Inspiration
We all have experience in Sports Analytics, and wanted to combine our skills into a deployable, fully-fledged project.
What it does
GrittyStats is a collection of NHL Win Probability and Passing Influence Models, along with an AI assistant GrittyLLM to teach you about hockey!
How we built it
GrittyStats is built with Python and Streamlit, and deployed via
Challenges we ran into
Time constraints have challenged the team to adapt fast and deliver a finished product. Initially we were planning to use a React project deployed with Vercel to present the data and a neo4j instance to store Passing Influence data, but instead decided to use Streamlit and networkx to work around the time constraints.
Accomplishments that we're proud of
We are proud of our ability to adapt on the fly and develop an interactive way to look at win probability, passing influence and learn about hockey.
What we learned
We learned a lot about Streamlit, using the pickle file format to store sklearn models, and how to use networkx and pyvis to deliver powerful graph visualizations.
What's next for GrittyStats
GrittyStats hopes to improve the accuracy of the Win Probability Models using more in-depth feature engineering and hyperparameter tuning, and to use these WP models in Monte-Carlo simulations for predictive analytics in Hockey.
Log in or sign up for Devpost to join the conversation.