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 2,101–2,200 of 3,633
Hybrid knowledge transfer for MARL based on action advising and experience sharing
Liu, F., Li, D., & Gao, J. (2024) · Frontiers in Neurorobotics · doi:10.3389/fnbot.2024.1364587 · Cited by 4
Damage-aware structural control based on reinforcement learning
Jedlińska, A., Pisarski, D., Mikułowski, G., Błachowski, B., & Jankowski, Ł. (2024) · e-Journal of Nondestructive Testing · doi:10.58286/29606 · Cited by 4
Majumder, S., & Sahoo, S. R. (2024) · European Control Conference (ECC) · doi:10.23919/ecc64448.2024.10591259 · Cited by 2
“Another prominent example of DRL’s success is AlphaZero [2] that excels in strategic board games like chess, shogi, and go, rivaling and often surpassing human-level performance.”
BackgroundA framework for the emergence and analysis of language in social learning agents
Wieczorek, T. J., Tchumatchenko, T., Wert-Carvajal, C., & Eggl, M. F. (2024) · Nature Communications · doi:10.1038/s41467-024-51887-5 · Cited by 4
“Both of these agents are modeled as deep neural networks, whereby the teacher network is trained in an RL framework, and the student learnstointerprettheinstructionsoftheteacher 29,49,50 .”
AlphaMut: A Deep Reinforcement Learning Model to Suggest Helix-Disrupting Mutations
Bhargav, P., & Mukherjee, A. (2025) · Journal of Chemical Theory and Computation · doi:10.1021/acs.jctc.4c01387 · Cited by 2
From fair solutions to compromise solutions in multi-objective deep reinforcement learning
Qian, J., Siddique, U., Yu, G., & Weng, P. (2025) · Neural Computing and Applications · doi:10.1007/s00521-024-10602-7 · Cited by 5
“While RL has demonstrated remarkable success in various domains [1, 2], its application to real-world problems remains challenging due to various limitations, including the difficulty of reward engineering [3, 4], the assumption of single-objective optimization based on expected accumulated rewards…”
Is It Time to Start Taking the Concept of a Holodeck Seriously?
Berryman, Q., Dolan, M., Boutouria, V., Thomas, G., Johnston, A., Fleischmann, F., Xu, X., & Campbell, A. G. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-031-84457-7_14 · Cited by 2
Mashloosh, W. M. (2025) · INTERNATIONAL JOURNAL OF PROFESSIONAL STUDIES · doi:10.37648/ijps.v19i01.004 · Cited by 1
Exploring AI Techniques for Yoga Image Classification: A Focus on Machine Learning and Deep Learning
Duggal, A., Kumar, S., & Tripathi, P. (2025) · International Conference on Intelligent Systems and Computational Networks (ICISCN) · doi:10.1109/iciscn64258.2025.10934197 · Cited by 2
“Kinect-based models further improved precision, reporting 99.33% accuracy in recovering body shapes [15].”
InfluentialHybrids of Reinforcement Learning and Evolutionary Computation in Finance: A Survey
Yadav, S., Ravi, V., & Kalyanakrishnan, S. (2025) · ACM Computing Surveys · doi:10.1145/3728634 · Cited by 4
Artificial Intelligence for Research (AI4R): Knowledge Base and Impact
Yu, X., Xia, J., & Shen, Z. (2024) · International Conference on Intelligent Education and Intelligent Research (IEIR) · doi:10.1109/ieir62538.2024.10959986 · Cited by 3
“Notably, the impact of the Man-AI collaboration paradigm is underscored by the surge in publications within prestigious journals like Nature [3], [8]–[13], [26]–[28], [35], [36], [45], [46], [52], [54] and Science [14]–[16], [50] over the past five years.”
