How does a person learn?

By doing the thing it wants to learn!

Famous quote:

If you do something for 1.000 hours, you will become good in it.

Chess (-student) implemented using pygame

This project will be used to train a model using deep reinforcement learning (deep RL) based on opponents play.
Nevertheless, the game can be played in multi-player mode without fancy ML features.

The project will eventually contain the following features (ticked ones implemented)

• [x] complete chess game

• [ ] history log and game loading

• [ ] deep learning reinforced model

• [ ] trained example model

• [ ] payable against trained model

• [ ] can train own model based on your moves!

Interesting problems to analyze

How would the model behave in such situation:

• King on e1, Pawn e2 moves up until e8 and gets promoted into Rook
  • Enable underpromoted Rook castling using Game(underpromoted_castling=True)

Usage:

clone the project

git clone git@github.com:philsupertramp/chess

Install requirements

pip install -r requirements.txt

Run the game in 2P mode

python main.py

Run the game in headless console mode (no dependencies required)

python main.py --headless

Development

clone the repository and install the dependencies using requirements-dev.txt.

Note:

• Do not, and I repeat not use Pawn-d4 as a first step for testing. This pawn is on the (4,4)-tile and will always work!

Demo

https://user-images.githubusercontent.com/9550040/126046245-74f7ed12-d695-4153-b909-2535f48e3706.mp4