Heatmaps, motion tracking and ‘the next frontier’ in mixed martial arts statistical data

Why didn’t Michael Johnson defeat Beniel Dariush at UFC Fight Night 73? The question isn’t meant to evaluate whether we personally agree with the judges awarding Dariush the split decision victory, but rather, why there was controversy in the first place? The judges themselves were not unanimous and neither was the audience watching. If there’s smoke, is there also verifiable fire?

It turns out, we might be able to tap into what drives the belief among some that Johnson deserved the nod if we look at more than just the aggregate striking totals. After all, those numbers are very close and while important, don’t settle the debate in and of themselves.

For most observers, there is general agreement Johnson’s striking was more effective in the first while Dariush the clear winner of the third, leaving the second round the epicenter of the dispute. Johnson outlanded Dariush numerically in that frame, but it was perhaps spatially that tells us the most about what happened in that round with the help of Fightmetric’s new ‘heatmaps’, a new data visualization tool that displays the movement of each fighter over the course of a round in the UFC:

This image doesn’t just show us where the fighters moved, but according to this representation of their movement in that pivotal second round, Johnson controlled the center of the Octagon for 57 percent of the round to just 4 for Dariush. Johnson advanced 29 percent vs. retreating only 13 percent, while Dariush advanced just 14 percent and retreated 27 percent. When one recalls that aggression and Octagon control are criteria judges must use to evaluate a winner, it appears the case of Johnson’s victory gets even stronger (note: the judges do not have heatmap visualization nor striking data to help them evaluate winners).

The key insight here from Rami Genauer, founder and director of Fightmetric, is that how much, where and why fighters move in the Octagon can be just as important to understanding a fight as the raw data about striking or takedown totals. “[This] is just another way of expressing the same kind of statistics,” he tells MMA Fighting. “We’re tracking the location and movement of the fighters and the heatmap is a nice way of plotting that on a graph or plane, so you can display frequency of time spent in a position in an easy way for people to consume.”

Collecting and making use of motion-tracking data has been a priority for Fightmetric for years, but heatmaps are the first use of them in mixed martial arts. Genauer says the first use of heatmaps came at UFC 180, but had been in the works for years. The value, he argues, is not the heatmaps themselves. While they can be a handy visualization tool, understanding just how much motion affects fighters, strategy, fans and especially judges can be a hugely important revelation for the sport is what’s really important.

“First and foremost, the important thing is to collect as much good data as we can,” he notes. “Ultimately, we’re trying to explain the sport and explain the fights. There’s going to be fights for which they strikes may not tell the story, the takedowns may not tell the story. One of the stories could be the use of space and the motion or direction of motion. We’ve seen that quite a bit. Regardless of who is landing the strikes, the fighter who is moving forward is considered winning. Judges certainly value it. I think fans and commentators do as well.

“So, if we can quantify those pieces of information, then we’ve added another layer onto the sport,” he continues. “We can understand it better. We can say there is tremendous value to being the fighter who is moving forward.”

That may not have not helped Michael Johnson in his close fight with Beneil Dariush, but from the motion-tracking data Fightmetric has collected to date, Johnson’s loss despite forward pressure and center control in the second round of their contest is aberrant, not indicative of what appears to be a larger trend.

Consider the following cases and Genauer’s explanation of what they mean as well as how heatmaps can differ fight to fight:

Conor McGregor vs. Dennis Siver, UFC Fight Night 59

“McGregor-Siver is a great example of center control success. What you see is the donut. There’s the center of the Octagon and Siver’s got a big blank spot. You take a look at McGregor’s and he’s filled in the donut hole, so all of his orange and yellow are right there smack in the center, where Siver is exactly not.

“If you watch that fight, it was one of those things where, ‘Why isn’t Siver doing anything about this? Why is he being complacent? Why is he allowing McGregor to let him pot shot?’

“It’s very, very difficult to win a fight from the outside if you’re opponent is constantly making you circle and circle because you can never set. You can never throw anything with power, as opposed to the guy who is stationary who can dart forward at a moment’s notice.

“For the most part, if you’re struck on the outside circling, you’re losing.”

Donald Cerrone vs. John Makdessi, UFC 187

“Neither fighter is on the perimeter. Both of them are standing in the center. They don’t move at all. They’re just in the pocket, taking shots at each other.”

Jon Jones vs. Daniel Cormier, UFC 182

“This is an example of a fight that’s all along the fence or in the clinch, at least. You’ve got vast segment of the Octagon in a five-minute round and yet there’s so much of it which is basically untouched because they did spend so much of it against the fence.

“This is not an interesting thing I’d put on screen to show people, but I would show it to you so can see there are different patterns of motion or action, which will play out differently in heatmaps.”

T.J. Dillashaw vs. Joe Soto, UFC 177

“They didn’t spend any time against the fence. This is all spent in free space. They didn’t cover the entire Octagon. You can see they still concentrated in the north-by-northwest portion, but at the same time, they’re constantly moving. You can see the total distance traveled between is 3,036 feet. That’s so far the highest we’ve seen combined in a round between two fighters.

“They’re covering a lot of ground and using a decent portion of the Octagon to get there.”

