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 2,401–2,500 of 3,633
Monte Carlo Search Algorithms Discovering Monte Carlo Tree Search Exploration Terms
Cazenave, T. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-86585-5_9 · Cited by 1
Bayesian curriculum generation in sparse reward reinforcement learning environments
Akgün, O., & Üre, N. K. (2025) · Engineering Science and Technology an International Journal · doi:10.1016/j.jestch.2025.102048 · Cited by 1
Hijji, M., Munsif, M., Alwakeel, M. M., Alwakeel, A. M., Aradah, F., Cheikh, F. A., Sajjad, M., & Muhammad, K. (2025) · Fractals · doi:10.1142/s0218348x25401711 · Cited by 1
Searching for ribbons with machine learning
Gukov, S., Halverson, J., Manolescu, C., et al. (2025) · Machine Learning Science and Technology · doi:10.1088/2632-2153/ade362 · Cited by 1
“We also implemented an adaptation of the AlphaZero Monte-Carlo tree search (MCTS) algorithm [SHS + 18] to work with single-player games.”
Прудков, П. Н. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2025.1588726 · Cited by 1
“Modern intelligent systems based on predetermined criteria of the optimization of a loss function belong to this architecture (Devlin et al., 2018; Silver et al., 2018; Vinyals et al., 2019; Krizhevsky et al., 2017 among others).”
Counter-productivity and suspicion: two arguments against talking about the AGI control problem
Stenseke, J. (2025) · Philosophical Studies · doi:10.1007/s11098-025-02379-9 · Cited by 2
“Examples of the narrow kind can be found in game play , where AI has mastered games such as Chess, Go, and Dota 2 better than any human player (Berner et al., 2019 ; Silver et al., 2018 ).”
Harshinni, B. (2025) · AIJR Proceedings · doi:10.21467/proceedings.7.4.4 · Cited by 1
Wang, Y., Lu, Z., Qian, P., Su, J., Zhou, M., Li, C. V., & Jiang, Z. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-9949-0_43 · Cited by 6
Demystifying MuZero Planning: Interpreting the Learned Model
Guei, H., Ju, Y.-R., Chen, W.-Y., & Wu, T.-R. (2025) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2025.3591082 · Cited by 4
“MuZero [5], a model-based reinforcement learning algo-rithm, builds upon AlphaZero [4] has been unrolled by the dynamics network for k steps.”
Gao, K., Liu, X., Hu, H., et al. (2024) · IEEE International Conference on High Performance Computing and Communications (HPCC) · doi:10.1109/hpcc64274.2024.00219 · Cited by 1
Human-Like Bots for Tactical Shooters Using Compute-Efficient Sensors
Justesen, N., Kaselimi, M., Snodgrass, S., Vozaru, M., Schlegel, M., Wingren, J., Barros, G. A. B., Mahlmann, T., Sudhakaran, S., Kerr, W., et al. (2025) · IEEE Transactions on Games · doi:10.1109/tg.2025.3593919 · Cited by 3
“In addition, academic models typically aim for optimal performance [1], whereas the industry prioritizes human-like behavior to improve the player experience [2], necessitating a different approach for defining reward functions.”
Artificial Intelligence Awareness Scale Development Study
Kaya, Ş. M., & Bayram, V. (2025) · OPUS Toplum Araştırmaları Dergisi · doi:10.26466/opusjsr.1680531 · Cited by 4
Efficient Neural Clause-Selection Reinforcement
Suda, M. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-99984-0_22 · Cited by 3
Expert Knowledge-Guided Deep Reinforcement Learning for Jiu Chess: A Hybrid Intelligence Approach
Liu, S., Yajie, W., Wu, Y., Xu, H., Feng, G., & Li, X. (2025) · Communications in computer and information science · doi:10.1007/978-981-95-0011-6_38 · Cited by 1
Decision-Making in Wargames: An E-CARGO Perspective
Li, Y., Sun, Y., Wang, M., et al. (2025) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2025.3589592 · Cited by 1
“Another approach is applying deep reinforcement learning (DRL), whose strategies have signifi-cantly surpassed human performance in complex competitive games such as Atari [4], chess [5], Go [6], Texas Hold’em [7], and StarCraft II [8].”
