Deep convolutional player modeling on log and level data

Abstract

We present a novel approach to player modeling based on a convolutional neural net trained on game event logs. We test our approach and a hybrid extension over two distinct games, a clone of Super Mario Bros. and Gwario, a human computation version of Super Mario Bros.: The Lost Levels. We demonstrate high accuracy in predicting a variety of measures of player experience across these two games. Further we present evidence that our technique derives quality design knowledge and demonstrate the ability to build a more general model.