How AI Could Diagnose Pitchers’ Deliveries In Real Time

Pitchers are not always the best at letting their team know when something feels a bit off. They are masters of pain management because they have to be, and sometimes it isn’t easy to discern the line between the normal soreness of wear and tear versus the pain that comes from an injury.

Sometimes the radar gun readings will point out a problem. Other times, a watchful pitching coach can pick up a subtle cue that a pitcher is just a little off. Often, if a pitcher doesn’t say anything, the problem goes undetected. Eventually the pain becomes too much, and the pitcher heads to the injured list.

Nowadays, computerized tracking systems are watching every pitcher’s delivery on every pitch at 300 frames per second. With markerless motion capture systems, teams may get to a point where a program using computerized vision will have artificial intelligence watching every pitch by every pitcher, looking at the pitcher’s biomechanics on every pitch. 

If a pitcher’s delivery starts to fall out of the normal range in any one of dozens of measures at any point in his delivery, the program could flag it and send a warning.

Avoiding false alarms will require a lot of programming and machine learning, but the utility of this eye in the sky is apparent. The Hawk-Eye system and other high-speed cameras could watch every pitch in games and bullpens and other throwing sessions. And it can monitor a pitcher’s shoulder rotation, elbow and knee flexion, release point, timing and many other measures simultaneously pitch after pitch, which could help teams spot delivery flaws before they lead to injuries.