Evidence policy

Every fact on this site carries a label saying how well it is evidenced. The labels are not decoration — they are absolute, and a claim resting on more than its source supports is never published.

The classes

Official — current

Taken from the operator's own current published documentation and re-checked on the date shown.

Taken from the operator's own current published documentation and re-checked on the date shown.

Official — historical

Taken from the operator's own documentation as it read on an earlier date. It may no longer be current.

Taken from the operator's own documentation as it read on an earlier date. It may no longer be current.

Independently verified observation

An observed outcome that we were able to corroborate against supporting evidence.

An observed outcome that we were able to corroborate against supporting evidence.

Moderated user report

A reader-submitted outcome that a moderator reviewed and accepted, with supporting evidence attached.

A reader-submitted outcome that a moderator reviewed and accepted, with supporting evidence attached.

Unverified user report

A reader-submitted outcome that we have not been able to corroborate. Treat it as an anecdote.

A reader-submitted outcome that we have not been able to corroborate. Treat it as an anecdote.

No official rule on record

We have not established this. It is not a value of zero and not an implied answer of any kind.

We have not established this. It is not a value of zero and not an implied answer of any kind.

“Not established” is a real answer

Where we have not established something, the page says so in words. It is never rendered as a zero, a dash, a blank cell or an omission — each of which a reader reasonably interprets as a value, and the last of which reads as an absence of problems.

A profile with very little on it can mean several different things, and they are not interchangeable: the operator may publish rules we read that simply do not state the figure; the document may not have been readable from where we looked; we may not have located it yet; or the brand may no longer publish under that name. Each profile states which of those applies to it rather than leaving a reader to assume the worst one.

This matters most on operators we know least about. Whichever of those causes applies, a profile with almost nothing on it does not mean the operator is hiding something — and the page says which cause applies rather than leaving the impression.

What one report can and cannot do

A single user report is never published as a fact about an operator. It cannot produce a processing-time figure, it cannot open an operational notice on its own, and it is never described in language that generalises it — one delayed cashout is not evidence that an operator delays cashouts.

What a single report can do is prompt us to look. If it points at something checkable against a stronger source, the stronger source is what gets published.

Language rules

These follow directly from the classes above and are applied to every page.

  • We do not call an operator a scam, a fraud, or a thief. Those are conclusions about conduct and intent that an operational tracker has no basis to reach.
  • We do not say “never pays” or “always delays”. We say what the recorded observations in this dataset showed, with the sample size attached.
  • We do not say “everyone is affected”. We describe the scope we can actually evidence.
  • We do say “evidence is currently insufficient to determine” when that is the honest position, including when it is unsatisfying.
  • We do not draw legal conclusions about any operator, any term, or any reader’s situation.

Evidence and commercial relationships

An evidence class is never influenced by whether a relationship exists with an operator. Whether one exists is not available to anything that produces a figure or a label, so a commercial arrangement cannot alter either, even accidentally. See the affiliate disclosure.

Challenging a label

If you believe a fact is mislabelled — including an operator who believes we have understated what their own documentation says — the corrections process exists for exactly that, and a citable source will change the record.