For the majority of its history, baseball has been widely embraced as a game of failure. In fact, many hitters are enshrined in the Baseball Hall of Fame having failed a staggering 68% of the time or more in professional plate appearances (this would translate to a .320 career batting average). This is what makes the game so beautiful in my opinion; finding the silver lining amidst a world of shortcoming. In no other sport would an individual be deemed a success with the percentages hitters face in baseball. A QB with a 32% completion percentage? No way. A point guard who shoots 32% from the field? Hello free agency. Failure is, and will forever be, a part of baseball….or will it?
Enter advanced analytics and Ray Hensberger, a baseball enthusiast and strategic innovator. Hensberger has been bold, and brilliant, enough to attempt to alter baseball’s failure rate. This entails pitchers not having the upper hand in baseball’s future. At a 2014 conference, Hensberger shared his innovative data crunching and the academic paper his team produced for the MIT Sloan Sports Analytics Conference. His team modeled MLB data to show with 74.5% accuracy what pitch a pitcher is most likely to select on a given count. Sounds ludicrous right? Actually, it’s quite the contrary.
Hensberger’s calculations are revolutionary; more accurate than any pitching analytics to date. How did they come this far? Hensberger and his team started with 900 pitchers on MLB rosters, and then excluded player who threw less than 1,000 pitches over a three season window. This drew an experimental sample of about 400 arms for his model to evaluate. Variables like the matchup of batter vs. pitcher (Righty/Lefty), current at-bat (pitch type and zone history, ball-strike count), game situation (inning, number of outs, and number and location of men on base); as well as other features from observations of pitchers varying across a span of games, such as curveball release point, fastball velocity, general pitch selection, and slider movement.
The result was essentially a set of pitcher-specific models as well as a detailed report about what those pitchers would throw in in-game situations.
The ‘Field’ Test
Hensberger and his team ran this model over previously unseen games from the 2013 World Series featuring the Boston Red Sox and the St. Louis Cardinals. The result was a staggering 74.5% pitch prediction efficiency. Being a former college pitcher, this terrifies me. However, I see how it can benefit the game as well. Adding runs usually adds excitement and attendance. While this process is clearly a ways away from being implemented (and I wouldn’t be surprised if the MLB instituted policies against it to maintain the human element of the game), it is no doubt an astounding testament as to how valuable a concise understanding of big data is. There’s a lot of power to be harnessed, and we are finally tipping the iceberg.
Want more on this topic? Check out the link below featuring a lecture on pitch prediction from the 2012 Sloan Sports Analytics Conference. This ‘throwback’ shows how they have definitely been onto something for quite some time.
Drive Home Safely,
Bryan White, MBA