Methodology & honesty

What this is

Fans use the phrase MMA math as a joke: A beat B, B beat C, so A beats C. Everyone knows it does not quite work. We wanted to know how far it goes when you stop joking about it: every result, every opponent's opponent, weighed at once, with honest uncertainty about how much any single fight can tell you. Plus some secret sauce of our own.

What MMA math cannot see

Results are not the whole story, and every fight fan knows it. A fighter can walk to the cage with a hidden injury, a brutal weight cut, a broken camp, or nerves that no record book will ever show. MMA math sees none of that. It can only tell you what the results themselves imply, which is exactly what we ask of it.

Why no women's divisions

Women's divisions are not rated here yet. The talent pools are smaller and the records are shorter, so the same MMA math gives shakier answers there. We would rather show no number than a shaky one, so women's fighters appear with their careers and results but without a rating.

Why debuts get no call

When either fighter is making their first UFC appearance, we make no call and grade nothing. Our numbers come from results, and a debutant has no UFC results to weigh yet; whatever we might compute would be closer to a guess than to the standard the rest of the record is held to. Debut fights are listed like any other, but they carry no probability and stay out of our record in both directions: no credit for the ones we would have gotten right, no cover for the ones we would have gotten wrong.

What we claim

We think our MMA math holds up better than the joke suggests, and we would rather show you than tell you. The full record is published, including the fights it called wrong, so you can judge it for yourself.

What we do not claim

We do not claim to beat the bookmakers. A closing line at a sharp bookmaker carries the wisdom of a very sharp crowd: professional bettors, people who hear about the injury or the bad weight cut before you do, and everyone in between. We built this to study the sport, and that is how it should be read.

On accuracy claims

Be suspicious of anyone claiming to pick fights with more than seventy percent accuracy. Numbers like that are not hard to produce when you build models: tune long enough against your held-out test fights and the score climbs. Such a person will tell you the holdout was never used for training, which can be technically true, but using it to make sure the accuracy stays high is overfitting all the same, and a model built that way scores far lower on fights it has truly never seen. Accuracy is also not the metric that matters most. Favorites win most fights, so a model that leans toward favorites scores well by default, and a model can be profitable at a much lower accuracy if it disagrees with the bookmakers often enough and wins enough of those disagreements. That is why we publish log loss, the closing line on the same fights, and our disagreement rate next to every accuracy figure.

Responsible gambling

Nothing on this site is betting advice. If you gamble, gamble responsibly and within your means.

Dataset generated 2026-09-13.