How often do UFC judges get it wrong? 3,955 decisions checked against the stats

One decision in five goes to the fighter who landed less

In 21% of UFC decisions the winner landed fewer significant strikes than the loser. That rises to 39% in split decisions, which makes sense: those are the fights the judges themselves couldn’t agree on. And it hasn’t got any rarer: from 2021 to 2026 it was 23%.

But “he landed more” isn’t how fights are scored. The rules tell judges to weigh effective striking and effective grappling, and a takedown or three minutes on top count even if they don’t show up in a strike total. So the real question is what a judge counts, and how much.

What a round is worth to a judge

Judges score each round, and UFCStats records what each fighter did in each round. So I built a model that predicts every judge’s scorecard from the round-by-round numbers: strikes landed, where they landed, knockdowns, takedowns, control time and submission attempts. The model learns how much each one moves a judge. Here it is, priced in the plainest unit there is, a significant strike to the body or legs:

A knockdown is worth about 16 strikes. A submission attempt is worth almost six, a minute of control three and a half, and a takedown two and a half. A strike to the head counts about 50% more than one to the body or legs. Two things count for almost nothing: light, non-significant strikes, and strikes that miss. Throwing a lot and missing doesn’t win rounds. So a fighter who lands ten fewer strikes but spends three minutes on top will usually take the round.

One popular idea doesn’t show up at all: that a big last round “steals” the fight. I let the model give the final round extra weight and it came out at 1.01 times the others. As far as the scorecards show, judges score the last round like any other.

Most rounds aren’t close, either. By the numbers, 60% of rounds are at least 80–20 for one fighter, and only about one in five is near even. Those close rounds are where decisions are won and lost.

So how many are real robberies?

With the model standing in for a typical judge, every decision gets a number: the chance a typical judge, looking at these stats, would score the fight for the man who actually won. When that chance is 25% or less, so the numbers say a typical judge goes the other way at least three times in four, I’ll call it a robbery by the numbers. That’s deliberately strict. A close fight that goes to the fighter with the slightly weaker case is a bad decision, but it isn’t a robbery.

That’s 171 of 3,955 decisions, about 4%. Since 2010 it works out to roughly ten a year, and between 3% and 7% of decisions in every full year, with no trend up or down. Just over half of them were split decisions, so at least one judge saw it the numbers’ way. But 42% were unanimous: all three judges against the stats. When that happens, the stats may well be missing something the judges saw, like how hard the shots landed.

Here are the ten decisions the numbers disagree with most. Strikes landed are significant strikes, winner first. Control time is rounded to the minute.

Some of these are famous: Pokrajac over Maldonado, Sherk over Dunham and Garcia over Phan are regulars on any list of the worst decisions in UFC history. Not every one is an outrage, though. Marlon Vera knocked Rob Font down three times; the numbers say Font’s extra 112 strikes should have outweighed that, and all three judges disagreed. Further down the list is a future champion: in 2017, Merab Dvalishvili lost a split decision to Frankie Saenz despite 11 takedowns and seven minutes of control.

If one fighter has the best claim to being the most robbed in UFC history, it’s Bobby Green. Four of his six decision losses were robberies by the numbers. He out-landed the winner every time (92–58, 50–31, 85–42 and 143–104), and every one of the four was unanimous. Evan Dunham is next, with three.

A thumb on the scale

The last question is whether anything besides the fight moves the judges. To test it, I let the model give one side a little extra credit every round, for being the betting favorite or for being in the red corner, and measured how much credit the scorecards actually call for:

The betting favorite gets about 1.6 strikes’ worth of extra credit per round, and a big favorite almost 3. The red corner gets about 1.4. Each effect holds when the two are measured together. On a round that’s otherwise dead even, that’s a 4- to 5-point lean. Across whole decisions, favorites won 67% where the stats alone predicted 63%, and the red corner won 58% where the stats predicted 55%. Experience and age make no difference.

There are two ways to read the favorite effect. Judges may lean toward the fighter everyone expects to win. Or favorites may do things the counts miss, like landing harder or avoiding the clean shots. The red corner is harder to explain away. The promotion assigns it, usually to the higher-ranked fighter, and nothing about it happens in the cage. The good news is that it’s fading. From 2010 to 2017 the red corner’s extra credit was about 2 strikes per round; since 2018 it’s been about 1.

And the judges themselves? Across all their cards, a judge was the odd one out, disagreeing with the other two, 7% of the time. Among the 19 judges who have scored at least 150 UFC fights, most went against the numbers on 6–9% of their cards. The differences between them are small, and they partly reflect which events each judge works.

Three famous decisions, round by round

Patterns across thousands of fights are one thing, but robbery arguments are about particular fights. So here are three of the most argued-about decisions in UFC history, run through the same model, with my own call on each. In the charts, each bar shows the chance a typical judge gives that round to each fighter, based only on what happened in the round. The champion, in the red corner, is on the left.

