NeymarL/ChineseChess-AlphaZero

Health v2 · maintenance only

Adoption not part of the score

1242 stars · 363 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

An open-source implementation of AlphaZero/AlphaGo Zero reinforcement learning methods applied to Chinese chess (Xiangqi). It includes self-play, training, evaluation, and supervised learning workers, plus a built-in pygame GUI for playing against the trained AI.

Use cases

• train a Chinese chess AI with AlphaZero methods
• run self-play to generate reinforcement learning training data
• play Xiangqi against a neural network with MCTS
• learn how AlphaZero works by studying a full implementation
• train a board game AI from scratch without human game data
• evaluate new model generations against a best model

When to choose

• you want to build or study an AlphaZero-style AI for Chinese chess
• you need a reference implementation of self-play, MCTS, and policy/value network training
• you want to contribute compute to a distributed Chinese chess training project

When to avoid

• you need a production-strength Xiangqi engine for competitive play
• you work with modern TensorFlow/PyTorch versions, since the stack targets old Python 3.6 and TensorFlow 1.x
• you want AlphaZero for a different game like Go or chess without porting

Facets

library · maturity maintenance

reinforcement-learning deep-learning machine-learning game-engine reinforcement-learning deep-learning artificial-intelligence python cross-platform alphazero chinese-chess mcts self-play tensorflow keras board-games game-development

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NeymarL/ChineseChess-AlphaZero")