Aniva, L., Sun, C., Miranda, B., Barrett, C., & Koyejo, S. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-90643-5_6 · Cited by 22
The Rise of Deep Learning and Neural Networks: Revolutionizing Artificial Intelligence
-, C. G. (2025) · European Journal of Computer Science and Information Technology · doi:10.37745/ejcsit.2013/vol13n178898 · Cited by 6
Mitigating data bias and ensuring reliable evaluation of AI models with shortcut hull learning
Zhou, W., Liu, F., Zheng, H., & Zhao, R. J. (2025) · Nature Communications · doi:10.1038/s41467-025-60801-6 · Cited by 13
A Robot That Plays Rubik's Race, a Slide Puzzle Game
Jadhav, S. S., Sharma, A. K., Anam, N. H., Yuan, J., & Duan, C. (2024) · International Conference on Automation, Control and Robots (ICACR) · doi:10.1109/icacr62205.2024.11053726 · Cited by 2
When machines take over: professional chess as a model case for the societal impact of superhuman AI
Anicker, F. (2025) · AI & Society · doi:10.1007/s00146-025-02465-w · Cited by 6
“Due to its superior evaluative capacities, it was much better at choosing the variations worth calculating—a capability very similar to human intuition (Silver et al. 2018).”
Quantum many-body simulations from a reinforcement-learned exponential Ansatz
Wang, Y., & Mazziotti, D. A. (2025) · Physical Review A · doi:10.1103/hrgr-klw3 · Cited by 2
Wang, C., Zhang, Y., Zhang, S., Wang, X., & Zhao, Z. (2025) · Computers & Operations Research · doi:10.1016/j.cor.2025.107221 · Cited by 2
Jia, L., Cai, C., & Wang, X. (2025) · Nonlinear Dynamics · doi:10.1007/s11071-025-11636-9 · Cited by 3
“It has been demonstrated that RL can achieve high decision-making performance through self-play [39].”
A Review of Deep Reinforcement Learning Methods for Production Scheduling
Di, Y., Deng, L., Wang, L., et al. (2025) · Tsinghua Science & Technology · doi:10.26599/tst.2024.9010247 · Cited by 2
Streaming system for detection and location of seismic and infrasound events PSDL
Asming, V. E., Федоров, А. В., & Vinogradov, Y. (2025) · Геофизические исследования · doi:10.21455/gr2025.3-1 · Cited by 2
CHEQ-ing the Box: Safe Variable Impedance Learning for Robotic Polishing
Cramer, E., Jäschke, L., & Trimpe, S. (2025) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2025.10.241 · Cited by 2
Dattijo, A., & Jo, S. (2025) · Discover Artificial Intelligence · doi:10.1007/s44163-025-00439-x · Cited by 7
“Starting from random play and given no domain knowledge except game rules, AlphaZero convincingly defeated world-champion programs in each of these games within 24 h of training [5].”
Assessing the inertial response time of grid-forming converters: Estimation and optimization
Sun, L. L., Pan, Y., & Wang, H. (2025) · Electric Power Systems Research · doi:10.1016/j.epsr.2025.112470 · Cited by 5
Adaptation of hippocampal spatial and contextual representations to task structure
Nyilas, R., Kelemen, A. B., Lükő, B., Sümegi, M., Ujfalussy, B. B., & Makara, J. K. (2025) · Science Advances · doi:10.1126/sciadv.adu4899 · Cited by 3
An Adaptive Threshold for the Canny Edge with Weak Label
Choi, K.-H., & Ha, J.-E. (2025) · Applied Sciences · doi:10.3390/app152212158 · Cited by 5
The Role of Artificial Intelligence in Gaming
del Bosque, A., Fernández‐Arias, P., Λαμπρόπουλος, Γ., & Vergara, D. (2025) · Applied Sciences · doi:10.3390/app152312358 · Cited by 7
Lv, M., Zhang, Q., Dong, Z., & Zuo, Z. (2025) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2025.3641526 · Cited by 3
“…progress in artificial intelligence, the question above could be resolved by intelligence control and reinforcement learning (RL) that have demonstrated their potential in finding an optimal decision maker or controller without using model information [27], [28], [29], [30], [31], [32], [33], [34].”
