EEmGuzman / ChineseCheckersAgent

A Chinese Checkers agent based on heuristics, MCTS, and self-play reinforcement learning.

Chinese Checkers Agent

• This is the repository for a Chinese Checkers agent trained with heuristics-guided Reinforcement Learning and Monte Carlo Tree Search.
• Clone the repository by running git clone https://github.com/kenziyuliu/PythonChineseCheckers.git in your terminal

Try for yourself:

1. Run python3 human_vs_ai.py <model path> to play against a pre-trained model. We have included two models for you to play against: good_model.h5 and good_model2.h5. Depending on your style of playing, you may find one model stronger than the other.

To play normal human vs human matches:

1. Run python3 human_vs_human.py

To play against a simple greedy player:

1. Run python3 human_vs_greedy.py

Some dependencies:

• Keras >= 2.1.6
• TensorFlow >= 1.6.0
• Python >= 3.6.5
• NumPy >= 1.14.3
• h5py >= 2.7.1