The Texas Athletics AI experiment is worth monitoring

Good morning, and thanks for spending part of your day with Extra Points.

Last week, Texas did something that caught the attention of regular sports fans, not just the SBJ-reading crowd that tends to occupy my corner of the internet. The university announced a massive new project: Bevo Lab.

Bevo Lab, to me, sounds more like a secret Texas A&M project meant to advance various Meat Sciences. But it’s actually an AI-centric laboratory centered on college sports at the University of Texas at Austin. Via CBS:

❝

Texas on Friday announced a new sports analytics and artificial intelligence hub on campus called the Bevo Lab, a 25,000-square-foot space inside Austin's Innovation Tower. Google Cloud, Dell Technologies and SHI International are the founding corporate partners for what they believe is the first-of-its-kind hub focused on sports analytics.

…

[Kirk Goldsberry, executive director of the Business of Sports Institute at Texas] has hired 24 undergraduate students for Bevo Lab's first team, and they begin training with Google next week. Texas says the lab will involve up to 40 student researchers in its first year, scaling to 200 a year. [Texas athletic director Chris] Del Conte said the lab's construction should be finished in about three months. Goldsberry said it will be fully operational in 2027.

…

Texas uses a tool called Football Scouting Intelligence. It automates the opponent data and video reports that Sarkisian's staff used to compile by hand each week, Goldsberry said. The AI can also digest data from human movement tools developed by companies like Catapult, which track players' acceleration, deceleration, heart rate, and wear and tear. The lab plans to pull data into the system, starting with football.

If you’re at all inclined to read more nerdy sports coverage, Goldsberry’s name should sound familiar. He’s a UCSB PhD-holder who helped bring basketball advanced stats coverage to the masses at ESPN. He also worked as the vice president for strategic research for the San Antonio Spurs.

He’s not some breathless AI-hype-bro from LinkedIn, convinced the world has changed because he figured out how to get Claude to clean up his inbox. He’s one of the most credible names in sports analytics, somebody who hasn’t just produced a bunch of whiz-bang fancy charts; he also got coaches and managers in professional sports to pay attention to them. If he’s leading the effort, I’m inclined to give it credibility.

The purpose of the institute, as I understand it, isn’t to build a chatbot so coaches can completely delegate play-calling responsibilities to computers. I know everybody got the same jokes off. You know, some variation of this:

BLEEP BLOOP TEXAS INSTRUMENTS LONGHORN BOT BLEEP BLOOP

Upon further reflection, I see that the corner fade I described did not result in a touchdown. The ball sailed out of bounds and into the third row, where it struck a booster named Greg in the face. This is entirely my fault, and I appreciate you bringing it to my attention.

I’ve noticed I suggested the corner fade three times in a row after you said it wasn’t working. You’re absolutely right to be frustrated. Here’s a new suggestion: the corner fade. 🏈

And yes, if the Texas coaching staff treats Bevo Lab the same way your dumbest coworker treats ChatGPT, the Longhorns will absolutely run into problems, no matter how fancy the computer lab is or how beautiful a chart Goldsberry makes.

But if everything goes according to plan, the biggest advantage will be improving the efficiency of the actual humans, giving them more time to do the important human football stuff.