AI Competition Framework

This is the website and game that was built by me for NeoDeerhunt. NeoDeerhunt was an overnight competition where teams of three competed to build the best bot for the given game. A public leaderboard was displayed and participants could challenge people on the leaderboard to take their position and knock everybody below them down.

The game was loosely inspired by starcraft. It has the following features:

• Turn based
• Horizontally symmetrical 2D grid maps
  • Contains walls and resources
• Two types of units per player
  • Worker units
    • Can mine resources
  • Fighter units
    • Can attack other units
    • Can spend resources to duplicate

The website was written in flask with react.js on the front end. It uses mongodb for storage and code submissions are run inside of a docker container. When a submission is made a new docker container is built and run. The replay and results of the match are extracted through docker logs.

The itself game was written in python and uses a simple server (to control the game state) and two simple clients (one for each player). The client was designed to be easy for people to understand and implement. Classes that manage the connection to the server and provide useful utilities are provided so the player can focus on implementing their strategy for the game.

The README s in each components folder provide more technical information on how to set up and run this for yourself.