A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
Open accessThe DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index; 5 of the 16 sources its claims cite were checked for support.
Abstract
One program to rule them all: Computers can beat humans at increasingly complex games, including chess and Go. However, these programs are typically constructed for a particular game, exploiting its properties, such as the symmetries of the board on which it is played. Silver et al. developed a program called AlphaZero, which taught itself to play Go, chess, and shogi (a Japanese version of chess) (see the Editorial, and the Perspective by Campbell). AlphaZero managed to beat state-of-the-art programs specializing in these three games. The ability of AlphaZero to adapt to various game rules is a notable step toward achieving a general game-playing system. Science , this issue p. 1140 ; see also pp. 1087 and 1118
Overview
CoChat Summary
TL;DR
AlphaZero is a single reinforcement-learning and MCTS algorithm that, given only game rules, learned superhuman play in chess, shogi, and Go through self-play. It defeated world-champion programs Stockfish, Elmo, and AlphaGo Zero using the same algorithm and network architecture across all three games.
Key findings
- In chess, AlphaZero defeated Stockfish 8, winning 155 games and losing 6 out of 1,000; it also won matches from common human openings and 2016 TCEC opening positions.
- In shogi, AlphaZero defeated Elmo, winning 98.2% of games as black and 91.2% overall; in Go, it defeated AlphaGo Zero, winning 61% of games.
- AlphaZero first outperformed Stockfish after 4 hours (300,000 steps), Elmo after 2 hours (110,000 steps), and AlphaGo Lee after 30 hours (74,000 steps); training ran for 700,000 steps, taking about 9 hours in chess, 12 hours in shogi, and 13 days in Go.
- AlphaZero searched only about 60,000 positions per second in chess and shogi, versus 60 million for Stockfish and 25 million for Elmo, yet still defeated Stockfish with 1/10 thinking time and won 46% against Elmo with 1/100 thinking time.
Methods
- AlphaZero uses a deep neural network \(f_\theta(s)\) that outputs move probabilities \(p\) and a scalar value \(v\), trained by reinforcement learning from self-play starting from random parameters; MCTS uses these outputs to guide search, with 800 simulations per move during training.
- The same algorithm, network architecture, and hyperparameters were used for chess, shogi, and Go, with no domain knowledge except game rules; no opening book, endgame tablebases, or domain-specific heuristics were used.
- Training used 5,000 first-generation TPUs to generate self-play games and 16 second-generation TPUs to train networks; evaluation matches used 3 hours per game plus 15 seconds per move, with opponents on 44 CPU cores and AlphaZero on four first-generation TPUs and 44 CPU cores.
Limitations
- Stated: AlphaZero did not use opening books, endgame tablebases, or domain-specific heuristics; the authors note that some such techniques could further improve performance but were left for future research.
- Stated: hyperparameters were reused across games without game-specific tuning, except for exploration noise and the learning rate schedule; matches against an opening book were low in diversity, and forced randomization increased losses.
- Evident from the text: results cover only three games with separate trained instances, no cross-game transfer is demonstrated, and evaluation compares AlphaZero on TPU hardware against CPU-based opponents, so hardware and time-control differences qualify direct strength comparisons.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
3 Valid · 2 Not valid · 8 Not checked · 43 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Apart from the representation of positions and actions described above, AlphaZero uses the same network architecture as AlphaGo Zero ( 9 ), briefly recapitulated here.”
What the sources say
- Supportedcites [9] source
The Neural Network Architecture Comparison (p. 29) identifies AlphaGo Zero’s architecture as a residual tower with policy and value heads. The citation supplies the AlphaGo Zero architecture that the citing authors say AlphaZero uses; the comparison to AlphaZero is their own statement. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“and in Go, AlphaZero first outperformed AlphaGo Lee ( 9 ) after 30 hours (74,000 steps).”
What the sources say
- Partly supportedcites [9] sourceExperimental — accuracy not yet measured
Section 2, p. 7 reports that AlphaGo Zero outperformed AlphaGo Lee, but says this happened after 36 hours, not 30. The passage does not give 74,000 steps, and it names the system AlphaGo Zero rather than AlphaZero. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The training algorithm achieved similar performance in all independent runs (see fig. S3), suggesting that the high performance of AlphaZero's training algorithm is repeatable.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“AlphaZero also defeated Stockfish when given 1 / 10 as much thinking time as its opponent (i.e. searching ~ 1 / 10 , 000 as many positions),”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“and won 46% of games against Elmo when given 1 / 100 as much time (i.e. searching ~ 1 / 40 , 000 as many positions), see Fig. 2.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
23 more claims
Create a free account to see all 56 claims and to verify the 3 references still open.
- Probably not a substantial claim. 10
- It may state two separate claims as one. 23
- The quote only partly states this claim. 1
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 source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Games have historically served as highly effective metrics for evaluating progress in deep reinforcement learning research”
From BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games (2024) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In 2017, AlphaGo, using the same learning algorithms in TD-Gammon and a billion times more computing power, soundly beat Ke Jie, the world Go champion, in a game that is to chess what chess is to backgammon in complexity (Silver et al., 2018).”
