Hi there, it’s Jim Gaddy, one of the deputy editors here at Pursuits. It’s Wild Card Weekend, and if you don’t know what that means, you might want to skip to the next section.
But if you’re one of the 30 million or so people who will be tuning in to the National Football League playoff games on Saturday, Sunday and Monday, there’s a fascinating angle on machine learning that most people don’t know about the popular sport in the US.
It’s been a strange year for the NFL. Sure, it’s as popular as ever: Pro football games accounted for 82 of the 100 most-watched television broadcasts in the US in 2022, a record. The combined value of its 32 teams is over $130 billion.
But the two biggest storylines of the year were about injuries, both during primetime. On Jan. 2, Buffalo Bills’ Damar Hamlin suffered cardiac arrest on the field after being hit in the chest. In another nationally televised game earlier in the season, Miami Dolphins quarterback Tua Tagovailoa’s hands showed signs of a neurological injury after he hit his head on the turf.
The league has not always had the players’s best interests at heart. It was late in acknowledging the impact of head trauma. It added a 17th game to the season. It started playing games on Thursdays, giving players only three days of rest in between. Theses moves have made it an easy target for people who say that it doesn’t care about the people who make it such compelling drama.
But there is some evidence that the game is safer today than 10 or especially 20 or 30 years ago. And it’s partly because of how the league is using machine learning.
For the most part, computer analytics is being used in baseball, basketball, and soccer as a way to gain a strategic edge. And NFL teams are doing that also, to a degree: You see more teams going for 2-point conversions, for instance, or going for it on 4th down.
But it’s also being used in simulations to determine how concussions and even soft-tissue injuries occur. Last year the league created the Digital Athlete, an initiative that uses computer simulation to reconstruct scenarios in which injuries of all types occur.
“It allows us to understand: Why did this player get injured? And then identify things we could do to change that outcome,” says Sam Huddleston, principal data scientist at Biocore LLC, the NFL’s engineering partner.
One of Biocore’s findings was that load management in preseason training has been completely backward. For decades, teams began preseason practices aggressively and then cut back on the amount of physical contact at the end. But by starting out hard, players often got injured early. “Slowly ramping the load will pay off for you in the end,” Huddleston says.
The NFL mandated this ramping-up effect last year, limiting how many minutes players could practice in their first four days back. The results, the league reports, are “significant,” including “a 26% year-over-year decrease in lower extremity injuries during the first two weeks.”
More of this sort of thing will be coming in 2023. Here’s the full story. And don’t forget to check out the Business of Sports podcast this weekend, where the author of the piece, Michael Croley, and I discuss it in more detail.