Using Reinforcement Learning to Learn Seega Board Game

Abstract

Computer games are one of the popular fields which use machine learning and artificial intelligence. Many of these games have a large search space, which makes them too vast for even supercomputers to brute force. Seega is an ancient Egyptian two-player board game similar to chess but has more difficult rules. It has two stages, In the first stage, the players position their stones on the board in strategic manner. In the second stage, the players move their stones vertically or horizontal to capture the opponent stones. In this work, we used deep reinforcement learning to train two co-operative agents to master the game of Seega and learn the game rules. We compared our proposed approach with the classical mini-max algorithm. We found that our approach is much more practical to be adopted to be used in Seega in terms of computational time and needed resources.