Abstract- The Virus Game (or simply Virus) is a turnbased two player perfect information game which is based on the growth and spread of competing viruses. This paper describes a CPU efficient and easy to use architecture for developing and testing AI for Virus and similar games and for running a tournament between AI players. We investigate move generation, board representation and tree search for the Virus Game and discuss a range of parameters for evaluating the likely winner from a given position. We describe the use of our architecture as a tool for teaching AI, and describe some of the AI players developed by students using the architecture. We discuss the relative performance of these players and an effective, generalisable scheme for ranking players based on similar ideas to the Google PageRank method.
Peter I. Cowling