Gao, G., Li, Y., Yang, S., et al. (2025) · IET Smart Grid · doi:10.1049/stg2.70028 · Cited by 1
Synthesizing Evolving Symbolic Representations for Autonomous Systems
Sartor, G., Oddi, A., Rasconi, R., Santucci, V. G., & Meo, R. (2025) · Progress in Artificial Intelligence · doi:10.1007/s13748-025-00394-9 · Cited by 1
“These include real-world games such as chess [1] and Go [2–4], video games such as Atari [5], and Dota [6], as well as various robotics tasks [7–10].”
Human Choice Prediction in Language-based Persuasion Games: Simulation-based Off-Policy Evaluation
Shapira, E., Madmon, O., Apel, R., Tennenholtz, M., & Reichart, R. (2025) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl.a.16 · Cited by 7
“Simulations also play a crucial role in the development of artificial agents proficient in gaming scenarios, e.g. by us-ing MCTS-like simulations to enhance agent performance (Silver et al., 2018, 2017; Schrittwieser et al., 2020; Oroojlooy and Hajinezhad, 2022).”
Xiao, J., Qiu, C., Xu, Y., Zhang, J., Qi, S., & Wang, X. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.129496 · Cited by 2
Xiao, C., & Yang, Z. (2025) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3748634 · Cited by 16
“Therefore, a growing body of recent research explores additional possibilities for using LLMs in different aspects of game development.”
Luyu, J., Cai, C., Wang, X., & Zhengkun, D. (2025) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2025.3620244 · Cited by 1
“Intelligent decision-making has attracted significant attention, with researchers successfully applying deep reinforcement learning (DRL) to tackle various decision-making tasks in games, including Go, Atari, and video games [6], [7].”
Learning Roles With Emergent Social Value Orientations
Li, W., Wang, X., Jin, B., Lu, J., & Zha, H. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3620954 · Cited by 7
“Recently, multi-agent re-inforcement learning (MARL) has been employed as an effective toolset for studying human cooperative behavior with notable success [8], [9], [10], [11], [12], [13].”
Explain to Gain: Introspective Reinforcement Learning for Enhanced Performance
Quintana-Amate, S., Stevens, D., Ketheeswaran, V., Capaldo, P., Sheldon, D., & Hall, M. (2025) · Communications in computer and information science · doi:10.1007/978-3-032-08324-1_11 · Cited by 1
Moving towards automated game play-testing
Malosto, C. G. D. A., Campos, L. C. D., & Knop, I. D. O. (2025) · Simpósio Brasileiro de Jogos e Entretenimento Digital (SBGames) · doi:10.5753/sbgames_estendido.2025.10236 · Cited by 1
“AlphaZero employs a reinforcement learning algorithm for network training, where the agent improves by playing self-matches through a process called self-play [Silver et al. 2016, Silver et al. 2017, Silver et al. 2018].”
Rossi, L., & Winands, M. H. M. (2025) · AI in Civil Engineering · doi:10.1007/s43503-025-00073-7 · Cited by 1
“…of knowledge, having been used for other board games (Chaslot et al., 2008a, 2008b; Arneson et al., 2010; Mehat & Cazenave, 2011; Winands, 2017; Silver, 2018), chemical design applications (Segler et al., 2018), robotics (Clary et al., 2018; Kartal et al., 2016; Labbé et al., 2020), planning…”
Liu, H., Yu, Y., Tian, Y., & Dang, C. (2025) · Journal of Systems Science and Systems Engineering · doi:10.1007/s11518-025-5699-z · Cited by 1
Clustering-Based Weight Orthogonalization for Stabilizing Deep Reinforcement Learning
Ma, G., Zhang, Y., Dai, Y., Hao, G., Chen, Y., & Yu, S. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11227331 · Cited by 3
“In recent years, reinforcement learning (RL) has made significant progress across various domains, ranging from gaming to robotic control, often surpassing human performance [1]– [6].”
