Ten checks, fixed weights, one denominator

Each check is a yes or no question about the operator record. Where the answer is yes the check contributes its weight; where it is no it contributes nothing. The weights sum to ten, which is also the fixed maximum, so a score reads directly as a proportion of what could have been published.

The two heaviest checks are a live licence status and verification wording, at one and a half points each: both mark the difference between an operator that has committed something to writing and one that has not. The lightest is the founding year, at half a point.

The obvious way to build a score of this kind is to divide by the number of fields you happen to hold for each brand. It is also the way that produces a nonsense ranking, and this code was written to avoid it after the mistake was made on a sister site: two facts, both good, becomes a perfect score on a denominator of two, while nine facts with one gap becomes a worse one. The table then sorts by how little we know rather than by how much the operator publishes.

So the denominator here is fixed at ten for every brand, and an unread field scores zero rather than being excluded from the calculation. A brand we know nothing about scores near the bottom, which is the correct place for it: an operator that has published nothing checkable has given a reader nothing to check.

The second rule the code enforces is that an empty array is not a finding. An operator with no restricted list in the record has not been shown to restrict nobody; the list simply has not been read.

This matters most for the one derived comparison the code produces, which sets each brand's own territory list against the subject's. That comparison is only computed where the brand's list is non-empty; where the list is empty the result is empty too, and the interface must not render that as a favourable finding. The output itself belongs on the licence page, and that is the only page on this site where it appears.

Curacao is dropped from that comparison for a separate reason. It is the subject's own licensing jurisdiction, and practically every operator licensed there excludes it as a matter of course. Crediting a brand for covering Curacao would reward an accident of incorporation rather than a real difference between two operators.

The matching underneath is fussier than it looks. Restricted lists are stored as two-letter codes while the subject's list holds full country names, so both forms are matched along with three-letter codes and common variants. The three BES islands share the code BQ, so a brand listing BQ is treated as naming Bonaire, Saba and Statia alike. Every value is stripped of zero-width characters first, because a captured terms page once hid a country name from a plain equality check with an invisible byte inside the word.

The status field is tested twice: it contributes one and a half points when a status is present and does not read as revoked, withdrawn or cancelled, and separately a record whose status does read that way is flagged and sorts below every unflagged record regardless of what else it holds.

No operator in this comparison currently carries that flag. Thirty-three of the forty-one alternatives have a status recorded as active or valid, eight have no status recorded at all, and none has one that reads as revoked. The rule is in the code because a licence can be withdrawn between one reading and the next, and a brand that loses one should fall immediately rather than gradually.

After the flag, ties break on score and then on brand name, alphabetically. There is no hidden tiebreaker and no manual override.

One check deserves its own note, because the register behind it was not usable on the day the records were read. On 2026-08-28 the Curacao Gaming Authority's certificate service returned no result for every domain queried against it, including brands whose Curacao licences are not in doubt.

The seventeen brands whose records carry a certificate identifier rather than a domain search are unaffected: those identifiers resolve. Roobet carries no certificate identifier, so its number, OGL/2024/687/0427, is unresolved.

Unresolved means unchecked, not invalid. The score treats it as an absence rather than a negative finding: the licence status check does not fire, so Roobet earns nothing for it, exactly as any brand with an unread status would.

On the scale described above, Roobet scores 6.0. It names an authority and a number, has verification wording, and has published coin, studio and restricted lists. It has no register link recorded, no status, no verification figure and no founding year.

Across the forty-one alternatives the range runs from 3.3 to 9.0. Wild.io is highest at 9.0, missing only the verification figure that thirty-six brands in the set also lack. Bitcasino.io, Empire.io and Sportsbet.io follow at 8.8. The bottom of the table is occupied by the two brands with no licence record at all, BetPlays at 3.3 and Nitrobetting at 3.5.

Thirty-one of the forty-one score above Roobet's 6.0 and three tie with it, which is a statement about how much each brand has published and not about which is the better place to play.

The commercial relationship is not an input. The scoring routine takes an operator record and returns a score, and the record has no field for a commercial arrangement to be read from. Where a brand is a paid placement it is labelled as one, separately and visibly, so that ordering and money stay apart.

Bonus terms and wagering requirements are not scored either: they change weekly, and a ranking built on them would be stale before it was indexed. Nor is anything scored that would require having played. No deposit was made and no withdrawal timed anywhere in the making of this site, so there is no payout-speed, support-quality or fairness check. Eleven brands carry an instant-withdrawal tag in the library and none is backed by a processing-time clause we have read, which is why the tag earns nothing.