A machine learning repository for efficiently and accurately classifying MLB pitch types given pitch data.
- Introduction
- Project Structure
- Installation
- Usage
- Data
- Experiments
- Results
- Contributing
- License
- Acknowledgments
This project was first created as a final project with Alex Rimerman for an undergrad Machine Learning course at Swarthmore College. Now, it is maintained as time allows out of interest in the subject. The inspiration for the original project came from the use of technology in modern baseball. Specifically, on the scoreboard at any professional baseball game, the pitch type is displayed immediately after the pitch is thrown, far too quick for human entry. So, we wanted to replicate this process as best as we could (at least for what time allows in a 5 week final project) by implementing machine learning models to predict pitch type based on pitch metrics. Currently, the best model works with better than 90% accuracy and runtime on the order of milliseconds.
Data can be scraped using pybaseball. Will include more instructions for how to in the future.
If you have any thoughts or ideas, please feel free to submit an Issue or a PR. I'm always looking to improve the project and meet new developers.
Thank you to Professor Ben Mitchell at Swarthmore College for his guidance and expertise in the project.