Game Feature Validation of a Real-Time Game Space with an eXtended Classifier System

Abstract

The objective of this paper is the proposal of a new approach for the game feature validation of a game space with the eXtended Classifier System (XCS) algorithm. For initial "proof-of-concept" evaluation we used the game space of the Tic-Tac-Toe game, which was placed in a context of real-time characteristics. Evaluation was done with the XCS algorithm without pre-processing internal knowledge programmed (online mode). Evaluation data results were acquired under real-time constraints. No-loss-strategy was implemented for the game, with various scenarios tested to find out if the algorithm has the capability of finding invalid game feature design flaws. Results indicated that the XCS algorithm is able to deliver stable validation of game feature testing results, can be a powerful tool and, therefore, worthy of further research in the Gaming domain. Confirmation of this core concept testing with this type of algorithm (and similar ones) is necessary, since it paves the way for further research in more complex game environments, where the dimension size, the number of aspects, game states and game actions rises intensely.