From Large Language Models and the Reverse Turing Test (2023) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Most of the attention was on who won, but more important was how it won: AlphaGo was creative and had discovered new moves and board positions that were outside human experience.”
From Large Language Models and the Reverse Turing Test (2023) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The primary method for applying more computation at inference, common in the AlphaZero family of agents2,72, is scaling the tree search and running multiple independent search attempts.”
From Olympiad-level formal mathematical reasoning with reinforcement learning (2026) · cites this paper as [2]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Modern AI can write computer programs (Li et al., 2022) or play chess at super-human level (Silver et al., 2018), but the ability of AI to perceive and act in the world remains far below human level.”
From Scaling Instructable Agents Across Many Simulated Worlds (2024) · cites this paper as Silver et al., 2018
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Citations
Cites
Mastering the game of Go with deep neural networks and tree search
Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., et al. (2016) · Nature · doi:10.1038/nature16961 · Cited by 15,612
Self-citationAdvances in neural information processing systems 7
(1996) · Computers & Mathematics with Applications · doi:10.1016/s0898-1221(96)90463-0 · Cited by 14,281
Identity mappings in deep residual networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_38 · Cited by 10,741
Identity Mappings in Deep Residual Networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · arXiv · doi:10.48550/arxiv.1603.05027 · Cited by 10,665
Mastering the game of Go without human knowledge
Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., Bolton, A., et al. (2017) · Nature · doi:10.1038/nature24270 · Cited by 9,084
Self-citationIn-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537
Encyclopedia of Artificial Intelligence
(2009) · IGI Global eBooks · doi:10.4018/978-1-59904-849-9 · Cited by 1,396
Campbell, M., Hoane, A. J., & Hsu, F.-H. (2002) · Artificial Intelligence · doi:10.1016/s0004-3702(01)00129-1 · Cited by 1,226
Lim, T.-S., Loh, W.-Y., & Shih, Y.-S. (2000) · Machine Learning · doi:10.1023/a:1007608224229 · Cited by 1,133
An analysis of alpha-beta pruning
Knuth, D. E., & Moore, R. W. (1975) · Artificial Intelligence · doi:10.1016/0004-3702(75)90019-3 · Cited by 1,004
Research works printed but not linked: 7
- C. J. Maddison A. Huang I. Sutskever D. Silver paper presented at the International Conference on Learning Representations 2015 San Diego CA 7 to 9 May 2015.
- D. N. L. Levy M. Newborn How Computers Play Chess (Ishi Press 2009).
- W. Steinitz The Modern Chess Instructor (Edition Olms 1990).
- O. Arenz “Monte Carlo chess ” master’s thesis Technische Universität Darmstadt (2012).
- 10.1023/a:1007634325138 · DOI
- T. Kaneko K. Hoki in Advances in Computer Games – 13th International Conference ACG 2011 Revised Selected Papers Tilburg Netherlands 20 to 22 November 2011 (Springer 2012) pp. 158–169.
- R. Ramanujan A. Sabharwal B. Selman in Proceedings of the 26th Conference on Uncertainty in Artificial Intelligence (UAI 2010) Catalina Island CA 8 to 11 July (AUAI Press 2010).
Cited by
Items 1,601–1,700 of 3,633
Qin, T. (2020) · Springer eBooks · doi:10.1007/978-981-15-8884-6_2 · Cited by 13
Gerdes, A. (2020) · Advances in intelligent systems and computing · doi:10.1007/978-3-030-63128-4_19 · Cited by 9
On the Redundancy in the Rank of Neural Network Parameters and Its Controllability
Lee, C., Kim, Y., Ji, H., Lee, Y., Hur, Y., & Lim, H. (2021) · Applied Sciences · doi:10.3390/app11020725 · Cited by 10
“Recently, neural networks have shown great success in a wide range of tasks, such as achieving superhuman performance in the game of Go [1] and Atari games [2,3], photorealistic image synthesis [4], and many natural language processing applications [5–9].”
BackgroundPredicting the Future With a Scale-Invariant Temporal Memory for the Past
Goh, W. Z., Ursekar, V., & Howard, M. W. (2022) · Neural Computation · doi:10.1162/neco_a_01475 · Cited by 9
“…processes (e.g., Watkins and Dayan, 1992; Sutton, 1988) have been extraordinarily successful in accounting for reward systems in the brain (e.g., Schultz et al., 1997; Waelti et al., 2001) and led to remarkable achievements in artificial intelligence (e.g., Mnih et al., 2015; Silver et al., 2018).”
