- Physical Review Journals
All JournalsPhysics Magazine

APS Open Science

Reuse & Permissions

It is not necessary to obtain permission to reuse this article or its components as it is available under the terms of the Creative Commons Attribution 4.0 International license. This license permits unrestricted use, distribution, and reproduction in any medium, provided attribution to the author(s) and the published article's title, journal citation, and DOI are maintained. Please note that some figures may have been included with permission from other third parties. It is your responsibility to obtain the proper permission from the rights holder directly for these figures.

Export citation

Export citation

Choose format for download:

Download Citation
  • Featured in Physics
  • Editors' Suggestion
  • Open Access

Artificial intelligence sustains higher strategic tension than humans in chess

Adamo Cerioli1, Edward D. Lee2,3,*, and Vito D. P. Servedio2,†

  • 1Department of Mathematics, Physics and Computer Science, University of Parma, Parco Area delle Scienze, 7/A, 43124 Parma, Italy
  • 2Complexity Science Hub, Metternichgasse 8, 1030 Vienna, Austria
  • 3Institute for Forest Ecology, BOKU, Peter-Jordan-Straße 82, 1190 Vienna, Austria

  • *Present address: Department of Physics and Astronomy & Center for Theoretical Physics, Seoul National University, Seoul 08826, Republic of Korea.
  • †Contact author: servedio@csh.ac.at
PDF

APS Open Sci. 1, 000131 – Published 11 September, 2026

DOI: https://doi.org/10.1103/63cv-52fj

Abstract

Strategic decision-making requires balancing immediate opportunities against long-term objectives: a tension fundamental to competitive environments. We investigate this trade-off in chess by analyzing the dynamics of human and AI (artificial intelligence) gameplay through a network-based metric that quantifies piece-to-piece interactions. Our analysis reveals that elite AI players sustain substantially higher levels of strategic tension for longer durations than top human grandmasters. We find that cumulative tension scales with algorithmic complexity in AI systems and increases linearly with skill level (Elo rating) in human play. Longer time controls are associated with higher tension in human games, reflecting the additional strategic complexity players can manage with more thinking time. The temporal profiles reveal contrasting approaches: Highly competitive AI systems tolerate densely interconnected positions that balance offensive and defensive tactics over extended periods, while human players systematically limit tension and game complexity. These differences have broader implications for understanding how artificial and biological systems navigate complex strategic environments and for the deployment of AI in high-stakes competitive scenarios.

View figure in article

Physics Subject Headings (PhySH)

Focus

Chess Reveals Differences Between Human and AI Strategies

Published 11 September, 2026

Chess grandmasters move toward a less complex game faster than AI players, suggesting different approaches for balancing short- and long-term prospects.

See more in Physics

Article Text

References (66)

