Guilherme Matos de Faria
Understanding and Visualising Matchups in Hearthstone
Collectable card games (CCGs) have been a big part of gaming for the past 2 decades. For
instance, Hearthstone, one of the most popular CCGs, has over 100 million players . Decks and their construction are a crucial part of CCGs. Players build synergies with a variety of cards in order to develop a strategy that will lead to a win. With constantly evolving deck landscapes, players need to adapt and learn new tactics. This can be a hard skill to develop: “How does one deck beat another?”, “How do I improve my chances against a specific deck?”, are common questions asked by players of all skill levels.
The research proposed will focus on leveraging AI and data visualisation to assist players and viewers in the interpretation of deck matchups, in the CCG Hearthstone. It will be supervised by Dr Florian Block (University of York) and is supported by HSReaply.net, the leading data provider for Hearthstone, who will provide access to their database of over 1 billion Hearthstone matches.
Research Question: Can the use of AI methods on CCG Hearthstone data from HSReplay.net, help in the investigation of deck strategy interactions, in order to obtain reasoning behind favoured and unfavoured matchups?
Prior work has explored the idea of automatic deckbuilding  and improving automatic play
. However, no work has focused on using AI to create explainable interpretation of deck
matchups during live Hearthstone games.
This research will aim not only to explore matchup interaction but also identify key cards in
deck strategies. HSReplay.net has highly detailed data, down to cards played in a turn and their targets. Statistical analysis and AI can be applied to sequence of cards played to investigate how these sequences influence the outcome of the game. As a starting point, clustering can be used to find archetypes and statistical power curve analysis will give insight into basic deck strategy. An iterative design process involving players and industry stakeholders will be used to continuously inform research and ensure its practical and commercial relevance.
The proposed research has three main application areas. First, the produced model and tools could help players to better understand crucial concepts in matchups and facilitate acquisition of skills. Secondly, developers can use it to investigate deck strategies and guide continued balancing. Furthermore, in  it has been shown that statistical information and info graphics can used by broadcasters to better engage their audiences with insightful and entertaining commentary.
 Stephanie Fogel. ‘Hearthstone’ Now Has Over 100 Million Players. Variety [online]. Available at: https://variety.com/2018/gaming/news/hearthstone-has-over-100-million- players-1203019919/ [Accessed 05 Mar. 2019].
 Garcia-Sanchez, P., Tonda, A., Squillero, G., Mora, A. and Merelo, J. (2016). Evolutionary deckbuilding in hearthstone. 2016 IEEE Conference on Computational Intelligence and Games (CIG). [online] Available at: https://ieeexplore.ieee.org/abstract/document/7860426 [Accessed5 Mar. 2019].
 Zhang, S. and Buro, M. (2017). Improving hearthstone AI by learning high-level rollout policies and bucketing chance node events. 2017 IEEE Conference on Computational Intelligence and Games (CIG). [online] Available at: https://ieeexplore.ieee.org/abstract/document/8080452 [Accessed 5 Mar. 2019].
 Block, F., Hodge, V., Hobson, S., Sephton, N., Devlin, S., Ursu, M., Drachen, A. and Cowling, P. (2018). Narrative Bytes. Proceedings of the 2018 ACM International Conference on Interactive Experiences for TV and Online Video - TVX '18. [online] Available at: http://eprints.whiterose.ac.uk/132551/1/TVX_Data_Driven_Storytelling_Camera_Ready_Final.pdf [Accessed 5 Mar. 2019]
Home institution: York
Supervisor: Dr Florian Block
Ready to apply?
Once you have identified your potential supervisor, we would encourage you to contact them to discuss your research proposal.Learn More