Written By: Jason Baik
On April 4th, 2017, the American Statistical Association held a Spring Banquet in the Wyndham Pittsburgh University Center. There was a 20-minute keynote panel of statistics in sports consisting of Bill Benter, Dan Fox, and Peter Carr. Here is a brief biography of the panelists and a summary of their presentations:
Bill Benter is a fellow of the Royal Statistical Society and the founder of Avenue Four Analytics, where he builds statistical models for horse racing markets around the world.
Dan Fox is the director of Baseball Informatics for the Pittsburgh Pirates. He joined the organization in 2008 and has “consulted on nearly every player acquisition decision, supplied the data behind defensive shifts, played a role in recommending a focus on increasing groundball rates, and conducted research on the draft and preventative health practices for the pitchers.” (Trib Live)
- The Pittsburgh Pirates have analysts “embedded” in the team. This means they always have an analyst traveling with them, giving the team personal insights throughout the season. Having an analyst always with the team has been more effective for the Pirates than relying on an analytics team based in Pittsburgh while the team is traveling.
- The Pittsburgh Pirates also have analysts “embedded” in minor league affiliates. These analysts, along with the ones at the major league level, come together to discuss trade acquisitions and draft picks for the team.
- A hitter’s vision, hand-eye coordination, and strength predict his swing rate and chase rate. These metrics ultimately indicate strikeouts, walks, hits, and speed.
Peter Carr is a Research Scientist at Disney Research. He has been investigating algorithms for generating sports broadcasts automatically.
- He found that sports teams at the bottom of the standings performed worse than his model and those at the top of the standings performed better. Interestingly, the top teams deviated from the model and used unique methods to achieve their success. Ultimately, these methods turned out to outperform Carr’s model.
Then, the panelists answered questions from the audience.
One question: Can you tell me a time when something did not work well on the job?
Bill Benter tried to bet on baseball plus and minus. His company went to great efforts to acquire weather data. He discovered that the number of runs scored increased when there was less air density. Fortunately, the world of betting did not catch on to this discovery. Thus, he formed a betting strategy that took advantage of this discovery. Ultimately, it failed, and Benter stopped betting on baseball.
Dan Fox: “Every time a baseball defense shift based on analytics does not work, it is a lot more painful than when it worked. Naturally, players only remember times it doesn’t work. Overall, it works more than it does not.”
Peter Carr developed algorithms that allowed for live tracking and tested them in Carnegie Mellon basketball games. After the experiment succeeded, Carr went to Disney’s Orlando Sports Complex and tested it in a national high school all-star basketball team. Unfortunately, his machines could not follow the speed the all-stars were playing at.
Bill Benter: “It is amazing to see that a computer model outperforms human decision and public accuracy.”
Dan Fox: “In the Pirates, buy-in by the coaches has gotten better. The Pirates are looking for people to assimilate analytic information. Information is given to team before each series.” (50 series in a year)