BackgroundDobriborsci, D., Osinenko, P., & Aumer, W. (2022) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2022.09.610 · Cited by 9
AI, IoT, and Blockchain: Business Models, Ethical Issues, and Legal Perspectives
Nehme, E., Salloum, H., Abdo, J. B., & Taylor, R. (2021) · Springer eBooks · doi:10.1007/978-3-030-74150-1_4 · Cited by 8
A Deep Learning Framework for Self-evolving Hierarchical Community Detection
Ding, D., Zhang, M., Wang, H., Pan, X., Yang, M., & He, X. (2021) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3459637.3482223 · Cited by 5
Iconary: A Pictionary-Based Game for Testing Multimodal Communication with Drawings and Text
Clark, C., Salvador, J., Schwenk, D., Bonafilia, D., Yatskar, M., Kolve, E., Herrasti, A., Choi, J., Mehta, S., Skjonsberg, S., et al. (2021) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2021.emnlp-main.141 · Cited by 13
Autonomous Learning in a Pseudo-Episodic Physical Environment
Haughn, K. P. T., & Inman, D. J. (2022) · Journal of Intelligent & Robotic Systems · doi:10.1007/s10846-022-01577-5 · Cited by 6
“…super-human performance in Atari game environments, Dota 2, and Go, as well as many simulated physics environments offered by the OpenAI libraries and MuJoCo to name a 1 The University of Michigan Department of Aerospace Engineering, 1320 Beal Ave, Ann Arbor, MI 48109, USA few [6, 27, 28, 35, 36].”
BackgroundSolving PBQP-Based Register Allocation using Deep Reinforcement Learning
Kim, M., Park, J.-K., & Moon, S. (2022) · Proceedings of the CGO/Proceedings of the ... CGO · doi:10.1109/cgo53902.2022.9741272 · Cited by 7
“That is, Equation 2 seeks a proper trade-off between exploitation and exploration using c puct [29]–[31].”
BackgroundXue, W., An, B., & Yeo, C. K. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i4.20389 · Cited by 9
Fabrication of a tactile sensor for artificial skin based on electrical impedance tomography
Liu, K., Qin, Z., Wu, Y., Chen, B., Li, F., Pan, H., & Yao, J. (2022) · Biosensors and Bioelectronics X · doi:10.1016/j.biosx.2022.100116 · Cited by 8
Rodríguez, S., Zaroukian, E. G., Hoye, J., & Asher, D. E. (2022) · Journal of Cognitive Engineering and Decision Making · doi:10.1177/15553434221129166 · Cited by 11
“…and reinforcement learning policies are often benchmarked in strategy-based games (Bard et al., 2019; Campbell et al., 2002; Gibney, 2016; Silver et al., 2018; Vinyals et al., 2017), with modern human-AI interaction and collaboration paradigms using game-based tasks as well (Ashktorab et…”
MethodChordal Sparsity for Lipschitz Constant Estimation of Deep Neural Networks
Xue, A., Lindemann, L., Robey, A., Hassani, H., Pappas, G. J., & Alur, R. (2022) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc51059.2022.9993136 · Cited by 28
“, in solving the game Go [1] and in handwritten character recognition [2].”
BackgroundToward Autonomous Cyber Defense for Protected Core Networking
Velazquez, A., Mathews, J., Lopes, R. R. F., Braun, T., & Nelson, F. F. (2023) · International Conference on Military Communications and Information Systems (ICMCIS) · doi:10.1109/icmcis59922.2023.10253614 · Cited by 7
“MCTS has been successfully used in two-player games such as Go [28] and chess [29] to determine the best next move based on a particular state of the game board.”
MethodRoute selection for a three-dimensional elevator using deep reinforcement learning
Hammoudeh, A. (2019) · Building Services Engineering Research and Technology · doi:10.1177/0143624419876079 · Cited by 6
Liu, X., Luping, F., Khalid, M. N. A., et al. (2020) · Entertainment Computing · doi:10.1016/j.entcom.2020.100358 · Cited by 6
Campbell, M. S. (2018) · Science · doi:10.1126/science.aav1175 · Cited by 7
Embracing curiosity eliminates the exploration-exploitation dilemmaPreprint
Peterson, E. J., & Verstynen, T. D. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/671362 · Cited by 15
Uncertainty-sensitive Learning and Planning with EnsemblesPreprint
Miłoś, P., Kuciński, Ł., Czechowski, K., Kozakowski, P., & Klimek, M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1912.09996 · Cited by 8
“This led, for example, to strong AI systems like Silver et al. (2017; 2018) or more recently to Lowrey et al. (2018), which is closely related to our work.”
BackgroundMethodInfluentialDeep Model-Based Reinforcement Learning for High-Dimensional Problems, a SurveyPreprint
Plaat, A., Kosters, W. A., & Preuß, M. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2008.05598 · Cited by 48
“Three wellknown examples are Samuel’s checkers player [Samuel, 1959], Tesauro’s Backgammon player [Tesauro, 1995a, 2002] and DeepMind’s Alpha(Go) Zero [Silver et al., 2017a, 2018].”
