AN INTELLIGENT PLAYER

Programming a perfect knowledge game that implements an artificial player that is capable of consistently beating a human player.

In this project, my goal is to program a two-player, perfect information game. This game is required to support players of different types, whether that be a player vs. player, player vs. computer, or computer vs. computer (relevant for AI training). Afterwards, I will build an AI that plays with a simple strategy to serve as a baseline to improve upon. In order to improve my AI player, I will need to develop a method of evaluating the worth of board positions. Key tools I will need to employ are artificial neural networks (ANN) and tree pruning algorithms. After building the AI player, I will finally need to iteratively improve it by generating game data, training the ANN, and experiment with altering the hyperparameters.

SANTORINI

On a 5x5 board, each player will select the locations to place two workers. On a player's turn, they are able to move one of their workers to one of eight adjacent spots, and afterwards place a tile on a adjacent spot relative to their new position, raising it a level. A player may only move up at most one tile, but they can go down as many as they like. Additionally, upon placing the fourth level, a "complete tower" is formed and a dome will block players from occupying that space. The goal is to reach floor three first or prevent your opponent from moving.