pith. sign in

arxiv: 1907.01623 · v1 · pith:ATIHNAO6new · submitted 2019-07-02 · 💻 cs.AI · cs.NE

Evolving the Hearthstone Meta

Fernando de Mesentier Silva , Rodrigo Canaan , Scott Lee , Matthew C. Fontaine , Julian Togelius , Amy K. Hoover This is my paper

Pith reviewed 2026-05-25 10:40 UTC · model grok-4.3

classification 💻 cs.AI cs.NE
keywords Hearthstonegame balanceevolutionary algorithmwin ratecard attributesmulti-objective optimizationmeta
0
0 comments X

The pith

An evolutionary algorithm finds card attribute changes that drive Hearthstone deck win rates toward 50 percent with minimal alterations.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines balancing Hearthstone by modifying attributes of its many cards to promote diverse strategies. It measures how changes affect simulated match-ups between decks and player strategies. An evolutionary algorithm searches for combinations of such changes that bring win rates close to equal. A multi-objective version of the algorithm adds the goal of minimizing the total number of changes made. Metrics are also introduced to help identify which cards would benefit most from adjustment.

Core claim

Simulations of deck match-ups show that an evolutionary algorithm can locate sets of card attribute modifications that move overall win rates toward 50 percent. The same search can be reformulated as a multi-objective problem to reach that target while altering the smallest possible number of cards.

What carries the argument

Multi-objective evolutionary algorithm that jointly optimizes for balanced win rates and minimal card modifications.

If this is right

  • Targeted attribute changes can equalize performance across a range of decks in simulation.
  • Multi-objective search reduces the scale of disruption needed to reach balance.
  • Proposed metrics can rank cards by their potential impact on meta balance.
  • Pre- and post-change win-rate comparisons provide a quantitative basis for patch decisions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same search process could be rerun periodically as new cards are added to keep the meta in check.
  • If simulation and live data diverge, the evolutionary fitness function could be updated with real match logs.
  • The minimal-change objective might generalize to other games where small patches are preferred over large overhauls.

Load-bearing premise

Simulated win rates between decks accurately predict the effects that real card changes would have on player behavior and overall balance.

What would settle it

Apply the algorithm's recommended card changes to the live game and measure whether observed win rates across decks converge to the simulated 50 percent target.

Figures

Figures reproduced from arXiv: 1907.01623 by Amy K. Hoover, Fernando de Mesentier Silva, Julian Togelius, Matthew C. Fontaine, Rodrigo Canaan, Scott Lee.

Figure 1
Figure 1. Figure 1: Average, Min and Max fitness over 12 generations. Baseline is the [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A scatter of points representing the individuals achieved through multi [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: A scatterplot of WRP (a) and WRD (b) versus WRN [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
read the original abstract

Balancing an ever growing strategic game of high complexity, such as Hearthstone is a complex task. The target of making strategies diverse and customizable results in a delicate intricate system. Tuning over 2000 cards to generate the desired outcome without disrupting the existing environment becomes a laborious challenge. In this paper, we discuss the impacts that changes to existing cards can have on strategy in Hearthstone. By analyzing the win rate on match-ups across different decks, being played by different strategies, we propose to compare their performance before and after changes are made to improve or worsen different cards. Then, using an evolutionary algorithm, we search for a combination of changes to the card attributes that cause the decks to approach equal, 50% win rates. We then expand our evolutionary algorithm to a multi-objective solution to search for this result, while making the minimum amount of changes, and as a consequence disruption, to the existing cards. Lastly, we propose and evaluate metrics to serve as heuristics with which to decide which cards to target with balance changes.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper proposes using an evolutionary algorithm (extended to multi-objective optimization) to search for minimal changes to Hearthstone card attributes such that simulated match-ups between different decks and strategies approach 50% win rates; it also introduces metrics to heuristically select which cards to target for balance changes.

Significance. If the simulation-based evaluation were shown to be a reliable proxy for live meta effects and the method produced validated, minimal patches, the work could offer a data-driven complement to manual balancing in complex card games. However, the manuscript supplies no experimental results, implementation details, calibration against historical patches, or out-of-sample validation, so any assessment of significance remains speculative.

major comments (2)
  1. [Abstract] Abstract: the description of the evolutionary search for 50% win-rate balance and the multi-objective minimal-change extension supplies no results, validation experiments, error analysis, or implementation details, making it impossible to determine whether the approach supports the stated claims.
  2. [Abstract] Abstract: the central claim requires that simulated win rates from fixed deck match-ups serve as a faithful proxy for how attribute tweaks affect emergent player behavior and long-term meta evolution, yet the manuscript provides no calibration against historical patch data, player logs, or adaptive-strategy tests to support this assumption.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the comments. The manuscript presents a methodological proposal for using evolutionary algorithms to identify minimal card attribute changes that balance simulated deck win rates near 50%, along with supporting metrics. We address the points below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the description of the evolutionary search for 50% win-rate balance and the multi-objective minimal-change extension supplies no results, validation experiments, error analysis, or implementation details, making it impossible to determine whether the approach supports the stated claims.