  1. J. Von Neumann and O. Morgenstern, Theory of Games and Economic Behavior: 60th Anniversary Commemorative Edition (Princeton University Press, Princeton, NJ, 2007).
  2. H. A. Simon, A behavioral model of rational choice, Q. J. Econ. 69, 99 (1955).
  3. R. Bellman, Dynamic programming, Science 153, 34 (1966).
  4. J. M. Weinstein, Inside Rebellion: The Politics of Insurgent Violence, Cambridge Studies in Comparative Politics (Cambridge University Press, Cambridge, 2007).
  5. J. C. Bohorquez, S. Gourley, A. R. Dixon, M. Spagat, and N. F. Johnson, Common ecology quantifies human insurgency, Nature (London) 462, 911 (2009).
  6. J. M. Smith and G. R. Price, The logic of animal conflict, Nature (London) 246, 15 (1973).
  7. S. DeDeo, D. C. Krakauer, and J. C. Flack, Inductive game theory and the dynamics of animal conflict, PLoS Comput. Biol. 6, e1000782 (2010).
  8. E. R. Brush, D. C. Krakauer, and J. C. Flack, Conflicts of interest improve collective computation of adaptive social structures, Sci. Adv. 4, e1603311 (2018).
  9. E. van der Woude, J. Groothuis, and H. M. Smid, No gains for bigger brains: Functional and neuroanatomical consequences of relative brain size in a parasitic wasp, J. Evol. Biol. 32, 694 (2019).
  10. A. Pitera, C. Branch, E. Bridge, and V. Pravosudov, Daily foraging routines in food-caching mountain chickadees are associated with variation in environmental harshness, Anim. Behav. 143, 93 (2018).
  11. J. C. Magee and A. D. Galinsky, 8 Social hierarchy: The self-reinforcing nature of power and status, Acad. Manage. Ann. 2, 351 (2008).
  12. K. N. Laland, J. Odling-Smee, and M. W. Feldman, Cultural niche construction and human evolution, J. Evol. Biol. 14, 22 (2001).
  13. E. D. Lee, J. C. Flack, and D. C. Krakauer, Constructing stability: Optimal learning in noisy ecological niches, Proc. R. Soc. B 291, 20241606 (2024).
  14. R. Ratcliff and J. N. Rouder, Modeling response times for two-choice decisions, Psychol. Sci. 9, 347 (1998).
  15. C. E. Shannon, Programming a computer for playing chess, Philos. Mag. Ser. 7 41, 256 (1950).
  16. H. A. Simon and J. Schaeffer, The game of chess, in Handbook of Game Theory with Economic Applications, edited by R. J. Aumann and S. Hart (Elsevier, Amsterdam, Netherlands, 1992), Vol. 1, pp. 1–17.
  17. M. H. Connors, B. D. Burns, and G. Campitelli, Expertise in complex decision making: The role of search in chess 70 years after de Groot, Cognit. Sci. 35, 1567 (2011).
  18. A. J. Puddephatt, Chess playing as strategic activity, Symbolic Interact. 26, 263 (2003).
  19. A. Kiesel, W. Kunde, C. Pohl, M. P. Berner, and J. Hoffmann, Playing chess unconsciously, J. Exp. Psychol. 35, 292 (2009).
  20. D. M. Lane and Y.-H. A. Chang, Chess knowledge predicts chess memory even after controlling for chess experience: Evidence for the role of high-level processes, Mem. Cognit. 46, 337 (2018).
  21. D. Kahneman, Thinking, Fast and Slow (Farrar, Straus and Giroux, New York, NY, 2011).
  22. D. Farren, D. Templeton, and M. J. Wang, Analysis of networks in chess, Technical Report, Stanford University, Stanford, CA, 2013.
  23. A loop consists of a closed sequence of interactions between pieces forming a cycle within the network (e.g., a white rook on b4 defends a white knight on e4; the knight, in turn, attacks a black bishop on d6, which attacks the rook on b4).
  24. M. Barthelemy, Fragility of chess positions: Measure, universality, and tipping points, Phys. Rev. E 111, 014314 (2025).
  25. S. U. Pillai, T. Suel, and S. Cha, The Perron-Frobenius theorem: Some of its applications, IEEE Signal Process. Mag. 22, 62 (2005).
  26. C. Castellano and R. Pastor-Satorras, Relating topological determinants of complex networks to their spectral properties: Structural and dynamical effects, Phys. Rev. X 7, 041024 (2017).
  27. R. L. Adler, A. G. Konheim, and M. H. McAndrew, Topological entropy, Trans. Am. Math. Soc. 114, 309 (1965).
  28. P. Van Mieghem, Graph Spectra for Complex Networks (Cambridge University Press, Cambridge, UK, 2023), Chap. 4.
  29. Although we treat all links equally for the main macroscopic metric, we also performed a structural deconstruction of the network. As we will show, isolating specific subsets of links (e.g., generating attack-only or attack-and-defense networks) yields qualitatively similar tension trajectories.
  30. https://www.pgnmentor.com.
  31. https://tcec-chess.com/.
  32. A. D. De Groot, Thought and Choice in Chess (Walter de Gruyter GmbH & Co. KG, Berlin, Germany, 2014), Vol. 4.
  33. G. De Marzo et al., Quantifying the complexity and similarity of chess openings with a network approach, Sci. Rep. 13, 5327 (2023).
  34. M. Campbell, A. J. Hoane, and F.-H. Hsu, Deep Blue, Artif. Intell. 134, 57 (2002); it includes discussion of Deep Blue’s use of extensive opening databases and grandmaster opening theory.
  35. J. Silman, How to Reassess Your Chess: Chess Mastery Through Imbalances, 4th ed. (Siles Press, Los Angeles, CA, 2010); it discusses the practical rule that when ahead in material, one should generally trade pieces, not pawns, to reduce counterplay.
  36. In principle, Stockfish has a search depth parameter that serves as a proxy for skill level. Leela has a time limit parameter during which it is permitted to assess the board state and make a decision. At a few dozen search depths, the tension load is limited to about 450, similarly for Leela at a few seconds. Since the games that are shown go far beyond these limited parameter values with supercomputers, this suggests that algorithmic complexity increases the ability to handle tension.
