A statistical approach to select the manipulated variables of a difficulty adjustment closed-loop control for an augmented reality video game

Dynamic difficulty adjustment (DDA) has become a popular method for enhancing the gaming experience in both serious and recreational video games. However, most DDAs are designed using heuristic rules, which can lead to suboptimal selection of the variables manipulated to affect the game’s challenge. To address this issue, we propose to perform a non linear open-loop identification of the interaction between the user and the game, in which the user parameterizes the difficulty manually, and the resulting data, along with the performance data from multiple game sessions, are used to train a machine learning model. Then, the predictive power statistical analysis is used to remove manipulated variables with reduced or null effect on players performance. In this paper, we report the experimental results obtained using a first-person shooter game (Invasion Victory) that we developed. Our findings demonstrate that this approach can successfully identify meaningful correlations between manipulated and controlled variables. The game is available for download from the Google Play Store. The proposed approach represents a promising direction for DDA design that incorporates principles of automatic control and statistical learning.