A Method for Using Player Tracking Data in Basketball to Learn Player Skills and Predict Team Performance

Area of Science:

• Sports Analytics
• Network Science
• Statistical Modeling

Background:

• Player tracking data offers unprecedented numerical insights into basketball performance.
• Traditional analysis often overlooks the complex interactions between players within offensive systems.

Purpose of the Study:

• To develop a method for automatically learning player skills using player tracking data.
• To predict the performance of novel five-man lineups by considering player skill interactions.
• To create a network-style model linking individual skills to team offensive success.

Main Methods:

• Coupling player tracking data with a network model of offensive plays.
• Implementing a simplified network model for analysis.
• Utilizing statistical inference schemes for skill estimation.
• Evaluating the model with simulated and real game data.

Main Results:

• Player skills can be accurately inferred even with limited data.
• The model effectively predicts lineup performance by accounting for player skill interactions.
• Player interactions within lineups are consistently described, even across different team compositions.

Conclusions:

• Player tracking data, combined with network modeling, provides a powerful tool for basketball analytics.
• This approach enables accurate skill assessment and reliable prediction of team performance.
• The model demonstrates robustness in analyzing player contributions and lineup dynamics.