Automated Playtesting With Procedural Personas Through MCTS With Evolved Heuristics
Open accessThe DOI matches its Crossref record; 26 of the 29 references it cites were found in a registry or an index.
Abstract
This paper describes a method for generative player modeling and its application to the automatic testing of game content using archetypal player models called procedural personas. Theoretically grounded in psychological decision theory, procedural personas are implemented using a variation of Monte Carlo tree search (MCTS) where the node selection criteria are developed using evolutionary computation, replacing the standard UCB1 criterion of MCTS. Using these personas, we demonstrate how generative player models can be applied to a varied corpus of game levels and demonstrate how different playstyles can be enacted in each level. In short, we use artificially intelligent personas to construct synthetic playtesters. The proposed approach could be used as a tool for automatic play testing when human feedback is not readily available or when quick visualization of potential interactions is necessary. Possible applications include interactive tools during game development or procedural content generation systems where many evaluations must be conducted within a short time span.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“In contrast to prior studies that rely on reward coefficient diversification or utility function adjustments [16], [28], [31], [32], the UBCL framework leverages actual game metrics in the reward function, enabling approximate reproduction of desired human-like behaviors by using game-play data, with accuracy varying across behavioral dimensions.”
From UBCL: A Reinforcement Learning Framework for Controllable and Diverse Player Behaviors (2026) · cites this paper as [16]
The cited source says what the claim attributes to it.
In the citing paper
“Procedural personas [4] [5] accomplish this by rewiring the feedback mechanism by a utility function.”
From Playtesting: What is Beyond Personas (2023) · cites this paper as [4]
The cited source says what the claim attributes to it.
In the citing paper
“Second, depending on the level layout, personas can execute a similar sequence [4].”
From Playtesting: What is Beyond Personas (2023) · cites this paper as [4]
The cited source says what the claim attributes to it.
In the citing paper
“As a result, personas enabled distinct playstyles and helped to playtest a game like distinct players.”
From Playtesting: What is Beyond Personas (2023) · cites this paper as [4]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
A Survey of Monte Carlo Tree Search Methods
Browne, C. B., Powley, E., Whitehouse, D., Lucas, S. M., Cowling, P. I., Rohlfshagen, P., Tavener, S., Perez, D., Samothrakis, S., & Colton, S. (2012) · IEEE Transactions on Computational Intelligence and AI in Games · doi:10.1109/tciaig.2012.2186810 · Cited by 2,970 · cited 4×
“1) Selection: MCTS chooses the next node to expand via the tree policy, starting at the root node and recursively picking the highest scoring child “until the most urgent expandable node is reached” [26] or a terminal state (i.”
BackgroundMethodInfluentialCooper, A. (1999) · Cited by 1,157
The Island Model Genetic Algorithm: On Separability, Population Size and Convergence
Whitley, D., Rana, S., & Heckendorn, R. B. (2015) · Journal of Computing and Information Technology · Cited by 380
Towards automatic personalised content creation for racing games
Togelius, J., De Nardi, R., & Lucas, S. M. (2007) · 2007 IEEE Symposium on Computational Intelligence and Games · doi:10.1109/cig.2007.368106 · Cited by 283 · cited 1×
“Perhaps the most obvious approach is to use some form of supervised learning to derive a model from play traces [18], [19].”
BackgroundSelf-citationGame Analytics : Maximizing the Value of Player Data
El‐Nasr, M. S., Drachen, A., & Canossa, A. (2013) · Cited by 260
Game design workshop : designing, prototyping, and playtesting games
Fullerton, T., Swain, C., & Hoffman, S. J. (2004) · 기초학문자료센터(KRM) (Korea Research Foundation) · Cited by 258
Game Analytics: Maximizing the Value of Player Data
(2013) · Springer eBooks · doi:10.1007/978-1-4471-4769-5 · Cited by 228 · cited 1×
“ase of continuously updated “live” games. Each approach has different strengths, weaknesses and costs, making it relevant for different game makers or different stages of the game making process [1], [2]. In this paper we suggest and demonstrate a method taking a new position on this spectrum (illustrated in Figure 1)…”
MethodSelf-citationDefining personas in games using metrics
Tychsen, A., & Canossa, A. (2008) · Proceedings of the 2008 Conference on Future Play: Research, Play, Share · doi:10.1145/1496984.1496997 · Cited by 169 · cited 1×
“They call this new conceptualization play personas and operationalized their definition through data mining, suggesting how the persona design process could be supported by analyzing quantitative game data gathered via telemetrics [13].”
