Inspiration
My curiosity and love of statistics drove me to create a tool that could answer the age-old question: "What if Team X from Year Y played Team Z from Year W?" The project combines advanced statistics, machine learning, and real-time simulation to bring historical matchups to life.
What it does
The MLB Historical Matchup Simulator creates pitch-by-pitch baseball games between any MLB teams from 1920 to present. It:
- Normalizes statistics across different eras for fair matchups
- Simulates realistic pitch sequences and outcomes
- Tracks detailed game states including baserunners and fielding
- Generates AI-powered game analysis and commentary
- Provides comprehensive post-game statistics and highlights
How we built it
The simulator was developed using:
- Python/Flask backend for game logic and statistical processing
- Socket.IO for real-time updates
- Hosted via Google Cloud Run
- Google Gemini API for natural language generation
- Advanced statistical models for era normalization
- MLB Stats API for historical data
- Custom algorithms for:
- Exit velocity calculations
- Hit distance estimation
- Pitch selection and outcomes
- Defensive positioning
Challenges we ran into
- Normalizing statistics across vastly different eras of baseball
- Creating realistic pitch sequences that reflect historical tendencies
- Implementing real-time game state management
- Balancing simulation speed with detail
- Handling edge cases in baserunning and scoring situations
- Developing accurate physics models for hit outcomes
Accomplishments that we're proud of
- Created a statistically accurate simulation engine
- Successfully normalized player performance across 100+ years
- Implemented realistic physics-based hit outcomes
- Developed natural language game summaries
- Built an engaging real-time interface
- Maintained historical accuracy while ensuring entertaining gameplay
What we learned
- Advanced baseball statistics and sabermetrics
- Real-time web application architecture
- Historical data normalization techniques
- Physics modeling for sports simulation
- Natural language processing integration
- Complex state management in simulations
What's next for Historic Matchup Simulator
- Season simulation mode
- Custom team creation
- Historical playoff brackets
- Player career trajectory modeling
- Advanced weather effects
Log in or sign up for Devpost to join the conversation.