How the matching works
You answer questions. Some answers rule casinos out entirely; the rest change the order. Then we show up to 3 of them with every reason attached, and one drawback each.
The two kinds of question
Must-haves are exclusions. If you say live dealer is a must-have, a casino without it doesn’t appear at all — not lower down with an apology. There are 3 rules of this kind. They are also the reason the finder can come back with nothing, which it will do rather than relax a constraint you set.
Preferences only move things around. There are 6 of these, each with a published weight. A casino earns all, half or none of a preference’s weight, and its position is the share of the weight it earned out of the weight you put on the table. Half credit exists because “nearly” is a real answer and pretending otherwise makes the ordering worse.
If the best thing we can find earns less than 45% of what you said mattered, we call that no strong fit and say so, with the near-misses underneath.
3 more rules are written down and not running. They read how an operator takes money in, and the records cannot answer that for every casino, so asking you about it would sort them on a gap in our data rather than on anything about them. The withdrawal is listed rule by rule on the methodology page, along with what would bring them back.
Finder answer
Must-have
A requirement. Failing it, or lacking the data to confirm it, rules the casino out.
Preference
Only reorders survivors. Credit is all, half, or none of that preference’s weight.
- All of the weight — the preference is met
- Half of the weight — the preference is partly met
- None of the weight — the preference is not met
Fit measure
Share of answered preference weight earned. Hard exclusions are not part of this denominator.
Explained shortlist
Every reason, plus one drawback.
- Meets this
- Partly meets this
- Doesn't meet this
- No data
No strong fit
Used when the best remaining fit is below the published minimum.
Complete text equivalent
- Each finder answer is either a must-have or a preference.
- A failing or unknown must-have is a hard exclusion. That casino is removed and does not contribute to fit share.
- A passing must-have lets the casino continue. Preferences then award all, half, or none of their weight. Unanswered or unknown preferences are skipped and do not count either way.
- Fit is the share of answered preference weight earned. The shortlist is at most 3 explained results, each carrying every reason and one drawback.
- If the best remaining fit is below 45 percent of what the visitor said mattered, the outcome is no strong fit.
Why it’s only 4 questions
Because a thirteen-question form is a survey, and people abandon surveys. The core sequence is 4 questions, some of which only appear if an earlier answer makes them relevant — nobody is asked about live dealer who didn’t mention table games. Everything else is offered next to the results, where you can see what each change does.
Skipping a question is free. An unanswered preference isn’t counted for or against anybody, so leaving one out doesn’t quietly penalise every casino.
What state is used for
One thing: leaving out casinos that say they don’t accept players from that state. It is optional and it isn’t used to say anything about the law. It is kept in your browser’s local storage so a refresh doesn’t lose it. If you want to know what’s permitted where you live, that’s a legal question and we’re not the right place.
What this can’t do
- It only knows about 12 casinos. A better fit for you may exist that we haven’t profiled, and the finder has no way to tell you that.
- It works from recorded attributes, not from playing there. Where an attribute was read from an operator’s own page the card shows the date it was read; where it was not, the card says so. Availability in particular changes without notice.
- Interface complexity and library breadth are classifications we make against a written rubric. They’re judgements. Reasonable people will disagree with some of them.
- It can’t tell you whether a casino will pay you, and neither can we. Payout speed is not confirmed for a single site in the data, and every result says so rather than filling the gap.
- There is no machine-learning model and no language model in the scoring. It’s arithmetic over published rules, which is why every line of reasoning can be shown to you.
What can’t influence it
Money. The code that scores casinos has no access to what any of them pays us — not a disabled feature, a missing connection. The whole set of recommendations is recomputed automatically with every commission figure altered, and the check fails if a single result moves by a single place. The commercial side, in full.
Every rule and weight is listed here, in the same words the code uses.