Travis Browne vs. Andrei Arlovski, UFC 187

“They use a lot of Octagon. It’s not like the Cerrone-Makdessi fight where they could be doing the same things – Arlovski and Browne are standing up and striking – but they’re doing it in a fundamentally different capacity because they are covering ground, they’re taking on different pieces of the Octagon, they’re moving.”

Cat Zingano vs. Amanda Nunes, UFC 178

“That was the shortest combined distance in a complete round between the two of them. They moved 186 feet. Zingano spent 30 of them getting across the cage. So, if you take out those 30, you’ve got 156 between the two in five minutes.”

Genauer says motion data is helping Fightmetric and the sport, generally, start asking and answering hugely consequential questions about the nature of mixed martial arts fights.

“With enough motion data, you can begin to answer certain questions. My most interesting one is, ‘Do judges only really look at the guys moving forward? Does moving forward matter more than strikes?’ Maybe it’s difficult to tell on a strike-by-strike basis who is landing each and every one of them, but it’s very easy to see who is moving forward or backwards.

“So, if you’re a fighter who is moving backwards, are you putting yourself at a severe disadvantage even if you’re able to outland because if it goes to a decision, the judges just aren’t going to give it to you?,” he asks. “They’re fundamentally predisposed to give it to the fighter who is moving forward. That’s the kind of the question that can alter strategy, can alter commentating or a lot of different things if it turned out to be true or false.”

He’s also able to ask a range of other questions related to fight performance or outcome, including whether ‘corner time’ – the amount of time a fighter spends within shouting proximity to their corner – affects their ability to win or perform better in particular physical or spatial contexts. Genauer also has suggested it could be possible with enough information to determine just how much a fighter’s odds of securing a takedown decrease the more there is distance between themselves and their opponent. That could reveal a takedown tipping point or a better explanation of how successful takedown artists manage distance or perhaps another revelation altogether.

There’s also the issue of what distance traveled could tell us about fights in different weight classes, genders, age and more.

Since the spatial data marries easily with the existing numerical data, Genauer argues, the motion-tracking data is a strong compliment to statistics observers already use.

As far using motion-tracking data that is useful or palatable to UFC television audiences, that’s where Genauer says the heatmaps come into play. “It helps illustrate some things the numbers are a little dry for,” he claims. “Center control for McGregor’s 64-percent to Siver’s 1-percent, everyone grasps that. Display it visually and you can really show the dominance.”

Still, for all the promise of motion-tracking data and its corresponding visualization, it’s not without its limitations, a fact Genauer is aware of and readily acknowledges.

For starters, he’s only tracking how much fighters travel, not ‘move’ aggregately. If they stay in place on the ground in a grappling exchange, the current state of motion-tracking doesn’t add much if they don’t compliment that by also traveling around the Octagon. “If something in staying in place and rotating, that wouldn’t count. They haven’t shifted center of mass,” he admits.

Second, visualization is a work in progress. As it stands, heatmaps can only show rounds, not entire fights (provided a fight is more than simply a round). While Genauer believes the way a fighter moves in the cage might show up visually in patterns over the course of a career, the picture can be a muddled mess after a three-round fight if there is an attempt to display the fight as a whole.

Lastly, Fightmetric can’t go back in time and measure the motion of older fights. A dedicated camera above the Octagon is used specifically and solely to collect the data. There’s no way to watch older PRIDE or UFC fights and get a raw data on motion (Genauer does note once we learn more about what motion-tracking data means and can show us, we can go back and watch old fights where we’ll be able to at least make some anecdotal notes or generalized points about a fighter’s movement). Yet, Genauer isn’t necessarily concerned about lacking that information in terms of creating adequately-sized data sets. He conteds the data size will grow fast with UFC’s aggressive schedule of shows. “Through 2006, UFC only ran 800 or so fights,” he recounts. “You can eclipse the first ten years of UFC existence in almost one year.”

But while the challenges with motion tracking are real, but the data is already as helpful as it is illuminating even in the infancy of its development. That’s why it’s not just part of what Fightmetric is doing, but in sports statistics, generally.

“The reason is why it was so important is it’s become such a fundamental part of the way other sports are analyzed,” Genauer argues. “MLB, this year, unveiled and rolled out this enormously expensive and tremendously sophisticated tracking system that’s based on cameras and radar, so that they’re looking at the motion of the players, the motion of the ball, bat speed, exit velocity of the ball off the bat, defensive efficiency.

“Motion tracking is the wave of the future,” he states further. “Basketball’s been doing it for several years. MLB did it in a more comprehensive way than they have in the past starting this season. The NFL introduced it in some games last year where they’ve got a chip in the shoulder pads of players. They can take a look at these things as well, like what it means if a corner is playing a guy off the line or giving him more space.”

The trend towards increased emphasis on motion-tracking data isn’t just widely adopted across sports as a blind mechanism to keep up with the Jones’s, he believes. As far as Genauer is concerned, it’s the next logical and evolutionary step from numerical metrics to something more robust that gives greater meaning and understanding to what’s happening on the field of play, be it a basketball court or Octagon canvas.

“This is really the next frontier. If looking at all the performance metrics was one phase, the next phase is turning those two dimensional statistics into three dimensions,” he says. “Now you can look at the outcomes or what happened, but also how did everybody get there? It should hopefully complete the story a little bit more, maybe explain those numbers better than they are in the abstract.”