Yuan, S., & Li, X. (2025) · Science China Earth Sciences · doi:10.1007/s11430-025-1738-8 · Cited by 2
“…in foundational architectures (Mnih et al., 2015) and later crystallized in game-theoretic domains through systems like AlphaGo and AlphaZero — which specialize in competitive board games such as Go and chess (Silver et al., 2017, 2018), have significantly addressed many of these challenges.”
Co-evolution of neurotechnology and AI: ethical challenges and regulatory approaches
Shenderyuk-Zhidkov, A., Maksimenko, V., Hramov, A., & Federal State Autonomous Institution "Digital Industrial Technologies Moscow, R. (2025) · Izvestiya VUZ Applied Nonlinear Dynamics · doi:10.18500/0869-6632-003196 · Cited by 4
A General Upper Bound for the Runtime of a Coevolutionary Algorithm on Impartial Combinatorial Games
Benford, A., & Lehre, P. K. (2025) · Genetic and Evolutionary Computation Conference (GECCO) · doi:10.1145/3712256.3726437 · Cited by 2
Estimating the motor exploration in reinforcement learning
Zai, A. T., Lorenz, C., Srikantharajah, S., Giret, N., & Hahnloser, R. H. R. (2025) · iScience · doi:10.1016/j.isci.2025.114398 · Cited by 3
“3 RL is a successful strategy for gaming 4,5 and holds promise as a theoretical framework for understanding neural processing, particularly in dopamine neurons, 6–8 but see.”
Artificial Intelligence in the Development of Sustainable Biopolymers
Oduro, K. O. A., Amponsah, L. A., Arthur, M., & Aboagye, S. A. (2026) · Materials horizons · doi:10.1007/978-981-95-2258-3_11 · Cited by 6
Wu, Z. (2026) · Scientific Reports · doi:10.1038/s41598-026-35877-9 · Cited by 2
Tang, C., Zhou, L., Chen, J., Lai, J., & Chen, G. (2026) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2026.3664774 · Cited by 2
“As a result, reinforcement learning finds many applications, such as backgammon [34], AlphaGo series [35], [36], automatic driving [37], robotic control [38], [39], etc.”
The Economic Impacts of Artificial Intelligence: A Multidisciplinary, Multi-book Review
Bryan, K. A. (2026) · Journal of Economic Literature · doi:10.1257/jel.20251799 · Cited by 9
Livada, Č., Duka, M., Keser, T., et al. (2026) · Electronics · doi:10.3390/electronics15050958 · Cited by 2
Cognitive state monitoring for neuroadaptive information visualization
Idesis, S., Kassinopoulos, M., Barreda-Ángeles, M., Leiva, L. A., & Arapakis, I. (2026) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2026.1793651 · Cited by 4
“…et al. (2017) argued that most AIs fall within a C0 stage of “unconscious processing,” although we can find examples of so-called super-human behavior in computer vision (Esteva et al., 2017), machine translation (Hassan et al., 2018), and game playing (Silver et al., 2018), among other tasks.”
A Brief View of Molecular Modeling Approaches to P2 Receptors
Alberto, A. V. P., Santos, L. H., Ferreira, R., Ferreira, D. N. M., & Alves, L. A. (2019) · IntechOpen eBooks · doi:10.5772/intechopen.86862 · Cited by 2
“AI comprises several definitions but can be defined briefly as a part of computer science capable of performing activities ascribed to human intelligence, such as solving problems, with similar and, in some instances, superior cognitive abilities, for instance, in some cases involving image recognition and playing gam…”
BackgroundLearning Context-aware Task Reasoning for Efficient Meta Reinforcement Learning
Wang, H., Zhou, J., & He, X. (2020) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/rveo2535 · Cited by 16
Spintronics for Neuromorphic Engineering
Lim, G. J., Ang, C. C. I., & Lew, W. S. (2021) · Springer eBooks · doi:10.1007/978-981-15-6912-8_9 · Cited by 3
Q-Value Weighted Regression: Reinforcement Learning with Limited Data
Kozakowski, P., Kaiser, Ł., Michalewski, H., Mohiuddin, A., & Kańska, K. (2022) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn55064.2022.9892633 · Cited by 9
“Deep reinforcement learning has been applied to a large number of challenging tasks spanning from games [1], [2] to robotic control [3]–[5].”
