This project develops a multi-stage framework to support Duke Women’s Soccer recruitment through advanced modeling and a web-service platform. In the first-stage of modeling, detailed game-level player statistics from the Wyscout API are transformed through extensive aggregation and feature selection to form the final set of average, percentage, and versatility metrics that are used to predict players’ impact when on the pitch. These predictions are then used as coefficient priors for ratings, which are derived through a regression and survival modeling approach, that estimate players’ expected goal contribution by regressing expected goals on player indicators and engagement. Observations are weighted by recency to generate detailed player impacts for every game in a player’s career, as well as a separate metric season over season. Finally, these insights are integrated into a web-service platform, whose internal architecture streamlines data access and visualization, providing users with a scalable and actionable tool for evaluating the success of past recruitment efforts and potential players of interest systematically.
Mentor: Leo Biral
