Custom AI chess engines trained with self-play and heuristics — compete online or explore the models.

• Multiple AI engines with different difficulty levels and evaluation strategies
• Deterministic and stochastic variants for varied gameplay
• Modular design for tweaking evaluation weights or search depth
• Web interface for live gameplay
• Clear structure for students, hobbyists, and researchers

• Frontend: Typescript, React, Next.js
• Backend: Python (self-play engines, evaluation heuristics)
• Search Techniques: Minimax, alpha-beta pruning, iterative deepening
• Training: Self-play loops, handcrafted evaluation functions, iterative tuning
• Deployment: Docker and Docker Compose for easy setup

No Python, Node.js, pip, or npm required. Docker Desktop is sufficient.

git clone https://github.com/maxtmiller/AlphaPoisson.git
cd AlphaPoisson
docker compose build
docker compose up

Open your browser at: http://localhost:3000

• Frontend: https://alpha-poisson.vercel.app/
• Backend (spin up before playing): https://alphapoisson.onrender.com/

• Experiment with new evaluation functions and heuristics
• Add engine rating system for ranking difficulty levels
• Explore reinforcement learning for adaptive AI