Predicting Game Difficulty and Engagement Using AI Players

Shaghayegh Roohi, Christian Guckelsberger, Asko Relas, Henri Heiskanen, Jari Takatalo, Perttu Hämäläinen

Research output: Contribution to journalArticleScientificpeer-review

5 Citations (Scopus)
242 Downloads (Pure)


This paper presents a novel approach to automated playtesting for the prediction of human player behavior and experience. We have previously demonstrated that Deep Reinforcement Learning (DRL) game-playing agents can predict both game difficulty and player engagement, operationalized as average pass and churn rates. We improve this approach by enhancing DRL with Monte Carlo Tree Search (MCTS). We also motivate an enhanced selection strategy for predictor features, based on the observation that an AI agent's best-case performance can yield stronger correlations with human data than the agent's average performance. Both additions consistently improve the prediction accuracy, and the DRL-enhanced MCTS outperforms both DRL and vanilla MCTS in the hardest levels. We conclude that player modelling via automated playtesting can benefit from combining DRL and MCTS. Moreover, it can be worthwhile to investigate a subset of repeated best AI agent runs, if AI gameplay does not yield good predictions on average.
Original languageEnglish
Article number231
Pages (from-to)1-17
Number of pages17
JournalProceedings of the ACM on Human-Computer Interaction
Issue numberCHIPLAY
Publication statusPublished - Sept 2021
MoE publication typeA1 Journal article-refereed
EventACM SIGCHI Annual Symposium on Computer-Human Interaction in Play - Vienna, Austria
Duration: 18 Oct 202121 Oct 2021
Conference number: 8


Dive into the research topics of 'Predicting Game Difficulty and Engagement Using AI Players'. Together they form a unique fingerprint.

Cite this