How the ordering works, check by check

Every check gets answered yes or no from the operator record, and every one has a weight that never varies. Added together the weights come to exactly ten, and ten is also what each score is divided by — the first of the two decisions explained further down.

Licence status and verification wording carry the most, 1.5 apiece. They were weighted up for one shared reason: those are the two fields where an operator saying nothing costs a player the most in practice.

A score here is always out of ten, whether the operator publishes ten fields or two. That sounds obvious and it is not: the natural implementation divides by the number of fields that were actually filled in, and that version quietly rewards silence. A brand publishing one fact and getting it right would score full marks.

Fixing the denominator makes the score mean what the page says it means. Two of ten is not a bad operator; it is an operator about which eight of the ten checks could not be answered, and it reads as a low number precisely because that is the finding.

The same principle governs empty fields everywhere on this site. An empty array in the record says the document has not been read yet; it is never a claim that the answer is zero, and a scoring routine that confuses the two will hand out marks for pages nobody has opened.

One of the ten checks asks whether an operator published a restricted-territory list. It does not ask what is on it. That is a disclosure signal in exactly the same way as publishing a coin list, and the contents are never an input to any position in any table.

The reason is a boundary this site keeps deliberately. It prints one restricted list in full — Rollbit's own, on the licence page — because that list is a clause of Rollbit's Terms and Conditions of Service, the one document this whole comparison is anchored to. That reasoning stops at Rollbit. It is not extended to the forty-two alternatives, and a scoring field that read their lists would be precisely how it slipped back in.

Twenty-six of the forty-two publish a list of some kind. That count is all the score uses.

Payout speed, support responsiveness, dispute outcomes, bonus fairness in practice, and whether an operator honours the clauses it publishes. None of these is an input, because establishing any of them would require money on the table and none was staked.

Reputation and size are also invisible to it. The scoring has no popularity term, no traffic term and no age bonus beyond the half-point for publishing a founding year at all, so a 2013 brand and a 2024 brand are treated identically if both say so.

And commercial relationships are invisible to it by construction. There is no field in the operator record that carries one, so there is nothing for money to attach to. The single paid placement on this site sits outside the ranked table and is labelled where it appears.

A score moves when the underlying record moves, which happens when a document is re-read and found to say something different, or when a field that was blank is filled in from a source we can point at. The new position then shows up first on the ranked table of Rollbit alternatives.

It does not move because a brand asked. An operator that wants a higher position has one route, and it is a real one: publish more. Most low rows here are low because a coin list, a studio roster, a verification threshold or a licence status could not be found on the operator's own site.

Each page is stamped with the day its facts were read. Where those dates differ across a table, the older ones are the ones to treat with more caution, and this site would rather show you a date than imply permanent currency.