BackgroundInfluentialContinuous Doubly Constrained Batch Reinforcement LearningPreprint
Fakoor, R., Mueller, J. W., Asadi, K., Chaudhari, P., & Smola, A. J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2102.09225 · Cited by 28
“[55] D.”
BackgroundContinuous Coordination As a Realistic Scenario for Lifelong LearningPreprint
Nekoei, H., Badrinaaraayanan, A., Courville, A. C., & Chandar, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.03216 · Cited by 41
“Deep reinforcement learning (RL) has shown an immense potential to achieve superhuman performance (Mnih et al., 2013; Silver et al., 2018) on some narrow and well-defined”
BackgroundDeep Learning: Implications for Human Learning and MemoryPreprint
McClelland, J. L., & Botvinick, M. (2020) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/3m5sb · Cited by 6
Methods for small area population forecasts: state-of-the-art and research needsPreprint
Wilson, T. G., Grossman, I., Alexander, M. J., et al. (2021) · doi:10.31235/osf.io/sp6me · Cited by 6
Continuous Control with Deep Reinforcement Learning for Autonomous VesselsPreprint
Zare, N., Brandoli, B., Sarvmaili, M., Soares, A., & Matwin, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.14130 · Cited by 16
“Since then DRL has been vastly employed for robotic control [8, 12, 24], strategic games [11, 20, 21], and autonomous car navigation [10, 17].”
MethodMEPG: A Minimalist Ensemble Policy Gradient Framework for Deep Reinforcement LearningPreprint
He, Q., Su, H., Gong, C., & Hou, X. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.10552 · Cited by 12
“…function approximation to solve sequential decision-making problems, and has achieved great success in a wide range of domains such as board games [Silver et al., 2018], video games [Mnih et al., 2015b, Vinyals et al., 2019a], robot manipulation [Haarnoja et al., 2018], etc. DRL algorithms are…”
MethodNature over Nurture: Functional neuronal circuits emerge in the absence of developmental activityPreprint
Barabási, D. L., Schuhknecht, G. F. P., & Engert, F. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.10.24.513526 · Cited by 6
Guo, G., Guo, L., Qian, J., He, X., Qian, X., Wang, L., & Huang, Q. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.01.657174 · Cited by 6
“Similar to AlphaZero 31 , AiDTA uses the encoding matrix of a given sequence s as the input to the policy-value neural network: f θ ( s ).”
Measuring Progress in Deep Reinforcement Learning Sample Efficiency
Dorner, F. E. (2021) · arXiv · doi:10.48550/arxiv.2102.04881 · Cited by 6
Gaze-informed signatures of trust and collaboration in human-autonomy teams
Ries, A. J., Aroca-Ouellette, S., Roncone, A., & de Visser, E. J. (2025) · Computers in Human Behavior Artificial Humans · doi:10.1016/j.chbah.2025.100171 · Cited by 12
RiskMiner: Discovering Formulaic Alphas via Risk Seeking Monte Carlo Tree Search
Ren, T., Zhou, R., Jiang, J., Liang, J., Wang, Q., & Peng, Y. (2024) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3677052.3698613 · Cited by 11
Applying opponent and environment modelling in decentralised multi-agent reinforcement learning
Chernyavskiy, A., Skrynnik, A., & Panov, A. I. (2024) · Cognitive Systems Research · doi:10.1016/j.cogsys.2024.101306 · Cited by 11
Space Trajectory Planning with a General Reinforcement-Learning Algorithm
Forestieri, A., & Casalino, L. (2025) · Aerospace · doi:10.3390/aerospace12040352 · Cited by 5
“Building on these advancements, DeepMind subsequently unveiled AlphaZero [31], a generalization of the AlphaGo Zero paradigm.”
Wang, S., Yue, Q., Xu, Z., Qiao, P., Lyu, Z., & Gao, F. (2025) · Mathematics · doi:10.3390/math13111738 · Cited by 8
Human–AI collaboration for modeling heat conduction in nanostructures
Ding, W., Guo, J., An, M., Tsuda, K., & Shiomi, J. (2025) · npj Computational Materials · doi:10.1038/s41524-025-01657-8 · Cited by 8
Quantum Circuit Design using a Progressive Widening Enhanced Monte Carlo Tree Search
Lipardi, V., Dibenedetto, D., Stamoulis, G., & Winands, M. H. M. (2025) · Advanced Quantum Technologies · doi:10.1002/qute.202500093 · Cited by 13
“Monte Carlo Tree Search (MCTS) is a best-first search method that has achieved notable results on several search-based problems, especially in games [45, 46, 42].”
Generalized synchronization between two distinct chaotic systems through deep reinforcement learning
Cheng, H., Li, H., Liang, J., Dai, Q., & Yang, J. (2025) · Chaos Solitons & Fractals · doi:10.1016/j.chaos.2025.116727 · Cited by 6
Lassoued, S., Baheti, L. S., Weiß-Borkowski, N., Lier, S., & Schwung, A. (2025) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2025.06.017 · Cited by 8
“Among approximate optimization approaches, RL, particularly with deep reinforcement learning (DRL) techniques that combine RL with deep neural networks, has gained significant attention in solving scheduling problems [28].”
