LeCoach
Agentic basketball analytics platform
Overview
LeCoach is a comprehensive analytics platform for coaches and players in Canadian usports basketball. It features an agent system for game analysis and predictive analytics powered by 10+ years of historical league data.
Motivations
Over the past year, I've been leading and expanding the data solutions team for the Waterloo varsity basketball program. The critical challenge we've identified is the lack of a centralized, user-first platform for performance analysis. While Synergy Sports is our current solution, the platform is overwhelmingly complex and cluttered. LeCoach would be a platform that bridges the gap between raw data and basketball-first minded coaches + players.
User Flow
Games Navigation
▼Users can navigate through all games dating back to the 2005-06 season.
Analysis Interface
▼The win probability graph is powered by a model trained on 1.4 million historical plays. More team metrics will be available soon (traditional stats + VORP, BPM, etc.)
Play by play tracking
▼View detailed play by play tracking for every game.
Agent chat
▼Our chat system uses RAG and chains multiple specialized agents together to handle different aspects of data retrieval, analysis, and response generation.
System Diagram
The system architecture for LeCoach integrates data collection, processing, and presentation layers to deliver real-time analytics and insights.
Design
LeCoach's design prioritizes simplicity and intuitive navigation, with bold colors against a monochromatic background. The interface employs thoughtful information hierarchy to ensure key metrics stand out, allowing coaches to make data-driven decisions quickly. I was inspired by modern data apps such as Real Sports and Bevel.
Technical Details
Frontend: Next.js, Tailwind CSS, Recharts
Backend: Python, FastAPI, Almebic, Docker
Cloud: S3, RDS, Lambda, Sagemaker
Other: Langchain, OpenAI, Scikit-learn