Zeng, Y., Lin, F., Zhang, L., Jiang, J., Zhang, C., Qin, D., & Zhang, Z. (2025) · Clinics · doi:10.1016/j.clinsp.2025.100843 · Cited by 1
Li, X., Fan, X., Yu, J., Dong, Z., Cairang, X., & Lan, P. (2025) · Frontiers of Information Technology & Electronic Engineering · doi:10.1631/fitee.2500287 · Cited by 2
Agentic AI: A Paradigm Shift in Autonomous Decision-Making and Intelligent Systems
Chahar, S., Gaba, P., Kaswan, K. S., Dhatterwal, J. S., & Baliyan, A. (2025) · IEEE International Conference on Recent Trends in Computing and Smart Mobility (RCSM) · doi:10.1109/aists66100.2025.11232814 · Cited by 3
Khodadadi, Z., Trpevski, D., Lindroos, R., & Kotaleski, J. H. (2024) · eLife · doi:10.7554/elife.97274.4 · Cited by 6
Outperforming Classical PI Tuning Methods via AlphaZero Algorithm
Queiroz, K. I. P. D. M., Dias, S. M., Oliveira, T. R., & Araújo, A. (2025) · Journal of Control Automation and Electrical Systems · doi:10.1007/s40313-025-01225-x · Cited by 1
Artificial intelligence for the new power system: Trends and challenges
TianShu, B. I., ZhaoHao, D. I. N. G., Hong, R. A. O., & Nian, L. I. U. (2025) · Scientia Sinica Technologica · doi:10.1360/sst-2025-0167 · Cited by 1
Impact of interaction diversity and interlayer coupling on the evolution of human-AI cooperation
Gao, B., Zuo, P., Xia, C., et al. (2025) · Applied Mathematics and Computation · doi:10.1016/j.amc.2025.129894 · Cited by 1
Benchmarking Deep Learning Architectures for Static Chess Position Classification
Muttaqin, M. Z., Indriani, F., Muliadi, et al. (2025) · International Conference on Informatics and Computing (ICIC) · doi:10.1109/icic68054.2025.11309414 · Cited by 1
“On the other, large-scale systems inspired by AlphaZero and transformer-based models can approach engine-level evaluation accuracy when trained with massive datasets and computational resources [3], [4].”
Athénan Wins 11 Gold Medals at the 2024 Computer Olympiad
Cohen-Solal, Q., & Cazenave, T. (2024) · ICGA Journal · doi:10.1177/13896911251315102 · Cited by 1
“Unlike Alpha Zero-like algorithms (Silver et al., 2018), Athénan is based on the Descent framework (Cohen-Solal, 2020).”
Wang, H., Zhang, X., & Mu, C. (2026) · Smart Cities · doi:10.3390/smartcities9020021 · Cited by 1
“In addition, based on the self-play training strategy, AlphaGo Zero [31] achieved superhuman performance on the game of Go.”
The use of AI in psychology: A historical perspective
O’Toole, A. J., & Ludvig, E. A. (2026) · British Journal of Psychology · doi:10.1111/bjop.70061 · Cited by 2
Relationship Between Artificial Intelligence and Carbon Footprint
Dere, G. (2026) · Springer eBooks · doi:10.1007/978-3-032-03718-3_2 · Cited by 3
Algorithmic and Human Collusion
Werner, T. (2021) · The Economic Journal · doi:10.1093/ej/ueag013 · Cited by 20
Integrating reinforcement learning with visual generative models: foundations and advances
Liang, Y., Fang, Y., Li, R., Ni, Z., Su, R., & Zhang, C. (2026) · Vicinagearth. · doi:10.1007/s44336-025-00030-z · Cited by 16
“Phase IV: The emergence of RL as a general substrate for decision-making [55], interfacing with planning [56], simulation [57], and generative modeling [7, 28].”
Budget-scalable inference-time hybrid MCTS for enhancing DRL-based flexible job shop scheduling
Zhang, Y., Dong, S., Wen, T., et al. (2026) · Computers & Industrial Engineering · doi:10.1016/j.cie.2026.111897 · Cited by 1
Yao, Q., Zhang, Q., Li, Q., Li, Z., Meng, L., Sun, Y., & Wang, C. (2026) · Neural Processing Letters · doi:10.1007/s11063-025-11826-x · Cited by 1
“Deep reinforcement learning (DRL) has shown remarkable performance in various decision problems such as games [37], chess [38], and path-planning tasks [39].”