    Authors: We agree that the current manuscript supplies no experimental results, error analysis, or out-of-sample validation. The work is framed as an algorithmic proposal with a description of the multi-objective extension and heuristic metrics. In revision we will add implementation details (e.g., evolutionary parameters, representation of card changes) and at least one illustrative run of the search to demonstrate feasibility. revision: yes

  2. Referee: [Abstract] Abstract: the central claim requires that simulated win rates from fixed deck match-ups serve as a faithful proxy for how attribute tweaks affect emergent player behavior and long-term meta evolution, yet the manuscript provides no calibration against historical patch data, player logs, or adaptive-strategy tests to support this assumption.

    Authors: We acknowledge that the manuscript offers no calibration or validation of the fixed-deck simulation proxy against historical patches or adaptive play. This assumption is implicit in the approach. The revision will include an explicit limitations paragraph stating the assumption and suggesting future calibration steps (e.g., retrospective comparison with known balance patches). revision: yes

Circularity Check

0 steps flagged

No circularity: evolutionary search optimizes simulation-defined objective without self-referential reduction.

full rationale

The paper applies an evolutionary algorithm to search for card attribute changes that drive simulated deck win rates toward 50%, with a multi-objective extension minimizing the number of changes. This is a standard optimization procedure whose objective (win-rate balance) is defined externally by the simulation match-ups; no step equates a derived quantity to its own fitted input, renames a known result, or relies on a self-citation chain for a uniqueness claim. The abstract and described method contain no equations or parameters that are presented as predictions yet reduce by construction to the search inputs. The central claim therefore remains independent of the patterns that trigger circularity flags.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract alone provides no information on free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5720 in / 1040 out tokens · 64006 ms · 2026-05-25T10:40:04.626190+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages · 1 internal anchor

  1. [1]

    Hearthbot: An autonomous agent based on fuzzy art adaptive neural networks for the digital collectible card gamehearthstone,

    A. R. da Silva and L. F. W. Goes, “Hearthbot: An autonomous agent based on fuzzy art adaptive neural networks for the digital collectible card gamehearthstone,” IEEE Transactions on Games , vol. 10, no. 2, pp. 170–181, 2018

    work page 2018

  2. [2]

    Predicting opponent moves for improving hearthstone ai,

    A. Dockhorn, M. Frick, ¨U. Akkaya, and R. Kruse, “Predicting opponent moves for improving hearthstone ai,” in International Conference on Information Processing and Management of Uncertainty in Knowledge- Based Systems. Springer, 2018, pp. 621–632

    work page 2018

  3. [3]

    Helping ai to play hearthstone using neural networks,

    Ł. Grad, “Helping ai to play hearthstone using neural networks,” in 2017 federated conference on computer science and information systems (FedCSIS). IEEE, 2017, pp. 131–134

    work page 2017

  4. [4]

    Monte carlo tree search experiments in hearthstone,

    A. Santos, P. A. Santos, and F. S. Melo, “Monte carlo tree search experiments in hearthstone,” in Computational Intelligence and Games (CIG), 2017 IEEE Conference on . IEEE, 2017, pp. 272–279

    work page 2017

  5. [5]

    Symbolic reasoning for hearthstone,

    A. Stiegler, K. Dahal, J. Maucher, and D. Livingstone, “Symbolic reasoning for hearthstone,” IEEE Transactions on Computational In- telligence and AI in Games , 2017

    work page 2017

  6. [6]

    Improving hearthstone ai by combining mcts and supervised learning algorithms,

    M. ´Swiechowski, T. Tajmajer, and A. Janusz, “Improving hearthstone ai by combining mcts and supervised learning algorithms,” in 2018 IEEE Conference on Computational Intelligence and Games (CIG) . IEEE, 2018, pp. 1–8

    work page 2018

  7. [7]

    Improving hearthstone ai by learning high-level rollout policies and bucketing chance node events,

