It’s not so often that you interview a professional athlete and he or she starts talking about the Nash equilibrium, but John Urschel is certainly not your average NFL player.
GeekWire caught up with Urschel, the Baltimore Ravens offensive lineman who also happens to be studying for his PhD in mathematics from MIT. For the second consecutive year, Urschel last week partnered with GE to be its “Calculations Correspondent” for the NFL Draft, offering up fun predictive equations that analyzed soon-to-be NFL rookies.
“I like math a lot, and I like football a lot, and I get to put these two things together,” he said.
— General Electric (@generalelectric) April 27, 2016
But Urschel noted that these calculations were just “basic stuff” meant to give the average fan more information about the players. It’s certainly not anywhere close to the Spectral Graph Theory, Numerical Linear Algebra, and Machine Learning Urschel is studying at MIT this offseason.
Starting my PhD in math at MIT this offseason!!! pic.twitter.com/vjcSctVx9i
— John Urschel (@JohnCUrschel) January 26, 2016
The 24-year-old Penn State grad, who is returning to his alma mater as the commencement speaker later this week, said he also conducts independent research into how data and analytics impact the game of football. He noted that relying on math equations to make football-related decisions can be a complicated grey area.
“If your mathematical calculations are correct, then the results you have are guaranteed to be right under your assumptions — but the question is, how good are your assumptions?” explained Urschel, who earned both his bachelor’s and master’s degrees with a 4.0 GPA at Penn State. “Analytics does have a place in football, but there is so much randomness and the sample size is small. Both are a hindrance to football analytics.”
Added Urschel: “Football is a very complex game, and therefore a model only captures a certain element of it in certain cases.”
It’s clear that using data to drive decision-making is becoming more and more important for teams in nearly every sport. But asked about the Seahawks’ decision to pass the ball on the goal line during Super Bowl XLIX, Urschel said that it wasn’t necessarily a scenario where Seattle head coach Pete Carroll should have completely relied on data. He explained a hypothetical situation with Ava from Ex Machina as the head coach — “she’s pretty much my future wife,” Urschel said of the fictional robot — in that same situation.
“If she’s coaching, you know she does her analysis, and for that situation there is a natural Nash equilibrium,” he said. “I can guarantee you that the equilibrium is not a pure strategy, which means that it’s 100 percent clear the other strategy is completely wrong.”
Say what, John?
More simply, Urschel said that NFL coaches will typically side with the probabilities when making play-calls, but sometimes they like to mix it up to keep the defense honest with what he described as “randomized strategy.” He said it’s not really fair for someone to say that Carroll’s decision to pass the ball was a terrible one.
“In a great majority of things involving sports, the optimal decision is a random one,” he said.
Urschel noted that computers won’t be replacing coaches on the sidelines any time soon — instead, “Ava will be eating dinner with me in our nice little suburban home hanging out with our robo kid,” he joked — but said that more artificial intelligence will be involved with football decision-making. He envisioned a day when offensive and defensive coordinators have supercomputers in the coaching box.
“I don’t see how having the extra tool could hurt, especially if it’s done the right way,” Urschel said.