Game studios have cemented their place among the fastest-growing media industries. In recognition of this, we hosted an event in June through the Royal Statistical Society (RSS) Merseyside Local Group to explore AI and data science in computer game development. This was an amazing opportunity to engage with a different, in-vogue domain that has unique ties to data science. We showcased two fantastic presentations covering both academic and industry perspectives.
Stanley Wang, a data scientist at SEGA Europe, opened the event by showing the methods that SEGA uses to collect, process, and apply data on player decisions in-game. It was a revealing glimpse at how smoothly in-game data collection is integrated into SEGA’s digital platforms and the ways these data can be used to engage game-centred communities – for example, running special celebrations once milestones are hit for in-game events (revenue made, goals scored, etc.) or offering real-time integration with streaming platforms so viewers can see detailed statistics on in-game progress. Stanley showed one particular example where data collection fed directly into development decisions for Endless Space, a competitive strategy game where players vie for galactic conquest. During the beta (a period where a game is available to play but still considered in-testing before commercial release), SEGA were able to monitor how well-balanced the playable alien factions were based on real-time win rate data, which led to improvements to game mechanics for the final release.
We also learned how SEGA’s data science teams are using clustering methods to identify different game-playing behaviours in Two Point Hospital, a simulation game where players design, build, and manage a hospital through various scenarios. After compiling high-dimensional in-game data such as objectives achieved, treatment of staff, and even furniture choices, various clustering algorithms (including k-means clustering) were used to identify common sets of player behaviour. Stanley highlighted that when using these sorts of unsupervised learning methods, it’s useful to get insights from multiple models to inform methodological decisions like number of clusters chosen or how to treat outliers. SEGA identified four distinct types of player from these analyses, which you can hear more about from Stanley in the video below. The approach allowed the company to better understand gamers’ motivations and experiences with a view to designing future game content.
Our second speaker, Dr Konstantinos Tsakalidis, a lecturer in the Department of Computer Science at the University of Liverpool, presented exciting new ideas to teach computer games developers of the future. Dr Tsakalidis walked us through the curriculum for a dynamic new undergraduate program that reflects the latest software development technologies and the theory behind them. The course outline was designed around building knowledge and practice from the fundamentals upwards, starting from game physics as a prerequisite for game mechanics, game mechanics being a prerequisite for game content, and game content being a prerequisite for game AI. Combined with the continuous active involvement of students at each stage, this represented a great model of constructivist teaching. Dr Tsakalidis also proposed that practical game development (and subsequent assessments) should follow the latest research on data science and AI in computer games.
- About the author
- Alice-Maria Toader is a PhD student at the University of Liverpool and a committee member of the RSS Merseyside Local Group. Liam Brierley is a research fellow in health data science at the University of Liverpool and chair of the RSS Merseyside Local Group.
- Copyright and licence
- © 2023 Alice-Maria Toader and Liam Brierley
This article is licensed under a Creative Commons Attribution 4.0 (CC BY 4.0) International licence. Thumbnail image by Jezael Melgoza on Unsplash.
- How to cite
- Toader, Alice-Maria and Liam Brierley. 2023. “Where do AI, data science, and computer games intersect?” Real World Data Science, August 17, 2023. URL