WARNING: Unlike other pieces where it's toned down, I don't hold back the nerd-speak today. You will see phrases like "rank-order competition," "non-linear compensation" and "optimum effort levels." Read at your own risk...but read anyway.
I love the end of June and early part of July. Yes, I love my country (July 4th) and UFC Fight Week, but it's also the time of year when the North American Association of Sports Economists (NAASE) holds their annual meeting for five glorious days of sports nerd nirvana. In the past, I've presented here on the behavior of NBA refs in make-up call situations (take a wild guess as to results), NBA refs and the Napoleon Complex and, last year, I presented the first ever analysis of the round-by-round scoring decisions of MMA judges, revealing the values they tend to place on different types of strikes and grappling techniques.
This year, I wanted to see if fight night bonuses have an observable influence on fighter behavior. If they don't affect behavior, they're a somewhat strange compensation mechanism - rewarding workers for no change in productivity. If they do, it's fascinating that fighters could explicitly or implicitly be influenced by money in the midst of battle with someone trying to rip their head off or snap their arm.
The NAASE meetings are one component of an even bigger nerd conference put on by the Western Economic Association (WEA). The difference between this conference and the more famous MIT Sloan Sports Analytics Conference is that the NAASE meetings are solely focused on rigorous research. Sloan has more panels with famous team executives, but it also has some research papers with an unknown quality control process. I went through the process two years ago and still don't know how it works.
The conference itself is incredibly fun and I'd encourage anyone to go. There's a lot of networking, glad-handing and team executives not wanting to reveal their secrets. My wife and others call it "white guys in suits" since it's packed with dressed-up MBA and college students wanting to land a job in sports. They've only ever had one MMA panel (interestingly enough titled "MMA Analytics") and that was a couple years ago with Rami Genauer of FightMetric, a few journalists and Dominick Cruz. Rami was essentially the only one who could truly speak on MMA analytics.
NAASE is much different. Each participant shares their paper in advance with someone who's qualified to review it (i.e., a discussant). You present for 15-20 minutes and then the discussant critiques and criticizes your work right to your face. This is science in action. If your work stinks or is fundamentally unsound, everyone in the room will hear about it while you sit there and take it like a (wo)man. It's not flashy, there's no media and we don't wear suits. But we get complements and criticism on our work in progress so as to improve and make it better.
The dedicated sports nerd room
My current project is very much a work in progress. I've got some initial results but I'm not 100 percent convinced I'm looking at it the right way. I'll share some of it today and let you guys be my discussant. Tell me what you like. Tell me what you don't like. I'm always nervous that a future journal will claim it's already been published if I share too much online, so I can't disclose most of the tables and figures.
Fighting in an MMA promotion is a series of tournaments. Fighters compete in a rank-order competition with the winner ranked higher than the loser, and the hope of progressing to the top of the division and titleholder status. Tournament theory was developed in the 80's and explains the highly non-linear compensations structures observed in tournament environments as a means to induce optimal effort levels. The primary example wasn't actually sports, it was executive compensation. "The salary of the vice president acts not so much as motivation for the vice president as it does as motivation for the assistant vice presidents." Throw in the superstar effect and we've got an even more skewed distribution.
Want an example of this non-linear compensation? Here you go.
This is the distribution of disclosed fighter show money (in thousands) throughout my sample of UFC and WEC events from 2001-2014. (I realize including the WEC skews the results a bit, but the shape is similar with UFC only.) The majority of fighters earn less than $25,000 show money while a select few elite fighters earn $150,000 or more (excluding PPV percentages and discretionary pay).
Within each event, there's another tournament going on - the tournament for bonuses. Everyone's eligible to win Fight of the Night (FOTN), but fighters must first qualify to win Knockout of the Night (KOTN) or Submission of the Night (SOTN). (My sample doesn't include events with Performance of the Night bonuses, but I suspect findings would still hold for POTN since the UFC appears to be using it pretty much like KOTN/SOTN unless there aren't any knockouts or submissions.) Bonus awards are large relative to fighter show money, coming in at an average of 360%. In theory, that should represent a strong incentive.
A fighter's effort levels should be affected by their perceived returns to effort. I calculate the potential returns to qualifying for a bonus based on the fact that all fighters have technically qualified for FOTN and, at the moment a fighter enters the cage, he knows or has the ability to know how many fighters have earned a qualifying TKO or submission (Does this seem reasonable? This is one potential area of concern. I'm also considering the role of expectations. To be clear, it's not suggesting that fighters have to watch the prior fights in the locker room, just that they know the outcome - who won and was it a decision, TKO or submission.) Since rich people don't value an extra dollar the same as poor people, I also compute salary-adjusted returns, dividing the nominal returns by each fighter's show money.
Now the question is what observable elements of performance should be examined? Obviously, I could look at the probabilities of getting a knockout or a submission. I'll eventually examine these, but I'd prefer something cleaner. In my mind, a cleaner element of fighter decision making is the choice to shoot for a takedown at distance. Takedowns with clinch control could also work, but the data isn't as clean in this situation.
At distance, if you're more inclined to be on the ground you can shoot for more takedowns. If you're less inclined, you can shoot for fewer. Is there still other stuff going on? Of course. But we can control for some of it as well as fighter histories and certain fixed effects. If a fighter is influenced by the KOTN return, we should see fewer distance takedown attempts (adjusted for time). If a fighter's influenced by the SOTN return, we should see more. What's the result? Some of the controls explain takedown attempts but the bonus returns explain nothing.
I also look at distance power strikes attempted, submission attempts (in good submission positions), bouts won by TKO and bouts won by submission. All come back with a benign effect of fight night bonus returns except for distance power strike attempts. Better FOTN returns may lead to more swinging with power at distance. SOTN bonus returns also have a possible effect here but I'm not fully convinced given the other results, and am somewhat concerned about the multiple comparisons problem.
I know there wasn't great detail and a lot was left out. But if you have any thoughts or suggestions, share them in the comments or e-mail me directly. This is all still very preliminary yet, as of now, it appears that KOTN and SOTN bonuses (now POTN) have little to no effect on fighter behavior in the cage while FOTN may incentivize fighters to swing away a bit more at distance. This is still all preliminary and everything I've just discussed needs more thought, more analysis and more time.
Other sports research
Joe Price presented a model of productivity spillovers in the NBA. To translate that into English, does the presence of certain people make their co-workers better? Do others make their co-workers worse? You can also think of him as addressing the Mark Cuban critique of the Wins Produced advanced stat (here and here). Joe looks at this in the NBA and finds, yes, spillovers matter. Sometimes it's claimed advanced stats don't treat Kobe and Carmelo well because they don't account for the double, triple teams and other attention that end up improving the performance of their teammates. This model does...and Kobe and Carmelo come out even worse.
Josh Price presented a model of positional wins above replacement (WAR) in the NFL. The tricky thing about assigning value to players in the NFL is determining who gets the credit. Is it the receiver who caught the ball, the quarterback who threw it, the linemen who blocked, another receiver who served as a decoy, etc.? Josh addresses this by essentially allowing the market to give the answer and looking at when players go down with injuries and their backups actually play. It's all still preliminary, but he finds the most value in QBs, FBs, WRs and the outer Oline. Remember, this isn't the highest absolute value, but WAR value.
That's a taste of some of the things that go down at the NAASE meetings. Even though I was the only MMA presenter, one positive is that a number of older, non-fighting, non-MMA types came up afterwards and talked about how glad they are someone is doing MMA research. Their kids and students are interested in it and there's nothing out there.