Play Ball! How AI is Changing Baseball in Covid and Beyond
The long wait for baseball is over, but the game looks a lot different from what fans are used to: no more flyovers, a shorter season, new rules, players wearing masks, empty stands, canned sound effects, and… dancing robots? First in Asia and then the US, baseball eventually returned to the fields during the 2020 on-again-off-again (and now on again) COVID-19 pandemic. For the games to resume under such circumstances, drastic changes and advanced technologies had to be applied everywhere on the diamond, some of which you might have seen on TV.
Since even before COVID-19, AI has been utilized in baseball for scouting advancement, player training, health improvement, marketing, and fan interaction. Here we have gathered some of the most interesting ways in which AI is shaping the game around the world.
Predicting Pitches
In the 2020 Japanese professional baseball pre-season practice game, the TV broadcast shows an “AI pitching predictor”. Where you would usually see a strike zone, there is now a new function that predicts the next pitch. In order to do this, AI experts studied four million baseball pitches from the last 16 years’ worth of games and trained AI using all the data from past pitchers in every pitcher batter situation – 1 ball 1 strike, 3 balls no strikes, and so on. Watch the “AI Catcher” in action below:
AI Umpires
Refereeing is an essential but often contentious part of any sport, where a call made by a referee (usually one that benefits one team and punishes the other) can be contested on the account of bias or human error. In the world of baseball, those days may be numbered with the advent of AI umpires, which could take over the interpretation of the highly scrutinized strike zone. The MLB has been testing a ball-tracking system named Trackman which can detect ball speed, movement, type of pitch and other factors with far greater precision than a human umpire. While the technology is still a long way from maturity, the prospect of an infallible AI umpire can have major implications on the dynamics of the game.
Stealing Signs
Many of us have heard of Mark Rober, who contributed to the Mars Rover during his time working at NASA. In 2019, he created an app that allows users to analyze and steal the baserunning signs that third-base coaches give to their runners. Given the fact that the signs in baseball are very secret, they are made to be confusing. After talking to around 50 baseball players and coaches, he discovered that almost every team has an indicator – a signal that denotes a real command – that precedes the instruction. With that in mind, it is possible to detect patterns and create a simple algorithm. Rober stated, “If you’re trying to decode signs, this simple [algorithm] should work 90% of the time.” The other 10%, which is any combination other than “indicator + sign”, can be cracked by machine learning. With enough training data, machine learning can efficiently sort through the complex data set and extract the right pattern from the irrelevant information. To be clear, neither we (nor Mark himself) endorse the use of machine learning to steal signs for the purpose of cheating, but it is a pretty cool demonstration of AI’s capabilities.
Virtual Fans
In order to simulate a realistic game atmosphere to encourage baseball players who are missing the support of their absent fans, Major League ballparks are now being filled with fake people. While some stadiums only have cardboard cutouts in the seats, others have progressed to include computer generated fans. Though these virtual fans only appear in the broadcast and have little effect on the players, organizations like FOX Sports are using them to create a more immersive experience for their viewers watching the game at home. Technologies used to generate these realistic virtual fans include augmented reality, animation, motion capture, and camera tracking, all of which are powered by NVIDIA’s GPUs and AI software. Fun fact: Pixotope, the augmented reality software used here, was previously used to create AR graphics for the Super Bowl and The Weather Channel.
Robotic Fans
Meanwhile across the globe, the SoftBank Hawks in the Nippon Professional Baseball League in Japan have taken an even more futuristic approach, replacing their fans with a legion of cheerleading robots sporting team jerseys, apparel, and colors that sing and dance to the music in the stadium. Though they lack the precision and verve of human cheerleaders, it is still a fun, inventive way to hype up the atmosphere in the empty stadium. (And for Japan to use robots is so on-brand.)
These changes signify only the beginning of a whole new kind of ball game. With the rise of technology, how much of the sport will be different, and how much will remain the same? AI could help us feel closer to the game by transporting fans at home to a full stadium during Covid. On the other hand, it could also make baseball feel more like a simulation, where viewers could control or predict aspects of the sport much like they would in a video game. Regardless of what tech means ultimately for the future of baseball, get your popcorn, crackerjacks, and hotdogs and let’s watch the season unfold!