BackgroundYannakakis, G. N., Spronck, P. H. M., Loiacono, D., & André, E. (2013) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/dfu.vol6.12191.45 · Cited by 108 · cited 1×
“Player modeling is the learning and use of computational models of player preference, experience, and/or behavior [17].”
BackgroundEvolving personas for player decision modeling
Holmgard, C., Liapis, A., Togelius, J., & Yannakakis, G. N. (2014) · 2014 IEEE Conference on Computational Intelligence and Games · doi:10.1109/cig.2014.6932911 · Cited by 106 · cited 1×
“featured in previous attempts at modeling personas via MCTS [3] or in the simpler MiniDungeons game [6], [7].”
MethodSelf-citation
Research works printed but not linked: 1
- hunicke, 0, MDA: A formal approach to game design and game research, Proc AAAI Workshop Challenges Game AI, 4, 1
Cited by
Items 101–125 of 125
Novak, D., Verber, D., Dugonik, J., & Fister, I. (2020) · Mathematics · doi:10.3390/math8050688 · Cited by 1
“With this paper, we tackle the problem of the automatic validation of game features for the game space specified in GDC and also address research requirements from articles (for instance, [43]), where the authors point out the need that games with a higher complexity have of scaling.”
MethodA Game-Based Approach for Helping Designers Learn Machine Learning ConceptsPreprint
Myers, C. M., Xie, J., & Zhu, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2009.05605 · Cited by 4
Synthesizing play: exploring the use of artificial intelligence to evaluate game user experience
Stahlke, S. (2020) · e-scholar@UOIT (University of Ontario Institute of Technology) · Cited by 1
Rinascimento: searching the behaviour space of SplendorPreprint
Bravi, I., & Lucas, S. M. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.08371 · Cited by 1
Exploring Game Design through Human-AI Collaboration
Alvarez, A. (2022) · doi:10.24834/isbn.9789178773084 · Cited by 8
UBCL: A Reinforcement Learning Framework for Controllable and Diverse Player Behaviors
Cilan, A., & Ozgovde, A. (2026) · IEEE Transactions on Games · doi:10.1109/tg.2026.3703355
“Themethodoffersascalablesolutionforautomated playtesting,gamebalancing,human-likebehaviorsimulation,andreplacingdisconnectedplayersinonlinegames.”
Expressive Response Curves: Testing Expressive Game Feel with A*
Junius, N., & Carstensdottir, E. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment · doi:10.1609/aiide.v19i1.27524
“…et al. 2006, 2008), is commonly used to drive automated playtesters (Zook, Harrison, and Riedl 2015; Keehl and Smith 2018; Horn et al. 2018; Holmg ˚ ard et al. 2018; Mugrai et al. 2019) as it requires no domain knowledge to be integrated into the al-gorithm (Jacobsen, Greve, and Togelius…”
BackgroundMethodClustering Player Strategies from Variable-Length Game Logs in DominionPreprint
Bendekgey, H. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.11273
Game AI hyperparameter tuning in rinascimento
Bravi, I., Volz, V., & Lucas, S. M. (2019) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3319619.3326842
“aggressive or defensive playing styles [6].”
Automatic Critical Mechanic Discovery in Video Games.Preprint
Green, M. C., Khalifa, A., Barros, G. A. B., Machado, T., & Togelius, J. (2019) · arXiv (Cornell University)
Self-citationTowards Designer Modeling through Design Style ClusteringPreprint
Alvarez, A., Font, J., & Togelius, J. (2020) · arXiv (Cornell University)
Self-citationRinascimento: Playing Splendor-Like Games With Event-Value Functions
Bravi, I., & Lucas, S. M. (2022) · IEEE Transactions on Games · doi:10.1109/tg.2022.3171908
“Holmgård et al. [14] associate a metric to each affordance and, then, design AI agents that plan their actions to maximize/minimize a selection of such metrics in order to show specific behavioral traits.”
