Keynotes

Gaming Focused Analytics

Adam Beame

Head of Data Analytics and Machine Learning at Hi-Rez Studios

Summary: The content of my presentation will focus on the role of Analytics at Hi-Rez Studios, how we expanded our analytics expertise, and common key decisions teams can run into when growing a gaming focused analytics discipline.

Bio: Adam Beame has been focused on empowering cross discipline game studios, eSports production companies, and publishers to leverage their big data as a revenue generating force multiplier. He currently leads the Hi-Rez Studios Data Analytics and Machine Learning team which has grown under his direction for the past 3 years. His team is currently focused on detecting and reducing toxicity, helping game designers find meaning in the data storm, and providing a leg up to our player outreach teams.

Esports: A new Domain to Explore

Simon Demediuk

Games Data Analyst at the University of York

Summary: When it comes to game analytics a large amount of work has gone into profiling players in order to either improve game development or predict churn for game developers. With the emergence of esports, it's growing popularity and data availability, the stakeholders for the results of game analytics has changed. In this domain we can use the vast amounts of player data in new and interesting ways for various stakeholders, from regular players to teams to broadcasters. In this talk I will delve into what research we are doing for the largest esports analytics project Weavr and explore how esports is changing the game analytics space.

Bio: Simon Demediuk is a Games Data Analyst, working at the University of York. Simon studied his PhD at the Royal Melbourne Institute of Technology in Australia, where he specialised in Player Profiling and Artifical Intelligence for games. Since then he has focused his research in player profiling, prediction models and churn analysis. Currently, he is working on a newly formed Government-funded audience of the future project called Weavr. This project aims to turn professional behavioural telemetry data, into hyper-personalised viewing experiences for esport audiences.

Tutorial

Matrix Factorization for Descriptive and Predictive Behavioral Analytics in Games

Rafet Sifa

Head of Cognitive Business Optimization at Fraunhofer IAIS

Summary: In this tutorial we will familiarize ourselves to the notion of matrix factorization and its use in descriptive and predictive analytics in the context of games. We will revise the theoretical basis behind factorizing data matrices by covering the low rank representations, the popular parameter optimization methods, the constrained factorization models and the relationships to well-known data science models. Following that we will present different application ideas and show a case study based on analyzing a dataset from a freemium mobile game, in which we will show how the learned representations can be used to better dissect the player behavior and predict player retention.

Bio: Rafet Sifa is a co-organizer of GAW'19. He is the head of Cognitive Business Optimization unit and a senior research data scientist at Fraunhofer Institute for Intelligent Analysis and Information Systems (IAIS). His current research focus is based on statistical data mining in the context of interpretable and informative representation learning methods. His PhD research from University of Bonn was about devising dedicated interpretable matrix and tensor factorization based behavioral representation learning methods for a wide range of behavioral analytics applications. Before joining Fraunhofer IAIS, he worked as a data scientist in the games industry, where he mainly concentrated on developing large scale machine learning solutions for business intelligence problems. He often gives talks and organizes workshops about informed machine learning and game analytics in scientific conferences and industrial summits.