Archetypal Analysis Based Anomaly Detection for Improved Storytelling in Multiplayer Online Battle Arena Games

Rafet Sifa, Anders Drachen, Florian Block, Spencer Moon, Anisha Dubhashi, Hao Xiao, Zili Li, Diego Klabjan, Simon Demediuk

Research output: Chapter in Book/Report/Conference proceedingConference contribution


Anomalies in esports refer to situations when something unexpected or unlikely happens. Rapid performance changes, unusual strategies, extraordinary plays, accelerated resource gains or team wipeouts comprise examples, but anomalies fundamentally comprise any situation where something unexpected happens. In multi-player online esports games such as multi-player online battle arena games, anomalies form a key component of the commentator-driven storytelling. In fast-paced, complex esports titles, anomalies can however go undetected until it is too late for commentators to note them and use them in their coverage, and for viewers they can take place outside the viewable area of the broadcast stream. Furthermore, there are limited tools available for commentators and players across professional and amateur levels for analysing or categorising anomalies. The research presented here provides a novel approach towards identifying one type of outliers in esports matches, via the application of archetype analysis to extract novel insights that can be used by commentators to improve esports coverage. As a case example, the major esports title League of Legends is used. We present a viable methodology for utilizing distributions resulting from the archetypal clusters and reconstruction errors to expose and explain anomalous events during gameplay.

Original languageEnglish
Title of host publicationProceedings of the Australasian Computer Science Week Multiconference 2021, ACSW 2021
EditorsNigel Stanger, Veronica Liesaputra Joachim
PublisherAssociation for Computing Machinery, Inc
Number of pages8
ISBN (Electronic)9781450389563
Publication statusPublished - 1 Feb 2021
Event2021 Australasian Computer Science Week Multiconference, ACSW 2021 - Virtual, Online, New Zealand
Duration: 1 Feb 20215 Feb 2021

Publication series

NameACM International Conference Proceeding Series


Conference2021 Australasian Computer Science Week Multiconference, ACSW 2021
Country/TerritoryNew Zealand
CityVirtual, Online

Bibliographical note

Funding Information:
Part of this research is supported by the Competence Center for Machine Learning Rhine Ruhr (ML2R) which is funded by the Federal Ministry of Education and Research of Germany (grant no. 01—S18038B). The authors would like to thank the anonymous reviews for their insightful comments.

Publisher Copyright:
© 2021 ACM.


  • Anomaly Detection
  • Archetypal Analysis
  • Maxoids
  • Multiplayer Online Battle Arena
  • Storytelling

Cite this