EAST LANSING – The World Series is around the corner and fans are anticipating the first pitch. But what exactly determines what that pitch will be?
What if we could add some science to this art? What if we could quantify a batter’s strengths and weaknesses and determine just how effective he is against each pitch in each situation? Could the pitcher use that information to choose a pitch to maximize the chance of an out?
Supported Intelligence has developed a model aimed at doing just that. Their model, dubbed “SmartPitcher,” recommends a pitch based on the batter and the count designed to maximize the chance of an out during the at-bat.
And just how is this possible? In 2007 the MLB installed a pitch tracking system in all of its thirty major league stadiums. PITCHf/x uses three cameras in each stadium to track the trajectory of the ball from the moment it leaves the pitchers hand to the moment it crosses home plate. If you’ve seen a pitch trajectory on television, it was likely generated using PITCHf/x data.
PITCHf/x is inspiring a true revolution in baseball analytics — and SmartPitcher is the next step. SmartPitcher uses data from the last three MLB seasons (over 1 million pitches!) to determine the likelihood of each pitch resulting in a called ball, a strike, the batter reaching a base, or the batter being called out. SmartPitcher feeds those likelihoods into a recursive model which determines the pitch to maximize the chances of an out during the at-bat.
Supported Intelligence will tweet pitch recommendations throughout the 2015 World Series next week with SmartPitcher’s own twitter handle: @SmartPitcher. We can only wait and see if SmartPitcher can improve on the now century old art of pitching.
This story was written by Neal Anderson, Lead Developer, Supported Intelligence. To reach Neal, Mailto:[email protected]
