The Latest Assistant Coach Is An AI Chatbot
Sports celebrate human excellence, but now there’s too much data to analyze to determine potential mismatches on game day.
(Bloomberg Opinion) -- Imagine it’s crunch time at the end of the fourth quarter of the Super Bowl. One team faces third-down and goal, five yards from a game-winning touchdown. On the sidelines, the head coach and his gut say run the ball, but he’s unsure where to send the running back. So, he turns to an assistant coach and asks: “When you spoke to our AI chatbot this morning, what did it think we should do in this scenario?”
For sports purists, the scenario is unthinkable. Sports are supposed to celebrate excellence, after all. But whether they like it or not, artificial intelligence is joining games on and off the field. Some teams are already utilizing it, and more will join them if only to keep up.
For all the concerns about AI jobs (bad), this is an example of how AI can help people do those jobs more efficiently by removing some of the tedious labor (good). Rather than replace the human element, generative AI is designed to recognize patterns and situations in specific sports that can leverage data and intensify competition.
To get an idea of what the future might look like, I recently joined over 2,500 sports industry professionals, including representatives of at least 120 teams, at the MIT Sloan Sports Analytics Conference. The forum, held since 2006, covered a range of analytically-oriented topics, from insights into how the US Olympic team maximized its figure skating medal results to whether the NFL Scouting Combine is still an effective way to judge talent.
Over the years, the job of analysts has become harder because of the growing volume of sports data. That’s not just in terms of hard numbers, such as yards gained in football or shots on goal in soccer. Now, there is an uptick in video being generated — from broadcast video or embedded cameras at events — to review individual athlete performance, too.
As the files accumulate, the challenge of actionable conclusions escalates. For example, in an average National Hockey League game, there are over 3,000 different measures of player strengths and weaknesses, according to Brian Hall, founder and head of artificial intelligenceat AlphaPlay AI, a generative platform used by professional hockey and soccer teams.
The human mind can’t comprehend that much data, and NHL teams, at least, lack good systems to process it, Hall says. As a result, they fail to see the potentially useful information — such as defensive mismatches — buried in the stats.
That’s where AI is helpful.
“All of a sudden, it can give you insights from its own thinking as to what it thinks is useful [for] winning,” Hall says. “What looks random may turn out not to be.”
The experience of using a sports AI, like AlphaPlay, is not unlike the experience of using ChatGPT. It is designed to learn from large data sets, recognize patterns within them, and then use that information to generate outputs understandable to non-computer scientists. For example, in advance of a recent NFL game between the Denver Broncos and the Jacksonville Jaguars (neither team is an AlphaPlay client), the system was asked: “We are playing the Broncos next week; what are some off-ball, often-ignored weaknesses in their rush defense?”
AlphaPlay answered by taking into account potentially millions of situations, player lineups, strengths, and weaknesses. Then it spoke the language of football: “Bradley Chubb has struggled to set the edge this season, exposing a potential weak spot in an otherwise strong Broncos defense.”
Coaches have carried clipboards of analytically derived tips like these for decades. What makes AI analytics different is two-fold. First, the tips come from a much deeper pool of information. Second, the results aren’t delivered as a table or chart; rather, they are presented in simple, direct and actionable language. When a situation requires a split-second decision, that straightforwardness can build confidence in a coach’s intuition about the right choice.
And notably, the decision is still the coach’s (a human being). The fast-changing dynamics of a real-time game still require a coach to decide when to deploy the AI-generated advice (a human won’t run the same play over and over just because the AI identifies a mismatch).
More From Bloomberg Opinion’s Adam Minter:
- Caitlin Clark Was Almost Paid to Stay in School. Good Idea.
- Women’s Hockey Is Breaking the Ice Ceiling
- College Sports’ Future Shouldn’t Be Determined in Courts
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Adam Minter is a Bloomberg Opinion columnist covering the business of sports. He is the author, most recently, of “Secondhand: Travels in the New Global Garage Sale."
More stories like this are available on bloomberg.com/opinion
©2024 Bloomberg L.P.
