THE MMA KING

Every fight is evaluated the same way: we weigh each fighter's track record and style against this specific opponent, then translate that into a probability for the win, the method, and how long it lasts. The percentages are calibrated — when we say 65%, we mean it lands about 65% of the time over the long run.

We treat thin résumés with caution. A fighter with only a fight or two of meaningful data gets flagged as limited data and is shown for interest only — never presented as a confident pick.

What you won't find here is hype. The numbers are the numbers, and we'd rather show a coin flip honestly than dress it up.

Win %

Our estimated chance a fighter wins the bout. The two sides add up to 100%.

Confidence feel

On Simple Mode, a plain-language label — from a coin flip to a heavy favorite — for how lopsided we think a fight is.

Method read

Whether a fight is more likely to end inside the distance (knockout or submission) or go to the judges' scorecards.

Why there is no edge or “VALUE” number

There used to be one on every bout and every prop. It is gone, because this site’s own graded files say it was not real. Counting props: the projection-graded ledger is 42-51-3 — 45.2% of 93 decided legs, below a coin flip, and far below the 54.34% a −119 leg or the 57.74% a two-pick entry has to clear. Significant strikes, the market this site called “our most trustworthy”, hit exactly 50.0% on 30 legs. Our projections lose to the posted line: mean absolute error against the realized total is 19.58 for us and 14.99 for the number the book put up (significant strikes 35.4 vs 26.5; knockdowns 1.14 vs 0.50). Moneylines: across 30 settled picks the vig-free market consensus scores a Brier of 0.1996 against our 0.2135, and our number lands closer to the actual result on only 7 of 25 comparable bouts. Derived prop markets: finish-round hits 75.2% of 137 but with a Brier of 0.6133 — a base-rate artifact, not skill; method of victory hits 33.3% of 63. An “edge” is a claim that our number beats the price. Ours does not, so we do not print one.

Power Rating

A single strength number for each fighter, updated after every bout. Higher is better; the gap between two ratings maps to a win probability.

Prop projection

Our expected figure for a stat — strikes landed, takedowns, and so on — that the over/under lines are measured against.

Closing-line value

Whether a price we logged beat where the market settled. It is the cleanest test that exists — and we cannot run it here: closing odds were never retained for MMA, so the tracker can only compare our probability with the price at the moment we logged it, and it says so rather than calling that CLV.

Fight markets are sharp, and ours has not beaten them. So we don't promise winners — we publish a transparent, forward record instead, and when that record came in against us we removed every edge figure, EV number, stake and “VALUE” tag from this site rather than shrink them. What is left is the projection, the model's probability, the posted line and the posted price. The tracker logs each pick at the price we first saw and grades it against the official result, win or lose.

Use the picks as a research signal and a starting point — not a guarantee. The judgment, the bankroll and the final call are always yours.

How the win model actually performs on a time-ordered out-of-sample test — trained only on fights before the cutoff, then graded on every bout after it. No peeking, no fitting on the test set. These are the receipts, not a sales pitch.

See the full reliability curve, bucket by bucket →