Zero Play

Teach a computer to play any game.

The zero play library is based on the ideas in the AlphaGo Zero paper and the example Python code in the alpha-zero-general project. The goal of this project is to make a reusable Python library that other projects can build on to make powerful computer opponents for many different board games. An example project that uses this library is Shibumi Games.

It includes a graphical display that you can use to play against the computer opponent or another human.


Installing Zero Play

Even though Zero Play has a graphical display, it is a regular Python package, so you can install it with pip install zero-play. If you haven’t installed Python packages before, read Brett Cannon’s quick-and-dirty guide.

Then run it with the zero_play command.

The default installation generates some errors about bdist_wheel that don’t seem to actually cause any problems. You can either ignore them, or install wheel before installing Zero Play.

pip install wheel
pip install zero-play

Known bug on Ubuntu 20.04:

qt.qpa.plugin: Could not load the Qt platform plugin “xcb” in “” even though it was found.

This is a PySide2 bug that is missing some dependencies. You can work around it by installing those dependencies like this:

sudo apt install libxcb-xinerama0

More Information

If you’d like to help out with the project, or add your own games, see the file in the source code. For all the details, look through the design journal for the project.

Here are some similar projects for inspiration or collaboration:

  • I already mentioned the alpha-zero-general project. It was a big inspiration, but I’m trying to build something that’s easier to add new games to, or use as a library within another project.
  • Galvanise looks interesting. It’s a mix of Python and C++, using Tensorflow. As of 2020, it looks like a single developer, without much documentation. The games are defined with GDL, not Python code.