Control Oriented Reinforcement Learning: A Survey of Recent Progress and Applications
Wang, X., Zhang, H., Liu, H., & Lewis, F. L. (2025) · International Journal of Robust and Nonlinear Control · doi:10.1002/rnc.70152 · Cited by 9
“It learns the target policy from collected data based on trial-and-error [11], and has many successfullyapplicationsinthefieldofartificialintelligence,such as AlphaZero [12] and AlphaStar [13].”
Li, F., Tian, Y., Noortwyck, R., Zhou, J., Kuang, L., & Schulz, R. (2025) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-025-02654-w · Cited by 9
“…in various decision-making fields, such as discovering novel matrix multiplication method (Fawzi et al., 2022), mastering strategy video games (Silver et al., 2018; Tian et al., 2023), controlling robotic behaviors (Han et al., 2024), as well as solving challenging planning (Esteso et al.,…”
Hu, H., Wu, W., Zhang, J., Wang, J., & Song, Y. (2025) · Astrodynamics · doi:10.1007/s42064-024-0252-2 · Cited by 5
“…learning (RL), deep neural networks (DNNs) and their combination, that is, deep reinforcement learning (DRL), have demonstrated rapid development and excellent performance in several fields, such as gaming [7, 8], path planning for different vehicles [9–12], and orbital pursuit–evasion gaming [13].”
Jin, X., & Li, J. (2025) · International Journal of Consumer Studies · doi:10.1111/ijcs.70133 · Cited by 12
Li, J., Wan, B., & Song, Z. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.113489 · Cited by 8
Wauters, M. M., & van Nieuwenburg, E. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-022-00535-y · Cited by 5
Deep Instance Segmentation and Visual Servoing to Play Jenga with a Cost-Effective Robotic System
Marchionna, L., Pugliese, G., Martini, M., Angarano, S., Salvetti, F., & Chiaberge, M. (2023) · Sensors · doi:10.3390/s23020752 · Cited by 6
Stockfish or Leela Chess Zero? A Comparison Against Endgame Tablebases
Sadmine, Q. A., Husna, A., & Müller, M. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-54968-7_3 · Cited by 10
“The original AlphaZero paper [19] compared the performance of AlphaZero with Stockfish for chess in terms of gameplay.”
BackgroundMethodResultInfluentialLearning Logic Specifications for Policy Guidance in POMDPs: an Inductive Logic Programming Approach
Meli, D., Castellini, A., & Farinelli, A. (2024) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.15826 · Cited by 19
M3ARL: Moment-Embedded Mean-Field Multi-Agent Reinforcement Learning for Continuous Action Space
Tang, H., Hu, Y., Zhao, F., Yan, J., Dong, T., & Ding, W. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10448058 · Cited by 6
“With the blooming of deep reinforcement research in recent years, multi-agent reinforcement learning (MARL) has enjoyed great achievements in strategic games [1, 2], multi-player computer games [3, 4], autonomous driving [5], and emergent behaviors [6, 7].”
BackgroundMultiagent Gumbel MuZero: Efficient Planning in Combinatorial Action Spaces
Hao, X., Hao, J., Xiao, C., Li, K., Li, D., & Zheng, Y. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i11.29121 · Cited by 5
“AlphaZero family algorithms have achieved great success in many applications, such as playing board games like Go and chess (Silver et al. 2016, 2017, 2018; Schrittwieser et al. 2020), making planning decisions in robotics and autonomous vehicles (Lenz, Kessler, and Knoll 2016), predicting the…”
BackgroundMethodImproved learning efficiency of deep Monte-Carlo for complex imperfect-information card games
Luo, Q., & Tan, T.-P. (2024) · Applied Soft Computing · doi:10.1016/j.asoc.2024.111545 · Cited by 7
Vallée, R., Gomez, T., Bourreille, A., Normand, N., Mouchère, H., & Coutrot, A. (2024) · Journal of Vision · doi:10.1167/jov.24.4.6 · Cited by 5
“…Received September 23, 2022; published April 8, 2024 ISSN 1534-7362 Copyright 2024 The Authors (Badia et al., 2020), various board games (Silver et al., 2018), the characterization of precise bacteria (Buetti-Dinh et al., 2019), and medical diagnosis (Abràmoff, Lavin, Birch, Shah, &…”
Sample Trajectory Selection Method Based on Large Language Model in Reinforcement Learning
Lai, J., & Zang, Z. (2024) · IEEE Access · doi:10.1109/access.2024.3395457 · Cited by 8
Vibration Control with Reinforcement Learning Based on Multi-Reward Lightweight Networks
Shu, Y., He, C., Qiao, L., Xiao, B., & Li, W. (2024) · Applied Sciences · doi:10.3390/app14093853 · Cited by 7
MiniZero: Comparative Analysis of AlphaZero and MuZero on Go, Othello, and Atari Games
Wu, T.-R., Guei, H., Peng, P.-C., Huang, P.-W., Wei, T.-H., Shih, C.-C., & Tsai, Y.-J. (2024) · IEEE Transactions on Games · doi:10.1109/tg.2024.3394900 · Cited by 29
“Each model is evaluated against a specific baseline model with Elo ratings [3], which is chosen to be neither excessively strong nor overly weak, for better discrimination.”
