Undergraduate Students dive into Sports Analytics

Per Tukey, one of the best aspects of being a statistician is getting to play in everyone’s backyard. While this adage was directed to fields like biology, psychology, and economics, statistics has recently found its place within the world of sports. Popularized by the work of Oakland Athletics General Manager Bill Beane as portrayed in the film “Moneyball”, sports analytics involves collecting and analyzing sports statistics to inform decision making on and off the field. As technological advances yield complex data such as motion tracking, professional teams are turning to sports analytics to gain a competitive advantage over their adversaries. Mentored by Dr. Bailey Fosdick and graduate adviser Connor Gibbs, four undergraduate students  tackled some tough questions regarding sports not for a competitive advantage but for the love of the game during the summer of 2020.

Quinn Johnson began the summer with a fundamental question: Can the National Basketball Association (NBA) create a more exciting playoff experience by changing the format of the playoffs? Defining three metrics associated with exciting games and leveraging various classification and regression techniques including support vector machines, elastic net regularization, and neural networks, Quinn found little to no differences between the NBA’s current format and other proposed formats.  Nevertheless, Quinn supports a new playoff format that involves and seeds the top 16 teams based on overall record, regardless of conference, to maximum fairness. Quinn presented this work at the 2020 UConn Sports Analytics Symposium.

Graduating senior Caroline Thomas teamed up with the CSU softball team this summer to develop novel data collection tools which allow CSU Coach Jen Fisher to assess the coach-ability of her players.  Players are often given instruction from the coach while at bat to target certain pitches.  Presumably a team’s ability to succeed depends on players’ abilities to execute the coach’s plan.  By creating a tool to record pitch count, Coach Fisher’s directions, and the result of the at bat, Coach Fisher can identify weaknesses among her team and refine practices accordingly to address these issues. Caroline equipped the team with an RShiny app to record practice or game data and automated summary reports to educate and motivate player improvement.

Data Science major Adam Kiehl sought to explore the validity of football’s ‘defense wins championships’ when applied to hockey. Unlike football, the same players in hockey contribute to offense and defense and the transition between the two is not always clear. Adam leveraged the idea of point shares, the proportion of points earned by offensive and defensive play, to address the question. Within the context of a generalized linear model, Adam found that defensive performance has a disproportionately large impact on the odds of winning the Stanley Cup; however there is little variability in defensive performances suggesting that marked improvements are difficult to attain in the National Hockey League.

Jonathan Olds turned to one of America’s most watched sporting events to inspire his project: the National Collegiate Athletic Association (NCAA) basketball’s March Madness tournament. Each season, 68 teams are selected for bracket play to compete for the year’s championship title. While 32 teams are automatically invited, the other 36 teams are selected at-large by a committee of individuals in a process void of transparency and details. Jonathan explored replicating these decisions through analytics, identifying aspects of the game the committee inherently deems as important. After accurately picking 90% of at-large bids algorithmically using a generalized linear model, Jonathan determined that team strength of schedule and conference strength of schedule are notably important attributes for the committee’s decisions.