Pradana, A., & Chew, L. Y. (2026) · Journal of Computational Science · doi:10.1016/j.jocs.2026.102807 · Cited by 2
Yang, T., Hou, J., Zhang, P., Hu, J., & Zhang, Y. (2026) · Neural Networks · doi:10.1016/j.neunet.2026.108735 · Cited by 1
An Empirical Study of Reinforcement Learning-based Class Integration Test Order Generation
Wang, J., Zheng, S., Zhang, M., et al. (2025) · Proceedings - Asia Pacific Software Engineering Conference/Proceedings · doi:10.1109/apsec66846.2025.00023 · Cited by 1
“As a powerful sequential decision-making tool, reinforcement learning (RL) has significant advantages in solving complex optimization problems [2, 3], attracting the attention of researchers.”
Abouelaziz, I., & Ghalmane, Z. (2026) · Multimedia Tools and Applications · doi:10.1007/s11042-026-21400-9 · Cited by 2
Robotic Arm Control Using a Q-Learning Reinforcement Algorithm
Timóteo, A. M., Barbosa, R. S., & Jesus, I. S. (2026) · Robotics · doi:10.3390/robotics15030050 · Cited by 1
What questions should a contemporary construct validation of an intelligence test address?
Sternberg, R. J. (2026) · Intelligence · doi:10.1016/j.intell.2026.102003 · Cited by 1
Opening the black box of neural variability: From noise to mechanisms
Rentzeperis, I., Li, J., Bauer, R., & van Leeuwen, C. (2026) · Neuroscience & Biobehavioral Reviews · doi:10.1016/j.neubiorev.2026.106661 · Cited by 2
Srdjević, B., & Srdjević, Z. (2026) · Water · doi:10.3390/w18080914 · Cited by 1
Curriculum reinforcement learning with measurable task representation learning
Wen, Y., Li, S., Fu, M., Yang, Y., Wang, X., & Liu, P. (2026) · Neural Networks · doi:10.1016/j.neunet.2026.109019 · Cited by 2
“Reinforcement learning (RL) [43] has emerged as a promising learning paradigm in challenging sequential decision-making tasks, including applications such as playing video games [32], chess [41], and robotic hand control [51, 1, 52, 12].”
Off by a beat: the effects of temporal misalignment in reinforcement learning for sepsis treatment
Tang, S., Yao, J., Wiens, J., & Parbhoo, S. (2026) · npj Digital Medicine · doi:10.1038/s41746-026-02625-2 · Cited by 4
A brief history of dopamine prediction errors
Dudhabhate, B. B., & Costa, K. M. (2026) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2026.1825622 · Cited by 2
“It was used to train algorithms to play complex games like back-gammon and chess at a master level (Tesauro, 1994; Baxter et al., 1998), and is a cornerstone of several modern machine learning and artificial intelligence algorithms (Mnih et al., 2015; Silver et al., 2018; Alkam et al., 2025).”
