Advanced statistical analysis has made a huge impact on baseball, and it's making inroads in football and basketball as well. Soon enough, it's going to come to mixed martial arts.
There will be skeptics -- the MMA versions of the old-school baseball people who were sure that they knew more than Bill James and that Moneyball was a sham -- arguing that the sport is "fought in a cage, not in a spreadsheet."
But Rami Genauer, who runs the innovative FightMetric.com, thinks MMA is ready for advanced stats. (I agree.) His web site is billed as the first-ever comprehensive mixed martial arts evaluation system, and the work he's doing has the potential to change the way we view the sport.
In a wide-ranging Q&A, I posed some questions to Genauer, starting with why MMA needs advanced stats.
Why does MMA need advanced statistical analysis?
Let's not get ahead of ourselves. Before we get to advanced statistical analysis we need basic statistical analysis and before that, we need statistics.
People have compared what we're doing to the early days of statistical analysis in baseball, but Bill James and the early guys at SABR (Society for Advanced Baseball Research) had a huge head start: Information already existed, in the form of box scores and other resources, so they could easily populate a database with nearly 100 years of baseball statistics. The genius of those people was in taking the information that everyone knew and creating meaningful metrics people had never thought of, like Runs Created and VORP.
MMA, on the other hand, doesn't have anything. There is no database. Every major fight in MMA history must be re-watched and fully scored.
But just imagine what one would do with all of that information. For instance, I'd be curious to see if fighters who have absorbed a large number of leg kicks early in the fight are less successful with their takedown attempts in the later rounds (normalized against those fighters' traditional late-round takedown success rates). It's exactly stats like those that will allow us to challenge orthodoxies and find statistical truths. But until we have that big database, things are limited.
Will we some day reach a point where we discuss an MMA fighter's stats the same way we discuss a football player's touchdowns or a baseball player's home runs? If so, which stats will we discuss the most?
I think it's reasonable to assume that some stats will catch on. While I'd love to believe that casual MMA fans will discuss some of the more innovative, comprehensive stats we've created, the reality is that people are most concerned with volume and accuracy. How many strikes or takedowns did a guy land and what was his success rate?
The problem is, there is not even consensus on what are the kinds of things that should be measured. Imagine if Bill James would have first had to convince the world that batting average should be Hits/At Bats instead of something like Hits/Total Swings. That's where MMA is currently. FightMetric has its designated 67 statistical categories, but we're fully aware that there are other systems measuring completely different things. As an example, we believe there is no qualitative difference between a punch or a kick of equal power to the same target; it doesn't matter what you strike with, only where you strike to and how hard. So while some folks will concentrate on a fighter's kick accuracy, we think that's a meaningless statistic based on a purely cosmetic factor.
In general, does statistical analysis lend itself more to team sports like football and baseball or to individual sports like MMA? Will statistical analysis in MMA ever reach the heights it has reached in baseball?
I don't think it has to do with teams vs. individuals. The issue is that MMA is a mess in terms of creating comparable data sets. For example, every major professional baseball game is nine innings long and the way to beat your opponent is by scoring more runs than they do. MMA has variable match lengths; some last 20 minutes, some last 20 seconds but result in the same outcome. MMA has variable rules and judging criteria; what won fights in Pride wouldn't work in the UFC. To make things worse, MMA has multiple victory methods or differing quality; you can dominate your opponent or score a lucky cut stoppage or eke out a bogus decision and they all equal a win on your record. Some would look at those factors and say MMA and stats are incompatible. I think it means we have to think more creatively.
What is your sense for how receptive the MMA establishment would be to statistical analysis?
There will be people who love it and people who hate it, same as for every sport. Baseball has its statheads and its people who still think Billy Beane wrote Moneyball. I do think that acceptance of stats will be a bit harder for MMA than for other sports. I've encountered people who are not comfortable with the idea that math can be used to measure something as primal as fighting. While I understand the urge to want to believe that you can't measure things like this because "in a fight, anything can happen," I completely disagree with it. If you asked them to, an actuary could tell you what the chances are that you will be killed on a Tuesday while choking on some Fruity Pebbles based on factors like frequency of Tuesday Fruity Pebble consumption and the ratio of average Pebble size to the circumference of your esophagus. Just because anything can happen, doesn't mean we can't measure it.
To what extent does statistical analysis help you predict the results of future fights? Could a person consistently make money in Vegas by using stats to pick winners?
If you subscribe to my newsletter, I can reveal to you the miracle secret THEY don't want you to know about., which will triple your gambling winnings with no risk...
One thing I should say is that stats can't make predictions, but they can make probability-based projections. You could say that Fighter X should beat Fighter Y 80% of the time, but that doesn't mean that Fighter X is going to win.
As for the Vegas part, I am fascinated by sports gambling from an economics perspective. The reason is that it's not like betting on red or black on a roulette wheel, where a known probability (47%) produces a standard payout (1:1). Fight odds are based on the public's perception of how close a matchup is. The oddsmaker needs to keep the odds at a place where bettors have the incentive to bet equally on both fighters. If someone came along with great stats that could predict winners with a high degree of confidence, all it would do would be to make the favorite a bigger favorite and the underdog a bigger underdog. If everyone had access to this information, the net gain in money for gamblers would be zero compared to what they have now. The market would correct for any improvement in information. The only way to beat the system is not to play.
If you're curious, I did a study of betting lines before and after the debut of The Ultimate Fighter and found that spreads have gotten wider in the last few years, as newer fans put disproportionate money on (usually better marketed) favorites.
Every college football fan understands that Hawaii's 12-1 finish last season was nowhere near as impressive as LSU's 12-2 finish, and we have metrics like the Sagarin ratings that help us assess teams' schedules. But do we have anything like that in MMA? Do we have any way of objectively comparing two fighters who have the same record against two totally different slates of opponents?
Not yet, but we're working on it. The goal is for it to function like the BCS, but less evil.
Has your statistical analysis led you to conclude that any particular fighter is overrated or underrated?
I can't speak to whether fighters are over- or underrated, but I know what kinds of fighters the system does and doesn't like. Lay-and-pray fighters do not get high performance quality ratings from our system because we don't give points for simply maintaining positions. You have to do something to earn points. On the flip side, the system loves quick finishers like Anderson Silva and hyperactive guys. Clay Guida and Karo Parisyan, win or lose, are two good examples.
What is your opinion of Kimbo Slice?
Don't have a large enough sample size to say anything statistically, but I admire the fortitude it takes to grow a beard like that in the face of what must have been constant pressure to shave.
Do you hear MMA announcers say things that your statistical analysis has proven wrong?
Sometimes you'll hear things that are based more on reputation than fact. For instance, every indicator we have says that Georges St. Pierre is a better MMA wrestler than Matt Hughes. But since Hughes is known as a great wrestler, that was considered a controversial statement.
What does the mainstream media consistently get wrong about MMA?
The general issue is that media is an echo chamber. When I worked as a journalist, I would often have to write a story about a topic with which I wasn't familiar and I didn't have a lot of time to become an expert. The way I educated myself was by reading articles others had written on the topic. While I didn't copy their work and I certainly conducted my own interviews, my article was based on their article, which was based on another, etc., etc. The fact that Senator McCain once called MMA "human cockfighting" is a worn out cliché, an opinion which he has since recanted. Yet you'll still see it mentioned in every new MMA piece as if it were news because the journalist who wrote it read that same thing in all the other articles he's seen. There's a reason why articles about MMA sound so similar. In fact, as a joke, I created some Mad Libs for MMA articles, one cautiously optimistic, the other a hatchet job. I've been curious to see if someone could really fill one out and get it published in the traditional media.