Rafet Sifa

Guns, swords and data: Clustering of player behavior in computer games in the wild

Behavioral data from computer games can be exceptionally high-dimensional, of massive scale and cover a temporal segment reaching years of real-time and a varying population of users. Clustering of user behavior provides a way to discover behavioral patterns that are actionable for game developers. Interpretability and reliability of clustering results is vital, as decisions based on them affect game design and thus ultimately revenue. Here case studies are presented focusing on clustering analysis applied to high-dimensionality player behavior telemetry, covering a combined total of 260,000 characters from two major commercial game titles: the Massively Multiplayer Online Role-Playing Game Tera and the multi-player strategy war game Battlefield 2: Bad Company 2. K-means and Simplex Volume Maximization clustering were applied to the two datasets, combined with considerations of the design of the games, resulting in actionable behavioral profiles. Depending on the algorithm different insights into the underlying behavior of the population of the two games are provided.

How players lose interest in playing a game: An empirical study based on distributions of total playing times

Analyzing telemetry data of player behavior in computer games is a topic of increasing interest for industry and research, alike. When applied to game telemetry data, pattern recognition and statistical analysis provide valuable business intelligence tools for game development. An important problem in this area is to characterize how player engagement in a game evolves over time. Reliable models are of pivotal interest since they allow for assessing the long-term success of game products and can provide estimates of how long players may be expected to keep actively playing a game. In this paper, we introduce methods from random process theory into game data mining in order to draw inferences about player engagement. Given large samples (over 250,000 players) of behavioral telemetry data from five different action-adventure and shooter games, we extract information as to how long individual players have played these games and apply techniques from lifetime analysis to identify common patterns. In all five cases, we find that the Weibull distribution gives a good account of the statistics of total playing times. This implies that an average players interest in playing one of the games considered evolves according to a non-homogeneous Poisson process. Therefore, given data on the initial playtime behavior of the players of a game, it becomes possible to predict when they stop playing.

Predicting player churn in the wild

Free-to-Play or “freemium” games represent a fundamental shift in the business models of the game industry, facilitated by the increasing use of online distribution platforms and the introduction of increasingly powerful mobile platforms. The ability of a game development company to analyze and derive insights from behavioral telemetry is crucial to the success of these games which rely on in-game purchases and in-game advertising to generate revenue, and for the company to remain competitive in a global marketplace. The ability to model, understand and predict future player behavior has a crucial value, allowing developers to obtain data-driven insights to inform design, development and marketing strategies. One of the key challenges is modeling and predicting player churn. This paper presents the first cross-game study of churn prediction in Free-to-Play games. Churn in games is discussed and thoroughly defined as a formal problem, aligning with industry standards. Furthermore, a range of features which are generic to games are defined and evaluated for their usefulness in predicting player churn, e.g. playtime, session length and session intervals. Using these behavioral features, combined with the individual retention model for each game in the dataset used, we develop a broadly applicable churn prediction model, which does not rely on game-design specific features. The presented classifiers are applied on a dataset covering five free-to-play games resulting in high accuracy churn prediction.

Clustering Game Behavior Data

Recent years have seen a deluge of behavioral data from players hitting the game industry. Reasons for this data surge are many and include the introduction of new business models, technical innovations, the popularity of online games, and the increasing persistence of games. Irrespective of the causes, the proliferation of behavioral data poses the problem of how to derive insights therefrom. Behavioral data sets can be large, time-dependent and high-dimensional. Clustering offers a way to explore such data and to discover patterns that can reduce the overall complexity of the data. Clustering and other techniques for player profiling and play style analysis have, therefore, become popular in the nascent field of game analytics. However, the proper use of clustering techniques requires expertise and an understanding of games is essential to evaluate results. With this paper, we address game data scientists and present a review and tutorial focusing on the application of clustering techniques to mine behavioral game data. Several algorithms are reviewed and examples of their application shown. Key topics such as feature normalization are discussed and open problems in the context of game analytics are pointed out.