Reinforcement Learning for Autonomous Driving: Optimization Strategies and Methodologies
Moussa, A. B., & Khazari, A. (2026) · Springer proceedings in mathematics & statistics · doi:10.1007/978-3-032-18042-1_30 · Cited by 2
Mahdaviara, M., Sharifi, M., Rafiei, Y., & Raoof, A. (2026) · Energy and AI · doi:10.1016/j.egyai.2026.100834 · Cited by 1
Computational methods toward ultrastable glasses
Leoni, F., Ozawa, M., Russo, J., et al. (2026) · The Journal of Chemical Physics · doi:10.1063/5.0341332 · Cited by 1
Wang, C., Zhu, C., Lu, M., Bai, J., & Qiao, L. (2026) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2026.115720 · Cited by 1
Effective Online 3D Bin Packing with Lookahead Parcels Using Monte Carlo Tree Search
Fang, J. H., Zhou, B., Wang, H., et al. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770854.3783946 · Cited by 1
Survey: Six Years of Neural Differential Cryptanalysis
Gérault, D., Hambitzer, A., Huppert, M., & Picek, S. (2026) · IACR Communications in Cryptology · doi:10.62056/a69qxruc2 · Cited by 1
Policy Optimization over General State and Action Spaces
Ju, C., & Lan, G. (2026) · SIAM Journal on Optimization · doi:10.1137/25m1791585 · Cited by 4
Reinforcement Learning for decision support in defense and security: A systematic review
Schadd, M. P. D., Berman, D. S., Chen, C. M., Cohen, M., Dorsch, J., GEGOV, A. E., Moll, M., Rose, O., Rösner, A., Saastamoinen, K., et al. (2026) · Annals of Operations Research · doi:10.1007/s10479-026-07205-x · Cited by 5
Radiomics: A Well-Intentioned Leap of Faith
Cho, C. (2019) · Annals of Surgical Oncology · doi:10.1245/s10434-019-07818-6 · Cited by 1
“Yet, there were the data, with their impressive concordance index of 0.77, and if hair color would have improved the ability to predict one’s likelihood of succumbing to sarcoma, it would have been incorporated into the nomogram.1 Last year, DeepMind Technologies published the initial results of a new machine learning…”
BackgroundInfluentialTraining of Mixed-Signal Optical Convolutional Neural Networks With Reduced Quantization Levels
Zhu, Z., Ulseth, J., Li, G., & Pang, S. (2021) · IEEE Access · doi:10.1109/access.2021.3072193 · Cited by 1
Raising Ethical Machines: Bottom-Up Methods to Implementing Machine Ethics
Kaas, M. H. L. (2020) · Advances in human and social aspects of technology book series · doi:10.4018/978-1-7998-4894-3.ch004 · Cited by 6
Towards a Distributed Framework for Multi-Agent Reinforcement Learning Research
Zhou, Y., Manuel, S., Morales, P., Li, S., Peña, J., & Allen, R. E. (2020) · IEEE High Performance Extreme Computing Conference · doi:10.1109/hpec43674.2020.9286212 · Cited by 1
“Such environments typically exhibit several characteristics that emphasize well known weaknesses in reinforcement learning algorithms [45]–[47].”
BackgroundRoyale Heroes: A Unique RTS Game Using Deep Reinforcement Learning-based Autonomous Movement
Ramadhan, F., & Suyanto, S. (2020) · International Seminar on Research of Information Technology and Intelligent Systems (ISRITI) · doi:10.1109/isriti51436.2020.9315441 · Cited by 1
“It implements complete autonomous movements (AMs) [39].”
Trends and Emerging Technologies in AI
Hoppe, D. (2021) · Springer eBooks · doi:10.1007/978-3-030-68049-7_12 · Cited by 3
An Empirical Study of Artificial Participants
Bart, W. M., Ritter, J., & Ritter, N. (2021) · Journal of Technological Advancements · doi:10.4018/jta.20210101.oa2 · Cited by 1
“AlphaZero is the first chess software that incorporated artificial intelligence concepts (Silver et al., 2018).”
BackgroundAI tool makes phase identification crystal clear
Sun, W. Q., & Toney, M. F. (2021) · Nature Computational Science · doi:10.1038/s43588-021-00071-6 · Cited by 1
Reinforcement learning approach for deterministic SOT-MRAM switching
Ender, J., de Orio, R. L., Fiorentini, S., et al. (2021) · doi:10.1117/12.2593937 · Cited by 1
“, after initially having made breakthrough advances in learning how to play games like chess or Go.(10) Based on the results previously published in Orio et al.”
ResultReinforcement Learning Applications in Health Informatics
Takiddin, A., Elhissi, M., Abuhaliqa, S., & Yang, Y. (2021) · Lecture notes in bioengineering · doi:10.1007/978-3-030-67303-1_12 · Cited by 2
Algorithms for Mastering Board Game Nanahoshi Considering Deep Neural Networks
Saito, M., & Miwa, H. (2021) · Lecture notes in networks and systems · doi:10.1007/978-3-030-84910-8_7 · Cited by 1
Coordinated Control Method for Ridesharing Service Area Using Deep Reinforcement Learning
Yoshida, N., Noda, I., & Sugawara, T. (2021) · Transactions of the Japanese Society for Artificial Intelligence · doi:10.1527/tjsai.36-5_ag21-d · Cited by 1
FoLaR: Foggy Latent Representations for Reinforcement Learning with Partial Observability
Meisheri, H., & Khadilkar, H. (2021) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9534127 · Cited by 2
“This set of algorithms has produced remarkable successes in fully observable settings, both deterministic [1], [2] and stochastic [3].”
