This book is aimed at giving readers an introduction to the practical side of game data science and thus can be used a textbook for game analytics or game user research class or as a reference to self learners and enthusiasts. Game data science is a term that we use to denote a process composed of methods and techniques by which an analyst or a data scientist can make sense of data to allow decision makers in a game company to make informed decisions. This process involves: statistical analysis, visualization, abstraction of low-level data, machine learning and sequence data modeling. The book introduces different methods borrowing from different fields including human computer interaction, machine learning, and data science, focusing on methods and techniques used by both industry and researchers within the field of games. The book examples and case studies specifically focus on gameplay log data. The book takes a practical stance on the subject by discussing theoretical foundation, practical approaches, and delves deeply into the different techniques proposed and used through labs, examples, and comprehensive surveys of various case studies from both industry and academia. Topics range from simple approaches to more advanced ones. No prior knowledge is required. The book is developed to be self contained and can be used as a good way to introduce the reader to data science and how it is applied to the filed of games.
19 August 2022 in ACM Transactions on Accessible Computing
Developing games is time-consuming and costly. Overly clinical therapy games run the risk of being boring, which defeats the purpose of using games to motivate healing in the first place [ 10 , 23 ]. In this work, we adapt and repurpose an existing immersive virtual reality (iVR) game, Spellcasters, originally designed purely for entertainment for use as a stroke rehabilitation game—which is particularly relevant in the wake of COVID-19, where telehealth solutions are increasingly needed [ 4 ]. In preparation for participatory design sessions with stroke survivors, we collaborate with 14 medical professionals to ensure Spellcasters is safe and therapeutically valid for clinical adoption. We present our novel VR sandbox implementation that allows medical professionals to customize appropriate gestures and interactions for each patient’s unique needs. Additionally, we share a co-designed companion app prototype based on clinicians’ preferred data reporting mechanisms for telehealth. We discuss insights about adapting and repurposing entertainment games as serious games for health, features that clinicians value, and the potential broader impacts of applications like Spellcasters for stroke management.
Video game analytics are widely adopted within academia and industrial development – yet, the player’s perspective is rarely considered as a driving factor, even though vast communities are interested in self-regulated learning, cross-comparisons and empirical trends or insights. In this work, we propose to empower the role of the player as the central impulse for analytics, collect requirements and data from the community of one of the most popular MMORPGs (Guild Wars 2), establish a tool over a 18-month period of participatory development and discuss the co-created features. With analytics for in-game logs from (n = 175, 099) unique players and atomic game actions of over 2 million hours played, we contribute a large-scale, long-term iterative implementation and evaluation of a player-driven instrument to quantify popularity, viability and hierarchical inspection between classes or individual performances. This tool found frequent usage among the actual game community, delivering game data science to the very player.
Educational technology is shifting toward facilitating personalized learning. Such personalization, however, requires a detailed understanding of students’ problem-solving processes. Sequence analysis (SA) is a promising approach to gaining granular insights into student problem solving; however, existing techniques are difficult to interpret because they offer little room for human input in the analysis process. Ultimately, in a learning context, a human stakeholder makes the decisions, so they should be able to drive the analysis process. In this paper, we present a human-in-the-loop approach to SA that uses visualization to allow a stakeholder to better understand both the data and the algorithm. We illustrate the method with a case study in the context of a learning game called Parallel. Results reveal six groups of students organized based on their problem-solving patterns and highlight individual differences within each group. We compare the results to a state-of-the-art method run with the same data and discuss the benefits of our method and the implications of this work.