Champion is a reinforcement learning model intended to optimally play Spy Cards from Bug Fables: The Everlasting Sapling.
The goal for Champion is to be able to provide a sufficient challenge for players, without cheating, that other decision-making bots have failed to accomplish.
Run the following in a terminal or command prompt. It will install the git to the directory specified in the terminal/command prompt.
git clone https://github.com/FoxtrotOnce/Champion.git
Run the model by running PPO.py, and view the data in GraphPPO.py.
- Create a Spy Cards environment accurate to Bug Fables for the model to interact with.
- Set up modeling framework and extra classes/functions for testing.
- Get >80% win-rate on a playing model using a Leafbugs deck VS. Bug Fables bot using a Thugs deck.
- Get >80% win-rate on a playing model using any meta deck VS. Bug Fables bot using any meta deck.
- Change modeling framework to make the playing model work with a model that predicts the opponent's move.
- Get >80% win-rate with the combined models using any meta deck VS. Bug Fables bot using any meta deck.
- Get data from the Spy Cards Tournament to use for supervised learning of the opponent prediction model, to predict human players.
- Get >80% win-rate with the combined models using any deck VS. a human player using any deck.
Champion is an open-sourced software licensed under the MIT license.