G-PCGRL: Procedural Graph Data Generation via Reinforcement Learning
Rupp, F., & Eckert, K. (2024) · IEEE Conference on Games (CoG) · doi:10.1109/cog60054.2024.10645633 · Cited by 13
“In the games community, RL has been widely applied to play games [25], [26], but recently also for PCG tasks [5], [27] as well.”
BackgroundTransfer learning to scale deep Q networks in the context of airline pricing
Nataraj, S., Varghese, J., Adarsh, R., Muralidhar, A., Joseph, E., Menon, R., & Westermann, D. (2024) · Journal of Revenue and Pricing Management · doi:10.1057/s41272-024-00493-7 · Cited by 4
“RL has shown notable promise in the field of robotics and gaming-achiev-ing par-human levels of performance in many use cases, including complex games such as Smash Bros and Alpha Go (Firoiu et al., (2017), Silver et al. (2018)).”
Su, P., Lin, X., Lu, W., Xiong, F., Peng, Z., & Lu, Y. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.125167 · Cited by 5
Competing for Pixels: A Self-Play Algorithm for Weakly-Supervised Semantic Segmentation
Saeed, S. U., Huang, S., Ramalhinho, J., Gayo, I. J. M. B., Montaña-Brown, N., Bonmati, E., Pereira, S. P., Davidson, B. R., Barratt, D. C., Clarkson, M. J., et al. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3474094 · Cited by 7
“This not only promotes faster learning compared to learning against a hand-crafted strategy, but also allows higher diversity in experiences for potentially improved generalisability [3], [4], [28], [29], [79], [80].”
InfluentialAI-powered mechanisms as judges: Breaking ties in chess
Anbarcı, N., & Ismail, M. S. (2024) · PLoS ONE · doi:10.1371/journal.pone.0305905 · Cited by 8
Ren, H., & Qureshi, A. H. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10801607 · Cited by 10
“The AlphaGo, AlphaZero [28]–[30] took advantage of MCTS and beat the top-ranked human go player in 2016.”
An open-source reproducible chess robot for human-robot interaction research
Zhang, R., de Winter, J., Dodou, D., Seyffert, H. C., & Eisma, Y. B. (2025) · Frontiers in Robotics and AI · doi:10.3389/frobt.2025.1436674 · Cited by 14
“Furthermore, chess has played an important role in advancing the field of AI, starting with Claude Shannon’s chess-playing algorithm (Shannon, 1950) to the success of IBM’s Deep Blue (Campbell et al., 2002) and DeepMind’s self-play learning algorithm (Silver et al., 2018).”
Galactic Center Shadows: Beyond the Standard Model
ZAKHAROV, A. F. (2025) · Physics of Atomic Nuclei · doi:10.1134/s106377882570019x · Cited by 5
Yang, Z., Yang, Q., & Lu, J. W. Z. (2025) · Advanced Engineering Informatics · doi:10.1016/j.aei.2025.103510 · Cited by 11
A taxonomy of spatial navigation in mammals: Insights from computational modeling
Vijayabaskaran, S., Zeng, X., Ghazinouri, B., Wiskott, L., & Cheng, S. (2025) · Neuroscience & Biobehavioral Reviews · doi:10.1016/j.neubiorev.2025.106282 · Cited by 7
Wang, D., & Liu, L. (2020) · ACM/IEEE Joint Conference on Digital Libraries · doi:10.1145/3383583.3398500 · Cited by 9
LBA: Online Learning-Based Assignment of Patients to Medical Professionals
Rosemarin, H., Rosenfeld, A., Lapp, S., & Kraus, S. (2021) · Sensors · doi:10.3390/s21093021 · Cited by 11
“MCTS has proven itself capable of achieving state-of-the-art performance in modeling both online dynamic behavior and the long-term effect of an agent’s decisions in complex environments such as the games of Go and Chess [57].”
BackgroundLuger, G. F. (2021) · Springer eBooks · doi:10.1007/978-3-030-71873-2_3 · Cited by 8
Information Bottleneck Theory Based Exploration of Cascade Learning
Du, X., Farrahi, K., & Niranjan, M. (2021) · Entropy · doi:10.3390/e23101360 · Cited by 6
Modular Networks Prevent Catastrophic Interference in Model-Based Multi-task Reinforcement Learning
Schiewer, R., & Wiskott, L. (2022) · Lecture notes in computer science · doi:10.1007/978-3-030-95470-3_23 · Cited by 6
“In the presence of a readily available analytical model of the environment, that does not have to be learned, perfect planning can be used to learn a policy with great success [29,31].”
