The Oakland Ballers let an AI manage the team. What could go wrong?

There’s a classic Simpsons episode in which the sly businessman Mr. Burns recruits real Major League Baseball players to join his company softball team in order to win a bet. But when the championship is on the line, Mr. Burns pulls eight-time National League all-star Darryl Strawberry for a substitute, Homer Simpson.

“You’re a left-hander, and so is the pitcher. If I send up a right-handed batter, it’s called playing the percentages,” Mr. Burns says to Strawberry. “It’s what smart managers do to win ballgames.”

High-level baseball is very mathematically-driven, with teams hiring dozens of data engineers to study granular statistics that can inform managerial decisions. But like Mr. Burns in that Simpsons episode, it’s tempting to overanalyze baseball statistics to the point of absurdity.

The Oakland Ballers, an independent Pioneer League baseball team, took that concept of “playing the percentages” to the next level: they let an AI manage the team for a game.

The Ballers were founded by edtech entrepreneur Paul Freedman as a salve to the departure of the beloved Oakland A’s, the Major League baseball team that owner John Fisher ripped away from local fans in what’s regarded as one of the most insidious managerial moves in sports history. Though they’re not a Major League team, the Oakland Ballers – coyly, the Oakland B’s – established an unprecedented national community of fans who rallied around the team in protest of the A’s departure. After just two seasons, the Ballers just won Oakland’s first baseball title since 1989.

“The Oakland Ballers uniquely have the experience of being like a major league team in a minor league market,” Freedman told TechCrunch. “We can have creative flexibility. We can play with things and experiment with things way before the MLB or NBA or any of those leagues could do something.”

Minor league baseball organizations are often called upon to test new technology before it gets implemented in the majors, like challenging calls with instant replay footage or the automated ball-strike system. The Ballers embraced this attitude, especially given Freedman’s own background in tech, but have added in a dash of whimsy, piloting things that would never actually debut in Major League Baseball.

Last year, this meant partnering with Fan Controlled Sports to let fans make managerial decisions during one late season game. The Ballers ended up losing that game, in part because fans advocated for the funniest possible managerial decisions, rather than the savviest – at one point, the fans called upon a pitcher to pinch hit.

This time around, once the team clinched their postseason berth, the Ballers partnered with the AI company Distillery to commission AI software that could manage a baseball game in real time.

“Baseball is the perfect place to do an initial experiment like this, because it is so data-driven, and decisions are made very analytically,” Freedman told TechCrunch. “You have the pace to be able to do something literally after every pitch.”

Distillery trained OpenAI’s ChatGPT on over a century’s worth of baseball data and analytics, including Ballers games, to approximate what decisions Ballers manager Aaron Miles would make.

“What the AI did was figure out what our human coach would have done – the ingenuity on strategy and the concepts came from [Miles], and the ability to use the data and recognize patterns… is what the AI did throughout the course of the game,” Freedman said. “So I think the role of human ingenuity is safe for now, and AI is a tool to be deployed to optimize decisions, but not to make them.”

The AI-controlled game went smoothly – in fact, the AI made all of the same choices regarding pitching changes, lineup construction, and pinch hitters that Miles would have made. The only time that Miles had to override the AI was to replace the starting catcher with his backup, since he was sick.

Miles took his temporary replacement by AI in stride – perhaps because he knows that his job is not actually in jeopardy. In a video posted to the Ballers’ Instagram, Miles walks to home plate before the game to shake hands with the opposing team’s manager – only instead of offering his own hand, he extends the tablet running the AI out for a handshake.

But the use of AI struck a nerve for Oakland fans, who see companies like OpenAI – which powered Distillery’s baseball AI – as enterprises that prioritize “winning” the AI race over shipping products that have been properly tested for safety. For many fans, the AI experiment felt like a betrayal, similar to the kind of corporate greed that pushed three professional sports franchises out of Oakland in five years.

“There goes the Ballers trying to appeal to Bay Area techies instead of baseball fans,” one commenter wrote. “It’s so over for Oakland.”

This backlash wasn’t what the Ballers expected, and Freedman does not intend to repeat this AI experiment. But the reaction of the fans models a larger cultural tension in baseball and beyond.

“It never feels good to have your fans be like, ‘We hate this,’” Freedman said. “But it’s not a bad thing that there’s more of a conversation about the pluses and minuses of this new technology now, as opposed to like, a decade later when it’s too late.”