built only from evidence with a source and a date behind it.

The licence and the entry on the regulator's own register. Payment research, with a distinction between retained estimates and any dated personal test. The terms as written, read line by line, with anything unfair catalogued as a numbered clause carrying a severity and a points value. The complaint record, and how the operator behaved inside it.

It opens at 10.0 and subtracts findings, and every deduction is published on the review with its points and its reason. The weights and caps are on the methodology page rather than described in adjectives.

One recent change to how it behaves is worth stating, because it made scores go down. A score that subtracts findings treats an unexamined file exactly like a clean one — an operator nobody has audited subtracts nothing and arrives at a perfect ten on the strength of everything we never checked. So the audit now carries a checklist, every unfinished check costs the operator points, and a score with open checks is published as provisional with the outstanding items named on the review. A provisional number may rise or fall as research gaps and unresolved findings are reviewed.

The JuryScore — what players report

Testimony from people who have actually had money with the operator. Noisier than the audit by nature, and that is the point.

It is the only place you find out that verification takes four days rather than the advertised one, that support stops answering at the weekend, or that a withdrawal that is supposed to be same-day is same-day only on the third attempt. No regulator collects any of that and no operator publishes it.

It has known biases and we would rather name them than have you infer them. Testimony skews to the extremes, because people write when something goes very badly or very well and almost never when a withdrawal simply arrives. An operator with few testimonies shows an early verdict rather than a confident number.

What neither of them could see

Both scores above have the same blind spot, and it took a year of building them to notice it.

The TrustAudit reads a licence and a set of terms. Both are documents the operator writes about itself, and a company can hold an immaculate licence and publish unobjectionable terms while its players spend a fortnight being asked for a fifth proof of address. The JuryScore was supposed to catch exactly that — and it only works once enough players have written. On a young register, it is silent precisely when a reader most needs it.

So there is now a third source, and it is the one that changed the numbers most. We read the published complaint databases and review platforms, and record a pattern only where at least two independent platforms report the same thing, graded by whether what is reported is service friction, a payment that is late, or a balance that is kept. Those are other people's accounts, sourced and linked on the page, and never restated as findings of fact.

It is normalised by operator size, because that turned out to matter more than anything else in the design. Documented complaints scale with customer numbers: an operator with twenty million accounts generates enough of any pattern for three platforms to report it whatever its conduct, while one with two hundred thousand does so only when the conduct is genuinely bad. Counting them unadjusted ranked the largest operators worst and called it evidence — it put bet365, on any reading the most established licensed operator on our register, at the bottom of the table.

The one average we do publish

Each review page shows an Overall: the mean of the readings we hold on that operator — the TrustAudit Index, an Experience mark for how good the product actually is, and the JuryScore. It appears only where at least two of the three exist, and it is printed underneath all three rather than in their place.

That is not the operation this piece opened by refusing. Refusing to blend meant refusing to let one reading stand in for another, and that still holds: the three are measured separately, shown separately, and the Overall summarises them rather than replacing them. If it is the only number you read, you have read the least informative thing on the page.

What we do not average is our audit with anybody else's. We read the published complaint databases and review platforms closely, and every report behind a finding is linked under it, so you can read the original and disagree with us. But their scores are theirs. Printing another reviewer's rating in the place a reader came for ours hands them the comparison and the credit at once, and this site is being built to do that job rather than to relay it.

When the two disagree

When they agree, you can move on. The interesting operators are the ones where they do not.

High audit, poor testimony is the most informative combination on a review page. The paperwork is in order — the licence is real, the terms are not predatory, the published payout time is short — and something is nevertheless going wrong in support or in payouts that the documents cannot show. That gap is usually worth more to a reader than either number alone.

Low audit, good testimony is the mirror, and it is the one people misread. An operator on a light-touch licence may pay perfectly well and treat players decently. The audit is not predicting that it will not. It is telling you what your recourse looks like if it stops — which is a different question, and the one that matters on the day it matters.

Why neither can be bought

The separation is worth nothing without a firewall, so the firewall is structural rather than a promise.

Whether an operator pays us, what the arrangement is worth and whether they are behind on it are stored where the scoring code cannot read them, and the recompute that produces every published score never touches those fields. The TrustAudit Index is computed from its recorded findings. Separate Experience marks are editorial judgments; the JuryScore is computed from eligible player testimony. These are different inputs, rather than a claim that every rating originates automatically. Paid placements are labelled, sit outside the ranked order, and are never mixed into a score-ranked list.

The same rule now covers the figures describing our own coverage. Numbers like how many operators we have audited are counted from the database when a page renders rather than typed into a form, because a statistic about ourselves deserves the treatment we already give a score.

What our coverage actually amounts to, including the figures that do not flatter us, is on the transparency page.