A Dataset to Investigate First-Person Shooter Players

Abstract

Datasets are multi-purpose research tools, enabling researchers to design, develop, and test solutions to classical computer sciences problems and novel research questions. In the gaming domain, however, there are few high-quality datasets providing both: (1) visual gameplay data and (2) additional information about the gameplay, such as user input. As a result, game researchers most of the time have to collect, process, and annotate gameplay data in time-consuming data collection studies themselves. We start closing this gap, by presenting a novel Counter-Strike: Global Offensive dataset. The contributed dataset is a collection of 12 high-skilled players playing Counter-Strike: Global Offensive. We showcase two deep learning-based examples using the presented dataset, demonstrating its versatility.