For a sport that’s more than 150 years old, the opener of the 2026 Major League Baseball season will feature an unusual number of firsts. The official Opening Day on March 26 is the earliest in baseball history. The first official game of the season tonight between the Giants and the Yankees, which is the Apertura game. EveningnoOpening Daytotally different, it will be the first game broadcast on Netflix.
And chances are, at some point during that game, a player will hit his helmet or cap after throwing a pitch, challenging the umpire’s call and triggering the first review of baseball’s Automated Ball-Strike System (ABS). The robot referees are here.
The system is remarkably simple. Each team receives two challenges per game, retains them if they are successful and loses them if they make a mistake. Only the pitcher, catcher or batter may contest, only on balls and strikes, and only within two seconds of the pitch.
Once a challenge is presented, a network of 12 high-speed cameras installed around the stadium tracks the exact location of the field and then the software creates a 3D model of the field’s trajectory (on the Jumbotron for all to see) against the batter’s individualized strike zone. The verdict is issued instantly. The referee doesn’t go to a monitor and reconsider for minutes, like in an NFL or NBA replay. He is simply the conduit to announce what the machine has decided.
In theory, this change should improve everyone’s situation. Teams have recourse in the event of a possible botched call at a crucial moment (such as the brutal strike that ended play for the Dominican Republic in this month’s World Baseball Classic). The challenges are limited and decided quickly, so the game doesn’t slow down. The automated system is accurate to within 0.25 inches (about the width of a pencil) and is fast enough to catch a 103 mph Aroldis Chapman fastball. Human referees are still largely in charge of the game.
All in all, the ABS system seems to be an ideal compromise: preserving human judgment while allowing machines to correct the worst errors. While the system is not powered by AI, it looks like an example of how humans and AI could work together fruitfully in the future, with humans firmly in the loop but aided by machines.
Except there’s a problem with splitting the difference between humans and machines. Once you’ve admitted that the machine is the final authority on whether a decision is correct (which is exactly what baseball has done here), you’ve quietly eliminated the argument for having the human there. What might seem like a stable equilibrium is not stable at all.
Call balls and strikes
This breakdown can already be seen in progress in the minor leagues, which have been experimenting with the ABS system for years. Baseball reporter Jayson Stark has written about umpires in the AAA minors who, tired of being flipped by the machine for all to see, began changing the way they handled the game, “calling balls and strikes the way they think the robot would call them.”
Because the league has given the machine the final say, the human behind the mask does not remain independent: it begins to imitate the machine. The referee, once the lord of the diamond, whose word was law, becomes the de facto eraser of the AI. Human knowledge and experience degrade.
To which a baseball fan might respond, perhaps with more colorful language, “they’re all bums anyway.” Which wouldn’t be entirely fair to our carbon-based umpires, even though fairness with umpires has once been a concern for baseball fans. MLB estimates that umpires throw 94 percent of pitches correctly, which on the one hand is good (I’m not sure I’m 94 percent accurate on anything) but on the other hand means they still make mistakes on about 17 or 18 pitches per game on average.
And while the data suggests that umpires have actually gotten better, we can now see replays and accurate pitch-tracking data that make it abundantly clear when a call was blown. A guy named Ethan Singer even created an independent project called Umpire Scorecards, which uses publicly available pitch tracking/Statcast data to grade every umpire, every game. The new ABS system only confirms what previous technology made evident years ago.
So the technological assault on the referee’s authority has been underway for some time, and while even the ABS system has its margin for error, the end result of the introduction of machines will be a more accurately called game. But true human abilities will be lost along the way. The best catchers are experts at structuring throws to make them. look as strikes, even if they are not. Good hitters learn the umpire’s individual strike zone and adapt game to game. (Red Sox great Ted Williams used to say there were three strike zones: his, the pitcher’s, and the umpire’s.) All of these skills were built on human imperfection, and they will all become less valuable even as machines make the game “fairer.”
The one-way street of automation
To glimpse the possible future of baseball, just look at tennis.
In 2006, professional tennis introduced Hawk-Eye challenges, which allowed players to appeal a limited number of line calls to an automated camera system. At first, the players were not fans. (As Marat Safin said: “Who was the genius who came up with this stupid idea?”)
But the logic, especially as the sport became faster and faster, was undeniable. By 2020, the US Open had eliminated human lines judging entirely, and Wimbledon followed suit in 2025. Human referees are still employed, but primarily for match management purposes; that is, to silence the crowd. The challenge system turned out to be just a stop on the path to almost full-scale automation. And now baseball is taking the same path.
ABS is what you get when an institution knows the machine is better at the job but isn’t willing to say so. That’s exactly the position many organizations find themselves in right now, as AI becomes increasingly more capable. The result, for the moment, tends to be a hybrid approach that leaves too many workers stressed and disempowered, while failing to reap the benefits of more complete automation.
But over time, automation tends to become a one-way street. The question is not whether machines will eventually call balls and strikes. It’s how much longer the middle ground can last, for those referees we love to hate and for the rest of us.
A version of this story originally appeared in the future perfect information sheet. Register here!