Online games are the integration of culture, art, and high-technology, which provide us with a new way of recreation and entertainment. As games become more complex and are reaching a broader audience, there is a growing interest and urgent need to analyze player behaviors and the impact of game design alternatives. However, due to large volumes and dynamic correlations of the gameplay data, as well as the high complexity of analytical tasks in real-world scenarios, it is still challenging for game analysts to conduct in-depth analysis and extract valuable information. Although many automatic approaches that can scale to massive data sizes for effective and rapid analysis are leveraged, the interpretation of the results can still be difficult to some extent. This triggers a broa...[ Read more ]

Online games are the integration of culture, art, and high-technology, which provide us with a new way of recreation and entertainment. As games become more complex and are reaching a broader audience, there is a growing interest and urgent need to analyze player behaviors and the impact of game design alternatives. However, due to large volumes and dynamic correlations of the gameplay data, as well as the high complexity of analytical tasks in real-world scenarios, it is still challenging for game analysts to conduct in-depth analysis and extract valuable information. Although many automatic approaches that can scale to massive data sizes for effective and rapid analysis are leveraged, the interpretation of the results can still be difficult to some extent. This triggers a broad use of visualization and visual analytics. By including human perception in the data exploration process, the flexibility, creativity and domain knowledge of human beings and the computational power of computer machines can be combined. This can further inform the basic organizing principles and patterns of in-game activities, such as understanding of game dynamics and design of novel, or augmentation of online games so as to support better user engagement.

In this thesis, we focus on online game community representation and two types of online game community dynamics, i.e., team-based combat dynamics and individual-based ego network dynamics. For the team-based combat dynamics, we propose a visual analytics system to help game designers discover patterns behind different occurrences in MOBA games. It produces a full gameplay visualization demonstrating detailed information of team formation, team combat, and team tactics. Then, to better facilitate the game occurrence analysis in breadth and depth, we propose a stepwise co-design process and enhance this visual analytics system by incorporating Machine Learning (ML) models to automatically recommend match segments of interest and further streamline the cross-match analysis. For the individual-based dynamics, we first study and identify an appropriate network representation for the online community which can greatly facilitate the downstream tasks. We then propose a visual analytics system to explore the evolution of the egocentric player social network. It not only provides a suite of novel visualization techniques to analyze the in-game ego network dynamics and impact propagation but also incorporates analytical metrics measuring structural changes during network evolution.

To the best of our knowledge, the above techniques are cutting-edge studies of visual analytics of online game dynamics. To validate the efficacy of our approaches, all the proposed techniques and systems are deployed in a game company to analyze real-world gameplay datasets and evaluated by domain experts.