The 2025 Wharton High School Data Science Competition challenged over 490 teams from 34 countries to analyze more than 5,300 NCAA Women’s Basketball games. Students explored real-world data, engineered new features like point differential and possession-adjusted stats, and tackled messy challenges like how to handle incomplete Division II records. 

Strong teams didn’t just build accurate models—they communicated their methods clearly and used basketball logic to ground their decisions. One of the main lessons we hoped students gained from the competition was this: in real-world data science, there’s rarely one “right” answer. That uncertainty, while sometimes frustrating, makes the field so creative, open-ended, and impactful.

Our expert team benchmarked submissions using ensemble models and found that approaches like Elo ratings (especially those with margin-of-victory), regression models with contextual features, and careful data cleaning consistently performed best. The best teams balanced simplicity with strategy, starting with point differential, then layering in strength of schedule, game context, and ensemble modeling.

View the top team presentations and explore the complete Playbook here.