In the latest issue (at the time of writing this, 2017. For a recent update, see arc.york.ac.uk) of Develop, we (Florian Block and Anders Drachen) discuss the use of data and analytics in esports: Whenever you play a game of DoTA 2, League of Legends or any other esports game, your behaviour and that of your teammates is recorded in great detail. The data is stored in replay files and made accessible online, providing us with the opportunity to not only visualise and analyse what is going on during matches but also to dive into the action post-match at a level that is unprecedented outside of esports.
Imagine if physical sports had the kind of detailed tracking that takes place in a digital game – it would revolutionise sports analytics. For many games such as Dota 2, we even have VR viewing modes that literally put you as an observer in the playing field during match runtime. The amount of data being generated from esports matches is huge. For instance, Dota 2 generates detailed gameplay data for over a million matches per day. Every single player action from all of these games can be download by the esports community, to analyse. The sheer size of data involved puts games on par with other big data accumulating fields like astronomy and finance.
While the data is available and the esports community has a growing appetite to engage with data, there are not many good tools around for working with it. If you are a professional team or a major broadcaster you may have an analyst working for you, dissecting matches and providing a detailed analysis. However, the vast majority of the 320 million-strong esports community is reliant on a few analysis providers and the tools built into the game clients. If you are a professional team or a major broadcaster you may have an analyst working for you, dissecting matches and providing a detailed analysis. However, the vast majority of the 320 million-strong esports community is reliant on a few analysis providers and the tools built into the game clients.
Current esports games have made great strides in providing great visualisation of metrics during and after matches. But, in terms of actual analysis of in-game performance, or the ability to share data and visualisations, we still have a long way to go. This problem is one of the key areas for us at Digital Creativity Labs (DC Labs) and the Intelligent Games and Games Intelligence programme (IGGI). DC Labs is the largest games research centre in the world, with over 30 scientists and dozens of PhD students, all working on games technology. Together with tournament organisers, broadcasters, teams, coaches and others from esports, we are working on democratising the access to analytical tools and techniques to help anyone extract value from gameplay data. This ranges from data platforms such as Dotabuff and OpenDOTA, developers who want to understand player behaviour, streamers and YouTubers to improve their coverage, all the way to our kids who want to share their stats with friends and maybe pick up some maths and data science along the way.
These challenges cover a lot of different stakeholders with varied needs and requirements. To develop solutions that are flexible, and tailored to non-experts, you need to think about adaptive interfaces. Within AI and user experience research, the notion of using AI assistants to work with the user directly to help with tasks is particularly interesting. We are also exploring recommender systems that assist with data analysis interpretation and generation of data-driven stories. In this context, machine learning has some exciting new application areas, such as automatically detecting ‘entertaining’ bits of gameplay among vast numbers of games, providing creative suggestions for data-driven commentary and providing on the fly coverage that intelligently adapts to a viewer’s preference.
These techniques could also radically transform production workflows in esports, especially allowing low-staffed, low-budget productions achieve a professional-grade quality of coverage. From the perspective of the player, being able to access your own data along with the ability to analyse and compare it with others, with AI assistants that help you improve, is of recurrent interest. While physical sports are still far away from realising those levels of data analytics, esports provides a fertile ground for innovation in this space that is ready for harvest right now. In the future, we think we can expect a plethora of tools becoming available that make it fun and engaging to interact with esports data.