Abstract

Esports broadcasting is a rapidly upcoming form of entertainment for gamers worldwide, and a new field of research for the academic community. Games broadcasted in this way are often fast-paced and complex, making it difficult for viewers to follow the action, especially for those relatively unfamiliar with the games being played, or the language used by the commentators. To understand esports games, game macro is a crucial concept, as it describes the overall strategy and decision-making in a game. This thesis presents a novel approach to categorizing, then automatically detecting and visualizing game macro in esports broadcasts. To facilitate this process, the formal game space of the game is used to formalize and simplify game states and events. Macro concepts are obtained through a combination of expert knowledge and analysis of esports broadcasts with a focus on game commentary. The resulting model is then used to detect and visualize game macro in real-time, using a combination of data sources to generate comprehensive game state snapshots. The framework is applied to the game League of Legends, using the Worlds 2023 tournament as the macro concept data source. A networked structure of the implementation is presented to allow for scalable and flexible deployment. Evaluation of the implementation is done through user interviews, which show that extra information about game macro can be beneficial to viewers, and that the system is able to provide this information in a way that is both informative and easy to understand. Only classical, rule based algorithms are used given the use of formal macro definitions, allowing for the potential of future improvements through machine learning techniques.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]