Kiavash Fathi - Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan, Iran
Mehdi Mahdavi - Associate Professor Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan, Iran

Game developers have used different techniques toincrease the entertainment quality of video games for many years.Methods used in conventional hard-coded methods like statemachinetend to create predefined and limited behavior patternsand thus not considered desirable. In the recent years, with thedevelopment and improvement of machine learning algorithms,developers have the tendency to use these novel tools to createdifferent agents in video games. Although, these new methodsexpand the behavior pattern of agents, but there is still a need formeasurements by which the believability of agents and playabilityof games can be determined. In this paper a novel method formeasuring the above-mentioned key issues is introduced. Theproposed method is based on environmental analysis of the gamelevel. Such a method can be used as a benchmark for comparingthe behavior of different agents according to the previouslygathered data from other players and agents.

Agent, Entertainment quality, Driving video game, Signal processing