    S. Zhang and M. Buro, “Improving hearthstone ai by learning high-level rollout policies and bucketing chance node events,” in Computational Intelligence and Games (CIG), 2017 IEEE Conference on. IEEE, 2017, pp. 309–316

    work page 2017

  8. [8]

    Exploring the hearthstone deck space,

    A. Bhatt, S. Lee, F. de Mesentier Silva, C. W. Watson, J. Togelius, and A. K. Hoover, “Exploring the hearthstone deck space,” in Proceedings of the 13th International Conference on the Foundations of Digital Games . ACM, 2018, p. 18

    work page 2018

  9. [9]

    Mapping hearthstone deck spaces with map-elites with sliding boundaries,

    M. F. S. Lee, L. B. S. F. de Mesentier Silva, and J. T. A. K. Hoover, “Mapping hearthstone deck spaces with map-elites with sliding boundaries,” in To be published at Proceedings of The Genetic and Evolutionary Computation Conference . ACM, 2019, p. 8

    work page 2019

  10. [10]

    Evolutionary deckbuilding in hearthstone,

    P. Garc ´ıa-S´anchez, A. Tonda, G. Squillero, A. Mora, and J. J. Merelo, “Evolutionary deckbuilding in hearthstone,” in Computational Intelli- gence and Games (CIG), 2016 IEEE Conference on . IEEE, 2016, pp. 1–8

    work page 2016

  11. [11]

    Automated playtesting in collectible card games using evolutionary algorithms: A case study in hearthstone,

    P. Garc ´ıa-S´anchez, A. Tonda, A. M. Mora, G. Squillero, and J. J. Merelo, “Automated playtesting in collectible card games using evolutionary algorithms: A case study in hearthstone,” Knowledge-Based Systems , vol. 153, pp. 133–146, 2018

    work page 2018

  12. [13]

    Q-deckrec: A fast deck recommendation system for collectible card games,

    Z. Chen, C. Amato, T.-H. D. Nguyen, S. Cooper, Y . Sun, and M. S. El- Nasr, “Q-deckrec: A fast deck recommendation system for collectible card games,” in 2018 IEEE Conference on Computational Intelligence and Games (CIG) . IEEE, 2018, pp. 1–8

    work page 2018

  13. [14]

    Hearthstone deck-construction with a utility system,

    A. Stiegler, C. Messerschmidt, J. Maucher, and K. Dahal, “Hearthstone deck-construction with a utility system,” in Software, Knowledge, Infor- mation Management & Applications (SKIMA), 2016 10th International Conference on. IEEE, 2016, pp. 21–28

    work page 2016

  14. [15]

    Deckbuilding in magic: The gathering using a genetic algorithm,

    S. J. Bjørke and K. A. Fludal, “Deckbuilding in magic: The gathering using a genetic algorithm,” Master’s thesis, Norwegian University of Science and Technology (NTNU), 2017

    work page 2017

  15. [16]

    Evaluation of hearthstone game states with neural networks and sparse autoencoding,

    J. Jakubik, “Evaluation of hearthstone game states with neural networks and sparse autoencoding,” in Computer Science and Information Systems (FedCSIS), 2017 Federated Conference on . IEEE, 2017, pp. 135–138

    work page 2017

  16. [17]

    Toward an intelligent hs deck advisor: Lessons learned from aaia18 data mining competition,

    A. Janusz, T. Tajmajer, M. ´Swiechowski, Ł. Grad, J. Puczniewski, and D. ´Slezak, “Toward an intelligent hs deck advisor: Lessons learned from aaia18 data mining competition,” in 2018 Federated Conference on Computer Science and Information Systems (FedCSIS). IEEE, 2018, pp. 189–192

    work page 2018

  17. [18]

    I am a legend: Hacking hearthstone using statistical learning methods,

    E. Bursztein, “I am a legend: Hacking hearthstone using statistical learning methods,” in Computational Intelligence and Games (CIG), 2016 IEEE Conference on . IEEE, 2016, pp. 1–8

    work page 2016

  18. [19]

    Investigating similarity between hearthstone cards: Text embeddings and interchangeability approaches,

    A. Janusz and D. Slezak, “Investigating similarity between hearthstone cards: Text embeddings and interchangeability approaches,” in 2018 IEEE International Conference on Systems, Man, and Cybernetics (SMC). IEEE, 2018, pp. 3421–3426

    work page 2018

  19. [20]

