Quantifying congestion with player tracking data in Australian football

Area of Science:

• Sports Science
• Biomechanics
• Data Science

Background:

• Congestion is a key factor in Australian football, influencing passing, fan experience, and rule evaluations.
• Objective measurement of on-field congestion is currently lacking.

Purpose of the Study:

• To develop and validate objective methods for quantifying player congestion in Australian football.
• To analyze the impact of player proximity on ball disposal during matches.

Main Methods:

• Developed a density-based clustering approach to categorize player proximity (primary, secondary, outside).
• Utilized the Random Forest algorithm to classify player-experienced congestion (high, nearby, low) during ball disposals.
• Employed player tracking and match event data from the Australian Football League (AFL) seasons 2019 and 2021.

Main Results:

• The Random Forest model achieved high accuracy in classifying high (0.89 precision, 0.86 recall) and low congestion (0.98 precision, 0.86 recall) disposals.
• The model demonstrated strong performance across various contextual variables like field position and game quarter.
• Both developed methods offer efficient, automated quantification of congestion, reducing reliance on manual coding.

Conclusions:

• The study successfully introduced objective, data-driven methods to measure Australian football congestion.
• These techniques provide a foundation for future research comparing congestion across different seasons, rounds, and teams.
• Automated congestion analysis enhances the understanding of game dynamics and potential impacts of rule changes.