anders drachen's Blog
I am a data scientist working with games.
My work centers on the analysis of user behavior in – and around – games. The reason I work with games is simple: They provide unprecedented access to incredibly detailed measures of behavior. Using logging technology, it is possible to capture the second-by-second interaction between user and game. This kind of precision is not possible in the real world, and provides incredible opportunities to study human behavior and how it can be affected – whether for the purposes of improving the user experience, optimizing design, facilitating learning, or monetization. One of the central challenges in this endeavour is the massive scale these kinds of behavioral data can reach, and the high degree of complexity. Imagine working with data describing thousands of actions across hundreds of variables from millions of players. Now imagine adding contextual data – demographics, video capture, physiological data, personality information, geographic information and much more … deriving actionable insights from such datasets requires not only analytical skills but also a deep understanding of an passion for games. I fundamentally form part of an international community of analysts in the private and academic sectors who try to derive meaning from user behavior in games, in order to inform game development.
I blog about all of this at www.andersdrachen.com
Behavioral profiling is among the most useful of the tools in the arsenal of a game analyst. Profiling allows a way to manage complex user data and discover patterns in behavior. This post covers the basics and presents notes on popular types of profiles.
Analytical systems designed to help players learn and improve can have a huge impact on games in the future, not the least in esports. Here we present the process behind building a recommender system for Destiny players.
Social connections are important to the experience of many online games. In Destiny´s PvP hub The Crucible, strong social connections also appear to be important for player performance.
Given a huge and varied game like Destiny, it is of interest to see if there are any patterns in how people play the game. Here we present player profiles from Destiny, and show how weapon use is a powerful characterization indicator in the game.
Here we introduce the idea of using heuristics models in predictive mobile game analytics. They address the problem of enabling small companies to access frugal but effective prediction models that are easy to understand, deploy and scale.
5 billion hours of play. 6 million players. 3000 games on Steam. 1 distribution model that explains how time spent playing games on Steam rises and falls – as a function of time.
anders drachen's Comments
[Blog - 11/23/2016 - 10:25]
Hi Gerry, if I understand ...
Hi Gerry, if I understand your comment correctly, you are querying about skill-matching. Destiny uses their own matchmaking algorithm so in theory at least matches should be of equal difficulty . So players cannot purposefully seek out weaker opponents as opponents will generally be matched in skill given the limitations ...
[Blog - 11/03/2016 - 04:05]
Hi Jacqueline, thanks for your ...
Hi Jacqueline, thanks for your kinds words about the report. The gameplay statistics has a focus on weapons because they are a core element of Destiny, and turned out to have a high explanatory value over a lot of other features we investigated in terms of the cluster solution. It ...
[Blog - 10/05/2015 - 01:52]
I often tell my students: ...
I often tell my students: there are four goals with empirical research: generalizability, validity, reliability and importance. You cannot have importance without satisfying the criteria for the first three. It simplifies a complex set of problems but is a good rule to keep in mind. Good piece Nick.
[Blog - 11/04/2014 - 02:42]
Great post Julian, and thanks ...
Great post Julian, and thanks for the reference. r n r nCross-promotion and transfer is a very interesting topic atm. with a ton of challenges, similar to retention and incentivization. We need more like this.
[Blog - 07/07/2014 - 08:48]
Hi everyone r n r ...
Hi everyone r n r nThanks for commenting on the post. I hope I can answer your questions. r n r n@Ben Weber: All agreed re. lifetime fitting being interesting. In this case we used bins of 1 hour duration per player, aggregating these across all players for that game. ...
[Blog - 09/11/2013 - 08:37]
Hi All r n r ...
Hi All r n r nApologies for responding to your comments late. r n r nArnaud: I take your point about rotating the graphs, I will remember this in the future. r n r nJohan/Isaac: yep, in the olden days some special characters were possible in WOW hyphens etc. . ...