    Proposed balance model for card deck measurement in hearth- stone,

    Y . Jin, “Proposed balance model for card deck measurement in hearth- stone,” The Computer Games Journal , pp. 1–16, 2018

    work page 2018

  20. [21]

    Mystical tutor: A magic: The gathering design assistant via denoising sequence-to-sequence learning,

    A. J. Summerville and M. Mateas, “Mystical tutor: A magic: The gathering design assistant via denoising sequence-to-sequence learning,” in Twelfth Artificial Intelligence and Interactive Digital Entertainment Conference, 2016

    work page 2016

  21. [22]

    Ai-based playtesting of contemporary board games,

    F. de Mesentier Silva, S. Lee, J. Togelius, and A. Nealen, “Ai-based playtesting of contemporary board games,” in Proceedings of the 12th International Conference on the Foundations of Digital Games . ACM, 2017, p. 13

    work page 2017

  22. [23]

    Ex- ploring gameplay with ai agents,

    F. D. M. Silva, I. Borovikov, J. Kolen, N. Aghdaie, and K. Zaman, “Ex- ploring gameplay with ai agents,” in Fourteenth Artificial Intelligence and Interactive Digital Entertainment Conference , 2018

    work page 2018

  23. [24]

    Evolving card sets towards balancing dominion,

    T. Mahlmann, J. Togelius, and G. N. Yannakakis, “Evolving card sets towards balancing dominion,” in 2012 IEEE Congress on Evolutionary Computation. IEEE, 2012, pp. 1–8

    work page 2012

  24. [25]

    Evaluating competitive game balance with restricted play,

    A. Jaffe, A. Miller, E. Andersen, Y .-E. Liu, A. Karlin, and Z. Popovic, “Evaluating competitive game balance with restricted play,” in Pro- ceedings of the Eighth Artificial Intelligence and Interactive Digital Entertainment International Conference (AIIDE 2012) , 2012

    work page 2012

  25. [26]

    Demonstrating the feasibility of automatic game balancing,

    V . V olz, G. Rudolph, and B. Naujoks, “Demonstrating the feasibility of automatic game balancing,” in Proceedings of the 2016 on Genetic and Evolutionary Computation Conference . ACM, 2016, pp. 269–276

    work page 2016

  26. [27]

    Algorithmically balancing a collectible card game,

    J. Krucher, “Algorithmically balancing a collectible card game,” Bach- elors Thesis. ETH Zurich , 2015

    work page 2015

  27. [28]

    Learning how design choices impact gameplay behavior,

    A. Zook and M. Riedl, “Learning how design choices impact gameplay behavior,” IEEE Transactions on Games , 2018

    work page 2018

  28. [29]

    Magic lessons: Designing and balancing game ob- jects,

    K. R. Gutschera, “Magic lessons: Designing and balancing game ob- jects,” http://twvideo01.ubm-us.net/o1/vault/gdc07/slides/S3709i2.pdf, accessed: 2019-03-24

    work page 2019

  29. [30]

    Rarity and power: balance in collectible object games,

    E. Ham, “Rarity and power: balance in collectible object games,” The International Journal of Computer Game Research, vol. 10, no. 1, 2010

    work page 2010

  30. [31]

    Card changes,

    “Card changes,” https://hearthstone.gamepedia.com/Card changes, ac- cessed: 2019-03-24

    work page 2019

  31. [32]

    Illuminating search spaces by mapping elites

    J.-B. Mouret and J. Clune, “Illuminating search spaces by mapping elites,” arXiv preprint arXiv:1504.04909 , 2015

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  32. [33]

    A fast elitist non- dominated sorting genetic algorithm for multi-objective optimization: Nsga-ii,

    K. Deb, S. Agrawal, A. Pratap, and T. Meyarivan, “A fast elitist non- dominated sorting genetic algorithm for multi-objective optimization: Nsga-ii,” in International conference on parallel problem solving from nature. Springer, 2000, pp. 849–858

    work page 2000

  33. [34]

    Generating beginner heuristics for simple texas holdem,

    F. de Mesentier Silva, J. Togelius, F. Lantz, and A. Nealen, “Generating beginner heuristics for simple texas holdem,” 2018

    work page 2018

  34. [35]

    Generating novice heuristics for post-flop poker,

    ——, “Generating novice heuristics for post-flop poker,” in 2018 IEEE Conference on Computational Intelligence and Games (CIG) . IEEE, 2018, pp. 1–8

    work page 2018

  35. [36]

    Non-cooperative games,

    J. Nash, “Non-cooperative games,” Annals of mathematics, pp. 286–295, 1951

    work page 1951