During my poster presentation at last Summer’s Joint Statistics Meeting (JSM) 2023 in Toronto, I was invited by a representative from the SCORE network to attend a session hosted by the American Statistical Association (ASA) Section of Sports Statistics. The speakers introduced the intriguing initiative of the SCORE network, a collaborative effort between academia and the sports industry aimed at developing and disseminating modules with sports-themed content for statistics and data science courses. Funded by the National Science Foundation (NSF), the project seeks to establish a sustainable national network focused on enhancing data science education through sports analytics.
Upon exploring the SCORE network's website, I discovered a diverse array of modules spanning baseball, football, lacrosse, marathons, triathlons, and even League of Legends. Intrigued by the prospect of contributing, I immediately asked whether I can contribute a module based on FIRST Robotics Competition (FRC) and received warm encouragements from the speakers.
Being a member of my high school’s FRC Team 341, Miss Daisy, has been a big part of my high school experience. I love it because FRC combines the excitement of sports with the rigors of science and technology. Because of my previous data science project, I was also familiar with FRC data sources. That was how I came up with the proposal to develop a module centered around the FRC data to teach some interesting statistics concepts.
Delving into the development process last December, I dug into my favorite data sources for FRC like The Blue Alliance and Statbotics. One of the key features of FRC is that the robot/team competes not individually, but in alliance with other teams. Because of this, it is important for teams to “scout” other teams as potential alliance partners. To aid in this process, various methodologies/models to evaluate each team’s potential contribution have been developed over the years. One of the most popular models is the Expected Points Added (EPA) model that produces predicted probabilities of winning for the alliance based on the past performances of each team in the alliance, as well as teams in the opposition alliance.
As such, there is a desire/need to assess how good the EPA model prediction is. Brier score is a method to evaluate the predicted probabilities against the actual outcomes. Originating with weather forecast research, it’s useful to assess the EPA predictions against actual FRC results.
The mission of the SCORE project resonates deeply with my passion for STEM education. Excited by the prospect of contributing to the SCORE network as a high school student, I recently wrote an outline for my module and contacted the SCORE network. Hopefully I can get some detailed guidance soon on how to complete my module to teach Brier score.