Strength and Numbers: Using Historical Data to Create Insights for Stanford Football Sports Performance
Lewis Burik, Mady Duboc, Collin Riccitelli, Justin Reid
The Client: Stanford Football Sports Performance
The mission of the Stanford Sports Performance department is to strengthen student-athletes’ pursuit of championship goals and high achievement throughout their life with a program aimed at enhancing athletic performance, reducing the incidence of injury, and developing mental discipline.
The Problem to Solve
The Sports Performance team wants to develop quantitative benchmarking and predictive modeling for accurate assessment in the following areas: performance projections for players, coach-player feedback, and exercise recommendations. They asked the team to help guide them toward data-driven analysis, evaluation and feedback mechanisms based on the the vast bank of data they currently collect. Specifically, they want to use data to:
- Determine which performance metrics are drivers of on-field success.
- Develop data driven tools to improve feedback mechansims.
Engineering a Solution
Tools and Methodologies
The team intends to deliver a comprehensive report that will empower the Sports Performance staff to make perfectly clear the standards of the program. The team investigated the following areas:
- Quartile analysis: This analysis uses games played as the success output.
- Feature selection analysis: This analysis uses on-field success to see if a given player made it to the NFL.
- Regularized predictive model: This model's output is whether a given player is predicted to be a starter or non-starter in the NFL using Lasso regression analysis.
Recommendations and Deliverables
The team delivered tools that are easily accessible for all of Stanford Football:
- Advice to coaches for data-driven benchmarking practices.
- Website where players can input their data and immediately see their quartile breakdown, as well as if they are projected to be a starter or non-starter.