Project Background
What makes one tennis shot better than another?
For decades, the sport of tennis has used unforced errors and winners to label poor shots from the best shots in the game. But did you know that these classifications are still done by humans?!
With modern tracking data, the sport is in a position to modernize the way it classifies error using machine learning approaches that could be more cost-effective and reliable than human approaches.
Project Goals
The student will use tracking data and human labels of errors and winners to build a machine learning classifier for each type of point outcome.
Specific Activities
- Get familiar with tennis tracking data
- Merge labels from IBM Slamtracker with tracking data of shots
- Evaluate the accuracy of different machine learning classifiers
- Use graphics and performance metrics to summarise findings
- Create report and presentation of findings
Requirements
This project is intended for students with experience with R, some familiarity with machine learning, and a love of sports.
Supervisors
The project will be supervised by Professor Di Cook (@dicook) from Monash University and Stephanie Kovalchik (@skoval), a tennis data scientist at Tennis Australia.
Contact
For inquiries and questions, please contact dicook@monash.edu or skovalchik@tennis.com.au