Jon Jones vs Dominick Reyes, UFC 247, 2020

Jones was the long-reigning light heavyweight champion and nearly a 4-to-1 favorite. Reyes had the better of the early rounds, and Jones came on late. All three judges gave it to Jones, two by 48–47 and one by 49–46. Dana White said afterward that he had Reyes ahead three rounds to one going into the fifth, and the 49–46 card was criticized by White and by Joe Rogan.

The numbers have Reyes winning the first three rounds and Jones the last two. Reyes landed 116 significant strikes to Jones’s 104, and threw a lot more, 259 to 166. But only the second round was clear-cut. The first and third were close enough that a typical judge gives each of them to Jones about one time in three. Put the five rounds together and a typical judge scores the fight for Jones 44% of the time. So Reyes had the better case, but this is a close fight, well above our robbery line of 25%. The 49–46 card is another matter. A typical judge gives Jones four of the five rounds only about one time in nine.

This is also the kind of fight where the thumb on the scale matters. Jones was the champion and the big favorite. If I add the extra credit that favorites and red corners get, measured together, a typical judge’s chance of scoring it for Jones goes from 44% to 56%. In a fight this close, that lean is enough to decide it.

My call: the wrong result, but not a robbery. I’d have scored it 48–47 for Reyes, so I think the wrong man won. But I’m keeping “robbery” for decisions where the numbers clearly point the other way, and this one doesn’t. Rounds 1 and 3 were close enough that a fair judge could give them to Jones. A judge who did that made a call I disagree with, not an indefensible one. The 49–46 card is the one part of this decision I can’t defend.

Lyoto Machida vs Mauricio “Shogun” Rua, UFC 104, 2009

Machida was the unbeaten light heavyweight champion. Shogun spent five rounds kicking Machida’s legs and body, and all three judges scored it 48–47 for Machida. The UFC booked a rematch straight away, and the following May Shogun knocked Machida out in the first round to take the title.

Here the numbers aren’t close at all. Shogun out-landed Machida in every round, 80 significant strikes to 35 overall, and 49 of his landed on Machida’s legs. A typical judge gives Shogun every round more often than not, and gives the fight to Machida only 7.5% of the time. Machida’s case was that his shots were cleaner and caught the eye, and he did land more to the body, 21 to 16. But the model already gives body strikes full credit, and it still has Shogun well ahead.

My call: a robbery. Shogun won at least three rounds on any reasonable reading of the numbers, and all three judges still found three rounds for the champion. Of the three fights here, it’s the only one where the numbers and the outrage agree.

Alexander Volkanovski vs Max Holloway 2, UFC 251, 2020

This is probably the best-known robbery claim of recent years. Holloway had the better of the first two rounds, Volkanovski came back, and he kept his title on a split decision, 48–47 on two cards and 47–48 on the third. FanDuel refunded bets on Holloway, and Holloway later said 80 to 90 percent of the world thought he had won.

The numbers say Volkanovski clearly won the last three rounds. He out-landed Holloway by at least 10 strikes in each of them and took him down three times in the fourth and fifth. On the stats alone, the first two rounds were dead even, so a typical judge gives Volkanovski the fight 91% of the time.

That isn’t the whole story, though. Live coverage at the time had Holloway dropping Volkanovski with a head kick in the first round and an uppercut in the second. UFCStats didn’t log either one as a knockdown, so the model never saw them. If I count both as knockdowns, Holloway wins the first two rounds clearly. But Volkanovski still gets the decision 76% of the time, because the last three rounds were his.

My call: not a robbery. Holloway won the first two rounds more clearly than the strike count shows, and Volkanovski won the last three. 48–47 Volkanovski is the right score. When they fought a third time in 2022, Volkanovski won every round on all three judges’ cards.

Other famous ones, briefly

Some other decisions that come up in every robbery argument, with the chance a typical judge scores each for the fighter who actually won. Georges St-Pierre over Johny Hendricks in 2013 (55%) and Sean O’Malley over Petr Yan in 2022 (49%) are coin flips, each decided by one round the numbers call nearly even. Jon Jones over Alexander Gustafsson in 2013 (74%) and Amanda Nunes over Valentina Shevchenko in 2017 (67%) lean toward the actual winner. Michael Bisping over Matt Hamill in 2007 (9%), Diego Sanchez over Ross Pearson in 2014 (9%) and Dricus Du Plessis over Sean Strickland in 2024 (13%) are robberies by the numbers.

The verdict

Most “robberies” aren’t. In most of the decisions where the loser landed more, the winner earned it on the mat or with knockdowns, and judges reward those consistently. Real robberies, where the numbers clearly point the other way, come to about ten a year. And close rounds tilt slightly toward the favorite and the red corner. It isn’t enough to swing many fights, but it’s there in two decades of scorecards, and in a fight as close as Jones vs Reyes it can be the difference. Of the three famous cases I went through round by round, only Machida vs Shogun holds up as a robbery.