BackgroundImproving failure rates in pulsed SOT-MRAM switching by reinforcement learning
Ender, J., de Orio, R. L., Fiorentini, S., et al. (2021) · Microelectronics Reliability · doi:10.1016/j.microrel.2021.114231 · Cited by 1
Gapoera: Application programming interface for AI environment of Indonesian board game
Rajagede, R. A., & Mahardhika, G. P. (2023) · AIP conference proceedings · doi:10.1063/5.0114841 · Cited by 2
“In game development, research related to the development of intelligent agents that utilizes the Deep Learning method is carried out to facilitate single-player and multiplayer games [2]–[4].”
MethodFree Will Belief as a Consequence of Model-Based Reinforcement Learning
Rehn, E. M. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-19907-3_34 · Cited by 2
“A notable example of this is the board game playing system AlphaGo and its descendants (Silver et al., 2017; Silver et al., 2018) that use a hardcoded model of the world (the rules of the game) for planning and, when doing credit assignment, simply assign the same credit to all moves of a game…”
BackgroundEmotions as Abstract Evaluation Criteria in Biological and Artificial Intelligences
Gros, C. (2021) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2021.726247 · Cited by 5
“Consider f.i. the generic deep architectures AlphaGo (Silver et al., 2017) and AlphaZero (Silver et al., 2018), which consist of layered networks with two heads, one for the policy and one for the value, together Frontiers in Computational Neuroscience | www.frontiersin.org 4 December 2021 | Volume 15 | Article 726247…”
BackgroundChess fortresses, a causal test for state of the art Symbolic [Neuro] architectures
Steingrimsson, H. (2021) · IEEE Conference on Games (CoG) · doi:10.1109/cog52621.2021.9619010 · Cited by 7
Hyperparameter Choice as Search Bias in AlphaZero
Weiner, E. M., Montañez, G. D., Trujillo, A., et al. (2021) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc52423.2021.9659008 · Cited by 1
“exploitation trade-off, and empirical evidence suggests that the optimal explorationexploitation trade-off varies across different games [2].”
BackgroundMakarov, I., Kovaleva, M., Fakhrutdinov, T., & Gorbachev, R. (2021) · International Conference on Engineering and Telecommunication (EnT) · doi:10.1109/ent50460.2021.9681753 · Cited by 1
Artificial Intelligence and Public Health: Opportunities Abound
Jacobson, S. H., & Jokela, J. A. (2022) · Cambridge University Press eBooks · doi:10.1017/9781108872188.003 · Cited by 2
BALANCING INTRANSITIVE RELATIONSHIPS IN MOBA GAMES USING DEEP REINFORCEMENT LEARNING
Exton, C. S. A. C. (2020) · IADIS International Conferences Interfaces and Human Computer Interaction; Game and Entertainment Technologies; and Computer Graphics, Visualization, Computer Vision and Image Processing (IHCI/GET/CGVCVIP) · doi:10.33965/ihci_get2020_202010l016 · Cited by 1
“Self-Play would facilitate more stable training and allow the agents to adapt to different playstyles with the same characters.”
MethodTCEC12: The 12th Top Chess Engine Championship
Haworth, G., & Hernandez, N. (2019) · ICGA Journal · doi:10.3233/icg-190090 · Cited by 1
Modular Reinforcement Learning for Playing the Game of Tron
Jeon, M., Lee, J., & Ko, S. (2022) · IEEE Access · doi:10.1109/access.2022.3175299 · Cited by 4
Bracing for Artificial General Intelligence
Growiec, J. (2022) · Frontiers in economic history · doi:10.1007/978-3-031-07195-9_11 · Cited by 2
A Robust Offline Reinforcement Learning Algorithm Based on Behavior Regularization Methods
Zhang, Y., Gao, T., & Mi, Q. (2022) · IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology (IAICT) · doi:10.1109/iaict55358.2022.9887435 · Cited by 1
“Deep reinforcement learning is now widely used in a variety of domains, including video games [1], board games [2], robotic arm simulators [3], etc.”