  37. A. A. Kotov, Think Like a Grandmaster (Batsford Books, London, 2012); it describes grandmasters’ deliberate search for complex positions and the need to calculate deeply in prolonged struggles.
  38. J. Nunn, Secrets of Practical Chess (Gambit Publications, London, 1997); it discusses practical decisions at GM level, including when to keep the game complex instead of simplifying into drawish positions.
  39. A. D. De Groot, Thought and Choice in Chess (Mouton, The Hague, Netherlands, 1965).
  40. R. Calderwood, G. A. Klein, and B. W. Crandall, Time Pressure, Skill, and Move Quality in Chess, Am. J. Psychol. 101, 481 (1988).
  41. J. Carow and N. Witzig, Time pressure and strategic risk-taking in professional chess, J. Econ. Behav. Organ. 238, 107218 (2025); working paper version 2024, Johannes Gutenberg University Mainz Discussion Paper 2404.
  42. D. R. Ferreira, The impact of the search depth on chess playing strength, ICGA J. 36, 67 (2013).
  43. G. Peri, L. Buffoni, G. Chiti, D. Fanelli, R. Marino, A. Nocentini, and P. P. Panti, Smart walkers in discrete space, Chaos, Solitons Fractals 208, 118356 (2026).
  44. R. Savit, R. Manuca, and R. Riolo, Adaptive competition, market efficiency, and phase transitions, Phys. Rev. Lett. 82, 2203 (1999).
  45. N. Charness, Expertise in chess: The balance between knowledge and search, in Toward a General Theory of Expertise, edited by K. A. Ericsson and J. Smith (Cambridge University Press, Cambridge, UK, 1992), pp. 39–63.
  46. H. Berliner, The System: A World Champion’s Approach to Chess (Deep Chess Press, New York, 1999); it emphasizes board control and the option principle, i.e., making moves that maximize future options and facilitate exploiting opponent errors.
  47. T. Küchelmann et al., Expertise-dependent perceptual performance in chess tasks with varying complexity, Front. Psychol. 13, 986787 (2022).
  48. P. Chassy and F. Gobet, Measuring chess experts’ single-use sequence knowledge: An archival study of departure from ‘theoretical’ openings, PLoS ONE 6, e26692 (2011).
  49. F. Gobet, Understanding Expertise: A Multidisciplinary Approach (Palgrave Macmillan, London, 2016).
  50. R. S. Lazarus, How emotions influence performance in competitive sports, Sport Psychol. 14, 229 (2000).
  51. V. de Lafuente, Flexible decisions and chess expertise, Front. Neurosci. 5, 4 (2011).
  52. E. C. Snell-Rood, D. R. Papaj, and W. Gronenberg, Brain size: A global or induced cost of learning? Brain Behav. Evol. 73, 111 (2009).
  53. J. R. Burger, M. A. George, C. Leadbetter, and F. Shaikh, The allometry of brain size in mammals, J. Mammal. 100, 276 (2019).
  54. D. Silver et al., Mastering the game of go without human knowledge, Nature (London) 550, 354 (2017).
  55. R. McIlroy-Young, S. Sen, J. Kleinberg, and A. Anderson, Aligning superhuman AI with human behavior: Chess as a model system, in KDD’20: Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (ACM Press, New York, NY, 2020), p. 1677.
  56. T. C. Schelling, Arms and Influence (Yale University Press, New Haven, CT, 1966).
  57. R. Simmons-Edler, R. P. Badman, S. Longpre, and K. Rajan, Position: AI-powered autonomous weapons risk geopolitical instability and threaten AI research, in Proceedings of the 41St International Conference on Machine Learning, Vol. 235 (PMLR, Vienna, Austria, 2024), pp. 45524–45548.
  58. R. Jervis, Cooperation under the security dilemma, World Pol. 30, 167 (1978).
  59. K. Guttieri, M. D. Wallace, and P. Suedfeld, The integrative complexity of American decision makers in the Cuban Missile Crisis, J. Conflict Resolut. 39, 595 (1995).
  60. K. Sakuwa and W. R. Thompson, On the origins, persistence and termination of spatial and positional rivalries in world politics: Elaborating a two-issue theory of conflict escalation, Int. Area Stud. Rev. 22, 203 (2019).
  61. https://lichess.org/.
  62. https://github.com/Draghellon/Tension_Chess.
  63. R. Bowen, Entropy for group endomorphisms and homogeneous spaces, Trans. Am. Math. Soc. 153, 401 (1971).
  64. E. I. Dinaburg, A correlation between topological entropy and metric entropy, Dokl. Akad. Nauk SSSR 190, 19 (1970).
  65. Stockfish originally relied on an HCE function, a traditional linear combination of human-defined heuristic features such as material balance, king safety, and piece-square tables. Modern versions achieve grandmaster-level positional understanding by replacing HCE with NNUE, which evaluate board positions via a shallow neural network trained on millions of positions. Disabling NNUE and restricting the engine to HCE allows us to isolate a more rudimentary, feature-driven form of tactical assessment.
  66. J. R. Steenhuisen, New results in deep-search behaviour, ICGA J. 28, 203 (2005).

Outline

Information

APS Open Sci. 1, 000131– Published 11 September, 2026
Vol. 1
  • Received 15 February 2026
  • Accepted 17 July 2026
Crossmark - Check for updates button
Creative Commons Logo

Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.

Published by the American Physical Society

Sign In to Your Journals Account

Filter

Filter

Article Lookup

Enter a citation