When statisticians, academicians, and sports enthusiasts with a keen interest in the overlapping fields of sports and statistics get together for the annual New England Symposium on Statistics in Sports on September 28 at Harvard University, Connecticut will have a strong presence.
Invited presenters include Barry Nalebuff, Professor of Management at Yale University’s School of Management for three decades. His presentation “Measuring Competitive Balance Correctly (in Sports)” will be the lead-off of three featured talks at the day-long symposium. The day will also include an expert panel on “The State of Soccer Analytics,” and Connecticut residents delivering formal talks or presenting their research on a range of topics and various sports, , including some research with surprising conclusions.
For example, Philip Maymin, Professor of Analytics and Director of the Master's in Business Analytics at Fairfield University’s Dolan School of Business, will outline research that much of the front-office functionality of professional sports teams can and should be automated.
“That includes decisions about who to draft, who to sign as a free agent, and who to trade for,” Maymin told CT by the Numbers. “It’s all based on machine learning techniques applied to a broad and rich historical dataset that includes college performance, combine measurements, mock drafts, and even the raw text of scouting reports. And, not only can the decisions be automated, the automated decisions are much, much better than human decisions: the difference is worth about a hundred million dollars for the average team.”
His paper suggests that “Brandon Clarke and Grant Williams will be the sleepers of the [2019 NBA] draft.” Clarke was the 21st selection, Williams the 22nd, at the NBA draft held in June.
"We have been running NESSIS since 2007. Our original conception of the conference was to cater to sports statisticians and researchers in New England, but has expanded so that we have a worldwide participation base,” Mark Glickman, Senior Lecturer in Statistics at Harvard University and co-founder of the Symposium, told CT by the Numbers.
“Even so, we find that many of our participants, including our invited presenters, are from Connecticut. This appears to be particularly true for this year's conference where we have quite a few Connecticut presenters, including those from Yale University, Fairfield University, and from ESPN," Glickman added.
Also among the research being shared with attendees: ESPN Analytics’ Neil Johnson will outline a novel use of tracking data in his presentation, “Extracting Player Tracking Data from Video Using Non-Stationary Cameras and a Combination of Computer Vision Techniques.” ESPN Analytics colleague Brian Macdonald will describe an innovation in sports applications with his work, “Analyzing Player Performance in eSports.”
Luke Benz, a 2019 graduate of Yale University and currently a data scientist at Medidata Solutions, Inc., examined “Timeout Value, Strategy and Momentum in NCAA Division 1 Men’s Basketball.” He looked at “whether there is any truth to the notion that timeouts stop opponent momentum by decreasing the rate of opponent scoring or swinging the rate of scoring in favor of the timeout-calling team.”
Noting that it often seems like teams take a timeout when their opponent is on a big scoring run, in hopes of "stopping opponent momentum" he sought to quantify the value of a timeout by comparing net score differentials in equal length time intervals before and after a timeout was taken.
Benz explained to CT by the Numbers that “Overall, I found that timeouts contribute positive points above expectation on average, though by how much decreases over the course of a game.” He also found that “timeouts decrease in value over the course of the first half, reaching a nadir right before half-time, before increasing in value over the course of the 2nd half.” His overall recommendation to coaches, based on his statistical analysis: “be more aggressive in taking timeouts early in the first half.”
Symposium registration is open to the public and now available. Online registration must be completed by September 14. The ticket price is $60 for students, $120 for non-students. The September 28 symposium will be held at Harvard University in the Science Center.
The Nalebuff talk outlines a new model for analysis that is neutral in the impact of season length in determining competitive balance. The results: “the NFL goes from having the most balance to being tied for the least, while MLB becomes the sport with the most balance.” And “the NBA uniquely stands out for having the most predictable results and hence the least amount of full-season competitive balance.”
Three UConn researchers examined whether the “success rate of a basketball shot may be higher at locations in the court where a player takes more shots.” They analyzed the shot intensity of four NBA players during the 2017-18 season, finding that “field goal percentages of these players are significantly positively dependent on their shot intensities, and that different players have different predictors for their field goal percentages.”
Presenting their research during the day will be individuals from Carnegie Mellon University, University of Western Ontario, Harvard University, Duke University, and the University of North Carolina. Additional research papers will be shared by individuals from Columbia University, Syracuse University, University of Tokyo and other higher education institutions.
Following the symposium, methodology-focused papers will be submitted to the Journal of Quantitative Analysis in Sports (JQAS); applied and practitioner-focused papers will be submitted to the Journal of Sports Analytics (JSA).