BackgroundLearning to Play Pursuit-Evasion with Visibility Constraints
Engin, S., Jiang, Q.-Y., & Isler, V. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9635959 · Cited by 5
“Recent advancements in Reinforcement Learning (RL) have led to progress in many different fields, including Atari games [1] and Go [2].”
BackgroundDeep Learning and the Game of Checkers
Popič, J., Boškovič, B., & Brest, J. (2021) · MENDEL · doi:10.13164/mendel.2021.2.001 · Cited by 6
“Deepminds programs AlphaGo [7], AlphaGo Zero [9] and Alpha Zero [8] are one such example of using a deep convolutional neural networks as part of a bigger algorithm that can learn to play a game of Go (as well as chess and shogi with Alpha Zero).”
BackgroundMethodInfluentialSEM: Safe exploration mask for q-learning
Xuan, C., Zhang, F., & Lam, H. (2022) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2022.104765 · Cited by 6
A Study of Neural Training with Iterative Non-Gradient MethodsPreprint
Karmakar, S., Mukherjee, A., & Muthukumar, R. (2021) · SSRN Electronic Journal · doi:10.1016/j.neunet.2022.03.040 · Cited by 8
“Human world champions of classic hard board games have famously been defeated by neural net-based approaches, the [1, 2, 3, 4].”
BackgroundOn Transitions Functions Model for Decision-Making in Offshore Operations
Mihai, R. G., Daireaux, B., Ambrus, A., Cayeux, E., & Carlsen, L. A. (2022) · International Conference on Control and Automation · doi:10.1109/icca54724.2022.9831911 · Cited by 6
Engineering Laboratory Experiments – a Typology
Zwart, S. D. (2022) · Engineering Studies · doi:10.1080/19378629.2022.2126783 · Cited by 5
Neuro-Inspired Deep Neural Networks with Sparse, Strong Activations
Cekic, M., Bakiskan, C., & Madhow, U. (2022) · Proceedings - International Conference on Image Processing · doi:10.1109/icip46576.2022.9897876 · Cited by 7
A Q-based policy gradient optimization approach for Doudizhu
Yu, X., Wang, Y., Qin, J., & Chen, P. (2022) · Applied Intelligence · doi:10.1007/s10489-022-04281-x · Cited by 5
“It extends AlphaGo Zero [28] to a multiplayer incomplete information game.”
BackgroundAdaptive cooperative exploration for reinforcement learning from imperfect demonstrations
Huang, F., Ji, N., Ni, H., Li, S., & Li, X. (2022) · Pattern Recognition Letters · doi:10.1016/j.patrec.2022.12.003 · Cited by 5
Optimal Control of Traffic Flow Based on Reinforcement Learning
Baumgart, U., & Burger, M. (2022) · Communications in computer and information science · doi:10.1007/978-3-031-17098-0_16 · Cited by 7
Distributional Actor-Critic Ensemble for Uncertainty-Aware Continuous Control
Kanazawa, T., Wang, H., & Gupta, C. (2022) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn55064.2022.9892771 · Cited by 11
“RL has already shown excellent performance in computer games [2] and even surpassed human champions in Go, Chess and Shogi [3].”
BackgroundLearning to play an imperfect information card game using reinforcement learning
Demirdover, B. K., Baykal, O., & Alpaslan, F. N. (2022) · TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES · doi:10.55730/1300-0632.3940 · Cited by 5
“Later in AlphaZero [12], the methods used in AlphaGo Zero were generalized so that the agent can learn more games simultaneously.”
MethodRole of reinforcement learning for risk‐based robust control of cyber‐physical energy systems
Du, Y., Chatterjee, S., Bhattacharya, A., Dutta, A., & Halappanavar, M. M. (2023) · Risk Analysis · doi:10.1111/risa.14104 · Cited by 5
“Model-free deep Q-network The model-free deep Q-network (DQN) algorithm was developed by DeepMind (Mnih et al., 2015).”
MethodSeismic profile denoising based on common-reflection-point gathers using convolution neural networks
Li, S., Zhang, J., Cheng, Q., Zhu, F., & Liu, L. (2023) · Journal of Geophysics and Engineering · doi:10.1093/jge/gxad008 · Cited by 5
Cai, H. G., Luo, Y., Gao, H., Chi, J., & Wang, S. F. (2023) · Computational Intelligence and Neuroscience · doi:10.1155/2023/2955442 · Cited by 5
Augmenting Multi-Turn Text-to-SQL Datasets with Self-Play
Liu, Q., Ye, Z., Yu, C., Song, L., & Blunsom, P. (2022) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2022.findings-emnlp.411 · Cited by 15
Scheduling Real-Time Wireless Traffic: A Network-Aided Offline Reinforcement Learning Approach
Wan, J., Lin, S., Zhang, Z., Zhang, J., & Zhang, T. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3304969 · Cited by 7
“RL seeks to learn the optimal policy that maximizes a long-term reward by interacting with the environment for the MDP problem, which has achieved astonishing successes in many applications, such as robotics [11], [12] and games [13], [14], [15].”
