Go

This is a python implementation of the board game Go. It is compatible with the same board game framework used by ultimate_tictactoe.

The project attempts to correctly implement the basic rules: Area scoring, positional superko, and termination of the game triggered by 2 consecutive passes.

The idea with this project is to enable some experimentation with machine learning inspired by the AlphaGo and AlphaZero projects.

Known Issues

Performance is not great, plenty of opportunities to avoid allocations/copies.
Territory scoring is not supported.

Requirements

Tested with this version of python:

Python 3.7.2

Getting Started

To set up your local environment you should create a virtualenv and install everything into it.

$ virtualenv go

Pip install this repo, either from a local copy,

$ pip install -e go

or from github,

$ pip install git+https://github.com/rtrusso/go

and then install the requirements

$ pip install -r requirements_server.txt
$ pip install -r requirements_player.txt

Please note these currently point to the forked copy of the repositories updated for python3, instead of the original repo created by jbradberry. This will be corrected once the python3 PRs are picked up in the original repos.

To run the server with Go

$ board-serve.py go

Optionally, the server ip address and port number can be added

$ board-serve.py go 0.0.0.0
$ board-serve.py go 0.0.0.0 8000

To connect a client as a human player

$ board-play.py go human
$ board-play.py go human 192.168.1.1 8000   # with ip addr and port

or with the AI player, see jbradberry/mcts (link to python3 update fork)

$ board-play.py go jrb.mcts.uct