BackgroundCautious curiosity: A novel approach to a human-like gameplay agent
Zhou, C. (2022) · Digital Repository Service (Northeastern University) · doi:10.17760/d20449077
Automatic Testing and Validation of Level of Detail Reductions Through Supervised Learning
Tamm, M., Shamon, O., Leon, H. A., Tollmar, K., & Gisslén, L. (2022) · IEEE Conference on Games (CoG) · doi:10.1109/cog51982.2022.9893682 · Cited by 2
Designing a Mood-Mediated Multi-Level Reasoner
Gomes, S., Rocha, J. B., Dias, J., & Martinho, C. (2023) · IEEE Transactions on Affective Computing · doi:10.1109/taffc.2023.3293310
“Given that MCTS scales particularly well for problems with continuous or massive discrete action spaces, and because its solutions are highly dependent on the way nodes are selected and expanded, several research applied the algorithm to construct decision-making profiles [33], [34].”
BackgroundMethodResult‘Journeys in the Dark’ - Towards Game Master AI in Complex Board Games
Best, T., Lucas, S., & Gaina, R. D. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment · doi:10.1609/aiide.v20i1.31861
“This element of tactical combat and multi-layered strategy can be realised by the AI agents within D2e similar to the approaches out-lined in Holmg˚ard et al.’s study (Holmg˚ard et al. 2018).”
Injecting Combinatorial Optimization into MCTS: Application to the Board Game Boop
Richoux, F. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-90065-5_32
Stuck in the Middle: Generating Levels without (or with) Softlocks
Cooper, S., & Bazzaz, M. (2025) · International Conference on the Foundations of Digital Games (FDG) · doi:10.1145/3723498.3723844 · Cited by 1
Pathways to Mastery: A Taxonomy of Player Progression Systems in Commercial Video Games
Guzsvinecz, T. (2025) · From Pixels to Play - The Art and Science of Video Games · doi:10.5772/intechopen.1011500
Testing Open-World Games: A Minecraft Case Study
Chen, Q., Zhang, M., & Zhang, X. (2025) · IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW) · doi:10.1109/issrew67781.2025.00059 · Cited by 1
“Holmg˚ard et al. [19] employed Monte Carlo Tree Search with evolved heuristics to traverse game states beyond predefined scripts.”
AI-Driven Player Behavior Analysis and Personalization
Zebua, N., Ziraluo, Y. P. B., Zebua, E. N. K., & Nazara, R. V. (2026) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3373-3690-9.ch007
Génération procédurale de niveaux de jeu alliant approche constructive et optimisation
Bailly, R. (2022) · theses.fr (ABES) · doi:10.70675/1adbf4c9z8da0z41bbz8063zec137916a1c3
Simulation Of Diverse Human Play-Styles In Video Games : A Reinforcement Learning Approach
de Woillemont, P. L. P. (2023) · theses.fr (ABES) · doi:10.70675/51b3c6d7z2e51z47fcz8d7azf1106953ba36 · Cited by 1
Generation and Evaluation of Exploration Terms for Monte Carlo Tree Search
Cazenave, T. (2026) · ICGA Journal · doi:10.1177/13896911261463788
“Closer to the design of simulation heuristics than to selection policies, Holmgård et al. (2018) evolved heuristics for MCTS to drive automated playtesting with procedural personas, and Alhejali and Lucas (2013) used Genetic Programming to evolve playout heuristics for an MCTS Ms. Pac-Man agent.”
Gil, H., & Kang, S.-J. (2026) · IEEE Access · doi:10.1109/access.2026.3712365
“Furthermore, Holmgard et al. [29] demonstrated that deploying procedural personas can improve the consistency of automated playtesting.”
Metadata
- Authors
- Holmgard, C., Green, M. C., Liapis, A., & Togelius, J.
- Affiliations
- New York University (US)
- University of Malta (MT)
- Venue
- IEEE Transactions on Games
- Published
- 2019-12
- Volume
- 11
- Issue
- 4
- Pages
- 352-362
- ISSN
- 2475-1502 (ISSN-L)
- 2475-1510
- OpenAlex
- W2789119104
- Cited by
- 125
- References
- 29
- Citations by year
- 2026: 4 · 2025: 10 · 2024: 7 · 2023: 14 · 2022: 25 · 2021: 17 · 2020: 24 · 2019: 16
- FWCI
- 9.15
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Games (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- GAANN Program
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence in Games
- Social Sciences › Sociology and Political Science › Digital Games and Media
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Keywords
- Monte Carlo tree search, UCB1, game development tools, procedural content generation
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.