BackgroundMjx: A framework for Mahjong AI research
Koyamada, S., Habara, K., Goto, N., Okano, S., Nishimori, S., & Ishii, S. (2022) · IEEE Conference on Games (CoG) · doi:10.1109/cog51982.2022.9893712 · Cited by 2
“Examples of reported AI progress through popular games include Backgammon [1], Chess [2], Go [3], [4], Shogi [5], Poker [6]– [8], and Atari 2600 [9].”
BackgroundStateless ARE: Action Recommendation Engine without Network State Measurement
Mohammed, A. R., Mohammed, S., Côté, D., & Shirmohammadi, S. (2022) · IEEE International Symposium on Measurements & Networking (M&N) · doi:10.1109/mn55117.2022.9887710 · Cited by 1
“Both could also have the superintelligent ability which has been demonstrated in computer games [17].”
BackgroundSize Scaling in Self-Play Reinforcement Learning
Neumann, O., & Gros, C. (2022) · The European Symposium on Artificial Neural Networks · doi:10.14428/esann/2022.es2022-53 · Cited by 1
“AlphaZero ( α Z) [1] is the extension of AlphaGo Zero, which attained human expert level performance at playing the game of Go.”
BackgroundMarkolf, L., & Stursberg, O. (2022) · SN Computer Science · doi:10.1007/s42979-022-01442-0 · Cited by 1
“The use of NN as function approximators for policies (control laws) and value (costto-go) functions, as in DeepMind’s popular computer program AlphaZero [23], is the core of numerous approximate dynamic programming and reinforcement learning algorithms [24–30].”
MethodAdvanced Mechanisms of Perception in the Digital Hide and Seek Game Based on Deep Learning
Livada, Č., & Hodak, D. (2022) · International Conference on Smart Systems and Technologies (SST) · doi:10.1109/sst55530.2022.9954814 · Cited by 1
“Reinforcement learning has been used to train artificial intelligence for games such as Chess and Go [2] and StarCraft [3].”
MethodDemocracy in the Age of Automation, Robotics, and Advanced AI
Bess, M. (2022) · Routledge eBooks · doi:10.4324/9781003218586-19 · Cited by 1
Contini, A., Piccolo, S. A., Zurita, L. L., & Šadl, U. (2022) · Frontiers in artificial intelligence and applications · doi:10.3233/faia220461 · Cited by 1
Sodobne perspektive družbe: Umetna inteligenca na stičišču znanosti
Bregant, J., Aberšek, B., & Borstner, B. (2022) · doi:10.18690/um.ff.11.2022 · Cited by 11
Nakagawa, Y., Ono, H., Hazui, Y., et al. (2022) · Journal of Process Control · doi:10.1016/j.jprocont.2022.12.010 · Cited by 1
Deep Reinforcement Learning with Parametric Episodic Memory
Chen, K., Gan, Z., Leng, S., & Guan, C. (2022) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn55064.2022.9891902 · Cited by 5
“Inspired by the AlphaGo series works [16], [23], [24], we consider using MCTS to obtain a distilled strategy.”
MethodHuman Language Explanation for a Decision Making Agent via Automated Rationale Generation
Nguyen, X.-B., Tran, T. H., Pham, N. B., Do, D. N., & Yairi, T. (2022) · IEEE Access · doi:10.1109/access.2022.3214323 · Cited by 1
“As in the previously mentioned article concerning AlphaGo [2], RL algorithms can generate moves that are beyond the scope of human understanding.”
BackgroundThe cost of passing – using deep learning AIs to expand our understanding of the ancient game of Go
Egri-Nagy, A., & Törmänen, A. (2022) · International Symposium on Computing and Networking · doi:10.1109/candar57322.2022.00019 · Cited by 1
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.