This project explores the impact of the Name, Image, and Likeness (NIL) policy implemented in July 2021 on NCAA athletes, particularly basketball players in the U.S. and Canada.
Using causal inference and clustering analysis, we aimed to uncover:
- How NIL changed athlete performance
- Whether different player styles responded differently
- What this means for branding and recruitment strategies
Since July 2021, NCAA athletes in the U.S. can earn income via NIL deals (e.g., sponsorships, social media). However, school payments based on performance remain restricted, and NIL policy varies across states.
To study NIL's influence, we used Canadian players as a control group (who were not affected by the U.S. policy).
- Has NIL affected on-court behavior of athletes?
- Can we quantify that effect with performance metrics?
- Do player types respond differently to NIL incentives?
Result:
U.S. players saw a 7% increase in Game Score relative to Canadian peers post-NIL.
- GmSc (Game Score) – holistic individual impact measure
- Box Plus-Minus (BPM) – team-adjusted performance estimate
- Fouls, turnovers, points, win shares, and more
We segmented 500+ players into 4 behavior-based clusters:
| Cluster Type | Traits |
|---|---|
| 🟢 Support Role Players | Low scoring, high defense, team-first |
| 🔴 Volume Scorers | High usage, solo impact |
| 🔵 Efficient Team Players | High BPM, disciplined & efficient |
| 🟣 All-Around Performers | Balanced offense & defense |
- +28% GmSc in All-Around performers
- +204% BPM in Team-first players
- +25% Points in Volume scorers
- -19.8% Fouls in disciplined defenders
-
Use Advanced Metrics Early
Spot undervalued athletes before their market value spikes. -
Prioritize Versatility & Team Fit
All-around or support players show consistent performance gains. -
Segment for Smart Branding
Scorers suit high-flash sponsorships; defenders suit leadership narratives.
- PowerBI & Excel for dashboarding
- Python for clustering (KMeans) and causal modeling (statsmodels)
- NCAA & U SPORTS data for U.S. and Canadian athletes
- DiD estimation via pre/post-performance comparison
- NIL did not uniformly improve performance — its effect depended on player type.
- Team-oriented players became more efficient, while scorers pushed harder.
- A player’s style shift can be an early signal of growth, discipline, or marketability.
📍 University of Minnesota - Carlson Analytics Lab
📁 Course: Exploratory Data Analytics
📆 Date: Spring 2025
🧑💼 Role: Data Science Lead
Made by Justin Varghese
Let’s connect if you're interested in sports analytics, policy impact evaluation, or causal ML.