The Playtime Principle: Large-scale cross-games interest modeling

The collection and analysis of behavioral telemetry in digital games has in the past five years become an integral part of game development. One of the key challenges in game analytics is the development of methods for characterizing and predicting player behavior as it evolves over time. Characterizing behavior is necessary for monitoring player populations and gradually improve game design and the playing experience. Predicting behavior is necessary to describe player engagement and prevent future player churn. In this paper, methods and theory from kernel archetype analysis and random process models are utilized to evaluate the playtime behavior, i.e. time spent playing specific games as a function of time, of over 6 million players, across more than 3000 PC and console games from the Steam platform, covering a combined playtime of more than 5 billion hours. A number of conclusions can be derived from this large-scale analysis, notably that playtime as a function of time, across the thousands of games in the dataset, and irrespective of local differences in the playtime frequency distribution, can be modeled using the same model: the Weibull distribution. This suggests that there are fundamental properties governing player engagement as it evolves over time, which we here refer to as the Playtime Principle. Additionally, the analysis shows that there are distinct clusters, or archetypes, in the playtime frequency distributions of the investigated games. These archetypal groups correspond to specific playtime distributions. Finally, the analysis reveals information about player behavior across a very large dataset, showing for example that the vast majority of games are players for less than 10 hours, and very few players spend more than 30-35 hours on any specific game.

Behavior evolution in Tomb Raider Underworld

Behavioral datasets from major commercial game titles of the “AAA” grade generally feature high dimensionality and large sample sizes, from tens of thousands to millions, covering time scales stretching into several years of real-time, and evolving user populations. This makes dimensionality-reduction methods such as clustering and classification useful for discovering and defining patterns in player behavior. The goal from the perspective of game development is the formation of behavioral profiles that provide actionable insights into how a game is being played, and enables the detection of e.g. problems hindering player progression. Due to its unsupervised nature, clustering is notably useful in cases where no prior-defined classes exist. Previous research in this area has successfully applied clustering algorithms to behavioral datasets from different games. In this paper, the focus is on examining the behavior of 62,000 players from the major commercial game Tomb Raider: Underworld, as it unfolds from the beginning of the game and throughout the seven main levels of the game. Where previous research has focused on aggregated behavioral datasets spanning an entire game, or conversely a limited slice or snapshot viewed in isolation, this is to the best knowledge of the authors the first study to examine the application of clustering methods to player behavior as it evolves throughout an entire game.

Archetypical motion: Supervised game behavior learning with Archetypal Analysis

The problem of creating believable game AI poses numerous challenges for computational intelligence research. A particular challenge consists in creating human-like behaving game bots by means of applying machine learning to game-play data recorded by human players. In this paper, we propose a novel, biologically inspired approach to behavior learning for video games. Our model is based on the idea of movement primitives and we use Archetypal Analysis to determine elementary movements from data in order to represent any player action in terms of convex combinations of archetypal motions. Given these representations, we use supervised learning in order to create a system that is able to synthesize appropriate motion behavior during a game. We apply our model to teach a first person shooter game bot how to navigate in a game environment. Our results indicate that the model is able to simulate human-like behavior at lower computational costs than previous approaches.

Beyond heatmaps: Spatio-temporal clustering using behavior-based partitioning of game levels

Evaluating the spatial behavior of players allows for comparing design intent with emergent behavior. However, spatial analytics for game development is still in its infancy and current analysis mostly relies on aggregate visualizations such as heatmaps. In this paper, we propose the use of advanced spatial clustering techniques to evaluate player behavior. In particular, we consider the use of DEDICOM and DESICOM, two techniques that operate on asymmetric spatial similarity matrices and can simultaneously uncover preferred locations and likely transitions between them. Our results highlight the ability of asymmetric techniques to partition game maps into meaningful areas and to retain information about player movements between these areas.

User Churn Migration Analysis with DEDICOM

Time plays an important role regarding user preferences for products. It introduces asymmetries into the adoption of products which should be considered in the context of recommender systems and business intelligence. We therefore investigate how temporally asymmetric user preferences can be analyzed using a latent factor model called Decomposition Into Directional Components (DEDICOM). We introduce a new scalable hybrid algorithm that combines projected gradient descent and alternating least squares updates to compute DEDICOM and imposes semi-nonnegativity constraints to better interpret the resulting factors. We apply our model to analyze user churn and migration between different computer games in a social gaming environment.

Predicting Retention in Sandbox Games with Tensor Factorization-based Representation Learning

Major commercial (AAA) games increasingly transit to a semi-persistent or persistent format in order to extend the value of the game to the player, and to add new sources of revenue beyond basic retail sales. Given this shift in the design of AAA titles, game analytics needs to address new types of problems, notably the problem of forecasting future player behavior. This is because player retention is a key factor in driving revenue in semi-persistent titles, for example via downloadable content. This paper introduces a model for predicting retention of players in AAA games and provides a tensor-based spatio-temporal model for analyzing player trajectories in 3D games. We show how knowledge as to trajectories can help with predicting player retention. Furthermore, we describe two new algorithms for three way DEDICOM including a fast gradient method and a seminonnegative constrained method. These approaches are validated against a detailed behavioral data set from the AAA open-world game Just Cause 2.