I started on this project for an Artificial Intelligence class I took in University. If you don't know what the Game of the Amazons is, you can check it out here: http://en.wikipedia.org/wiki/Game_of_the_amazons It turned out to be a really fun project, we ended up with a few unique features and winning first place to boot.

It's been a while between last looking at the project and putting it up on GitHub, but here is what I remember!

Algorithmic Details:

• Concurrent MinMax search with Alpha-Beta pruning for non-closed games.
• Concurrent State Space Search for games that become closed (player and opponent's pieces can not interact with one another).
• Action based searches, as opposed to state base searches.
  • Instead of generating a game tree of successive boards, compact actions are generated instead (an action is that of moving a queen to a square and firing an arrow).
  • Thus, states are stored as an initial board and action chain.
  • This method greatly reduces the required resources to search a game tree.
  • The MinMax and State Space searches operate by applying and undoing actions.
  • Accordingly, the Successor Function produces all possible player actions for a given state, as opposed to all possible end states.
• The Evaluation Function is divided into two stages: early game and end game.
  • Early game algorithm evaluates a Queen's 'limited mobility' for one and two-hop jumps.
  • Early game algorithm also weights a Queen's escape routes.
  • Late game is based on a Territory-Mobility (TM) algorithm.
  • Territory (how close one's piece is to a square) is weighed against Mobility (how many moves it takes for one's piece to get to a square).
  • Of course, blocked Queens are also taken into consideration. This AI will play aggressively for them.
• Cutoff test is a simple memory and time check - moves must be completed under 30 seconds.

Other Fun Facts:

• Impressive interactive GUI!
• Customizable runtime configurations.
• Networked HvH, AIvAI, and AIvH action.

Because this project was developed for the sole purpose of the course, it doesn't have support for offline, Human vs AI games. For the competition, everything was routed through a SmartFox server hosted somewhere in my Professors office. Chances are, that server is long since dead, meaning this project is best used as a code sample or to inspire future projects. I have plans in the future to whip out a build that enables you to play local games against the AI.