There are more UFC cards than anyone has time to watch live, and most reviews worth reading have already spoiled the result. This is our attempt at solving both at once: one spoiler-safe number per fight, so you can decide what's worth your time before you know how it went.
No spoilers
Nothing here reveals a result before you ask. Not the method, not the round, not the winner. A high score never implies an outcome — a first-minute knockout and a five-round war can land in the same place.
Data driven
Every fight is scored by the same formula, from the same public record, with no manual overrides. And the formula isn't guesswork — it's tested against how thousands of fans rated these same fights, and it changes only when a change provably helps.
What the score rewards
A finish counts for more than a decision. It's the biggest single input.
Sustained output and relentless pace both register — a short frantic scrap and a long war land the same way. Knockdowns count extra — in our data they do more for a fight's rating than almost anything else we measure. Long fights that coast get docked.
Hunting for a submission counts too, and it's how often, not how many: three attempts in four minutes reads as a scramble, three across twenty-five minutes doesn't. It's a small term, and it exists because a grappling match with almost no strikes thrown used to score as though nothing happened in it.
A belt, a ranked opponent, a rematch. These say why a fight mattered, and they count a little, on purpose. In our data, big billing predicts how much attention a fight got more than how good it was.
Fight of the Night goes to the bout, not a fighter, so we can weight it heavily and even show it without spoiling anything. It's the most reliable signal we track. Performance of the Night counts too, but naming it would give the result away, so it stays hidden.
The methodology, for those who crave it
The complete version: what we measure, what we can't measure yet, and how the weights are set.
Spoiler-safety is a hard constraint. The formula uses result facts internally, but nothing it shows you can be worked backwards into a winner, a method, or a round.
One formula, every fight, no exceptions. Scores are recomputed from the raw record on every read — there are no stored numbers to go stale, and no per-fight adjustments.
Changes must earn their way in. Any proposed change is fitted on one era of fights and tested on a later era it has never seen. If it doesn't beat the current formula on those unseen fights, it doesn't ship. Most proposed changes fail this test.
The method and round of the finish. Combined striking output and its pace, from the official stat sheet. Knockdowns. Submission attempts per minute. Fight duration. Titles, billing, and rankings coming in. The promotion's post-fight awards. Injury stoppages, which count against the score.
Each is verified against the official record before it enters the score. Where our data disagrees with the official record, the score waits until the data is fixed.
Momentum swings. Comebacks. Upsets. The story two fighters carry into the cage. We've measured this: a model free to use everything we track still misses some fights that critics and fans call all-time great, because they're great for reasons the stat sheet doesn't record.
Grappling is now partly covered and partly not. Submission attempts count. But control, positional dominance and the slow work of passing guard don't — we tested control time and it turned out to measure quiet fights rather than good ones. The clearest cases we still get wrong are grappling matches with no submission attempts at all, where nothing in the stat sheet distinguishes a masterclass from a stalemate.
Until these can be measured properly, we leave them out.
Around a dozen weighted inputs are summed and normalised onto a 0–9.9 scale. The weights are calibrated against how thousands of fans rated more than 2,500 of these fights, and validated on fights the formula never saw during tuning.
Two things are kept even though they don't improve the fan-rating fit: a small premium for stakes, and a rule that a handful of consensus-great fights must stay near the top through any change.