Bill Belichick's Superiority Has Been Proven By Science

Jerry Thornton3/07/2019 4:51 PM

The mathletes at MIT just met for their annual Sloan Sports Analytics Conference, where they basically engage in the nitro-powered thrill ride that is presenting research papers on athletic performance and sports management.

Among the finalists at this year’s Analytics ComicCon was “How Much Do Coaches Matter,” by University of Chicago public policy professors Christopher Berry and Anthony Fowler. Which examines the effects a coach can have on team in pro football, baseball and college basketball using their probably-exactly-as-much-fun-as-it-sounds “Randomization Inference for Leader Effects” method:

1. Residualize game-team level data by season, home-field, and quality of opponent (optional).
2. Aggregate to season-team (optional).
3. Regress the outcome on coach fixed effects and record the r-squared statistic.
4. Randomly permute coaches within each unit, sampling each coach’s tenure as a block.
5. Regress the outcome on permuted coach fixed effects and record the r-squared statistic.
6. Repeat steps 4 and 5 many times, recording the proportion of cases where the r-squared from the permuted data is greater than that from the real data

And if you think that kind of talk wasn’t making panties hit the convention space floor from one end of the hall to the other, you’ve obviously never been to the Sloan Analytics event. A little r-squared statistics and the custodians start reaching for the mop bucket.

To summarize, this is what they found:

Baseball – “managers have a greater impact on runs allowed than runs scored.”
Football – “coaches ‘significantly affect’ a team’s fumbles and penalties.”
Basketball – “for college basketball, coaches had a greater impact on points scored and allowed, than on point margin between teams.”

And, most significantly:

Bill Belichick – “Using available data through 2016, the researchers found that New England Patriots head coach Bill Belichick was 18 percent more likely to win a game (based on the quality of his opponent and home field advantage) than an average coach. While that isn’t the highest percentage of his peers, it is “extremely unusual” given his 17 seasons with the team.
“Could Belichick be the luckiest coach in the NFL? The professors ran a simulation based on 10,000 hypothetical NFL games, and in only five cases would a coach with a 17-year tenure have Belichick’s same record.
“’In that sense,” Berry and Fowler write, ‘we can strongly reject the null hypothesis that Belichick is no better than an average coach.’”

Now before anyone goes to the Belichick Hater default setting of “That’s only because he has Tom Brady,” this research specifically “factors in player quality and strength of opponents, along with the ability to separate team performance based on coaching versus performance by chance.” So factor Brady out of it. R-square the “AFC East is a bunch of tomato cans” argument straight to hell. Remove all elements of luck and what have you got?

Out of 10,000 hypothetical games, Belichick does better than any other coach in 9,995 of them. I’m not as good as maths as the guys at MIT, but I think those are the same odds Dr. Strange gave the Avengers in Infinity War:

Of course, by their own admission, these professors only used data up to 2016. So it’s fair to ask what Belichick’s done since then? Oh, right. He’s gone 29-9 (including postseason) taken his team to both Super Bowls, won the most recent one and added another ring to the name of a boat, in order to hold the next one, is going to need a bigger “R.”