BackgroundA Unified Framework of Ethical Principles for AI
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0004 · Cited by 10
Load Frequency Control with Deep Reinforcement Learning under Adversarial Attacks
Liu, X., Jiao, Q., & Yan, Z. (2023) · IEEE Conference on Industrial Electronics and Applications (ICIEA) · doi:10.1109/iciea58696.2023.10241803 · Cited by 9
“At the same time, the application of the DRL model also includes real-time economic dispatch of electric vehicles [20], [21], grid voltage control [22], dynamic distribution network reconfiguration [18], and demand-based energy price mechanism [24].”
MethodHow Terminology Affects Users’ Responses to System Failures
Candrian, C., & Scherer, A. (2023) · Human Factors The Journal of the Human Factors and Ergonomics Society · doi:10.1177/00187208231202572 · Cited by 11
Toward Artificial General Intelligence: Deep Reinforcement Learning Method to AI in Medicine
Nguyen, D. S. W., & Odigie, R. (2023) · Journal of Computer and Communications · doi:10.4236/jcc.2023.119006 · Cited by 5
Understanding Human-AI Teaming Dynamics through Gaming Environments
Zhang, Q. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment · doi:10.1609/aiide.v19i1.27541 · Cited by 5
Enhancing Stockfish: A Chess Engine Tailored for Training Human Players
Manzo, A., & Ciancarini, P. (2023) · Lecture notes in computer science · doi:10.1007/978-981-99-8248-6_23 · Cited by 9
Mirage: Towards Low-interruption Services on Batch GPU Clusters with Reinforcement Learning
Ding, Q. L., Zheng, P., Kudari, S., Venkataraman, S., & Zhang, Z. (2023) · International Conference for High Performance Computing, Networking, Storage and Analysis (SC) · doi:10.1145/3581784.3607042 · Cited by 11
“These recent innovations motivate us to design a smart resource provisioner using RL techniques, and particularly in our framework, deep Q network [34] (DQN) and policy gradient (PG) [43], which we show can help HPC practitioners perform adaptive, proactive resource provisioning.”
BackgroundMethodInfluentialReinforcement Learning in an Adaptable Chess Environment for Detecting Human-understandable Concepts
Hammersborg, P., & Strümke, I. (2023) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2023.10.135 · Cited by 13
Mastering Spatial Graph Prediction of Road Networks
Anagnostidis, S., Lucchi, A., & Hofmann, T. F. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00498 · Cited by 8
“Un-visited graph nodes are selected based on an upper confidence score, balancing exploration with exploitation, similar to [51].”
BackgroundGeneration of probabilistic synthetic data for serious games: A case study on cyberbullying
Pérez, J., Castro, M., Awad, E., & López, G. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.111440 · Cited by 10
Artificial Intelligence and Public Health Context: What We Should Know?
Madli, F., Janin, Y., Gulabdin, S. B., Lada, S., Yun, W. S., Adis, A.-A. A., & Jafar, A. (2024) · Journal of Advanced Research in Applied Sciences and Engineering Technology · doi:10.37934/araset.39.2.96109 · Cited by 8
SoRTS: Learned Tree Search for Long Horizon Social Robot Navigation
Navarro, I., Patrikar, J., Dantas, J. P. A., et al. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3370051 · Cited by 5
“Finally, the ego agent uses the updated statistics to select the action that maximizes the normalized visitation count as in [15].”
MethodSalience Interest Option: Temporal abstraction with salience interest functions
Zhu, X., Zhao, L., & Zhu, W. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106342 · Cited by 5
Metadata
- Authors
- Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., Lillicrap, T., Simonyan, K., & Hassabis, D.
- Affiliations
- Google DeepMind (United Kingdom) (GB)
- Corresponding author
- Silver, David
- Hubert, Thomas
- Schrittwieser, Julian
- Hassabis, Demis
- ORCID
- Venue
- Science
- Published
- 2018-12-07
- Volume
- 362
- Issue
- 6419
- Pages
- 1140-1144
- ISSN
- 0036-8075 (ISSN-L)
- 1095-9203
- OpenAlex
- W2902907165
- Cited by
- 3,562
- References
- 30
- Citations by year
- 2026: 257 · 2025: 427 · 2024: 512 · 2023: 514 · 2022: 569 · 2021: 546 · 2020: 462 · 2019: 256 · 2018: 13
- FWCI
- 179.89
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Science (publishedVersion)
- PubMed (publishedVersion)
- UCL Discovery (University College London) (submittedVersion)
- Funders
- DeepMind
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence in Games
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence Applications
- Keywords
- AlphaZero, chess, shogi, Go, evaluation function learning, game tree search
- MeSH terms
- Algorithms, Artificial Intelligence, Humans, Reinforcement, Psychology, Software, Video Games
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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.