BackgroundA Cognitive FMCW Radar to Minimize a Sequence of Range-Doppler Measurements
Altmann, M., Ott, P., Stache, N. C., Kozlov, D., & Waldschmidt, C. (2021) · European Radar Conference (EuRAD) · doi:10.1109/eurad48048.2021.00065 · Cited by 2
“Other tasks include playing Atari Games [8] or chess, shogi, and Go achieving superhuman performance [9].”
Hedging of financial derivative contracts via Monte Carlo tree search
Szehr, O. (2023) · The Journal of Computational Finance · doi:10.21314/jcf.2023.009 · Cited by 4
Cazenave, T. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-34017-8_13 · Cited by 3
“It was used in the AlphaGo program [8] as well as in its descendants AlphaGo Zero [9] and Alpha Zero [1].”
MethodDelay Aware Universal Notice Network: Real world multi-robot transfer learning
Beaussant, S., Lengagne, S., Thuilot, B., & Stasse, O. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9635917 · Cited by 3
“From reaching human-level control on Atari games [1] to beating world class champions at complex games [2], [3], the RL field is full of promises.”
BackgroundAnatomizing the Elo transfer network of Weiqi players
Li, M., & Zhou, W. (2021) · The European Physical Journal B · doi:10.1140/epjb/s10051-021-00180-1 · Cited by 2
Scalable sub-game solving for imperfect-information games
Li, H., Wang, X., Li, K., Jia, F., Wu, Y., Zhang, J., & Qi, S. (2021) · Knowledge-Based Systems · doi:10.1016/j.knosys.2021.107434 · Cited by 2
Descent wins five gold medals at the Computer Olympiad
Cohen-Solal, Q., & Cazenave, T. (2021) · ICGA Journal · doi:10.3233/icg-210192 · Cited by 2
“Unlike AlphaZero-like algorithms (Silver et al., 2018), the Descent framework uses a variant of Un-bounded Minimax (Korf and Chickering, 1996), instead of Monte Carlo Tree Search, to construct the partial game tree used to determine the best action to play and to collect data for learning.”
MethodOptimised Playout Implementations for the Ludii General Game System
Soemers, D. J. N. J., Piette, É., Stephenson, M., & Browne, C. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-11488-5_20 · Cited by 2
“When offline training is used to train policies, for instance based on deep neural networks [16] or simpler function approximators and state-action features [19], it is also customary to use such a distribution with τ = 1 (leading to a softmax distribution) and the Q(m) values referred to as logits.”
Neural Maximum Independent Set
Pontoizeau, T., Sikora, F., Yger, F., et al. (2021) · Communications in computer and information science · doi:10.1007/978-3-030-93736-2_18 · Cited by 2
“Deep Reinforcement Learning gave state of the art results on complex two-player complete information games such as Go, Chess and Shogi [30].”
BackgroundTCEC14: The 14th Top Chess Engine Championship
Haworth, G., & Hernandez, N. (2019) · ICGA Journal · doi:10.3233/icg-190117 · Cited by 2
Predictive Modeling of Aircraft Dynamics Using Neural Networks
Soleyman, S., Chen, Y., Fadaie, J., Hung, F., Khosla, D., Moffit, S., Roach, S. M., & Tullock, C. (2022) · SAE International Journal of Aerospace · doi:10.4271/01-15-02-0010 · Cited by 2
Heuristics: Human and Nonhuman
Voiskounsky, A. E. (2022) · Вестник Российского университета дружбы народов. Серия: Психология и педагогика · doi:10.22363/2313-1683-2022-19-2-195-208 · Cited by 2
“Поскольку программный комплекс AlphaZero обыгрывает игроков самой высокой квалификации, то такого рода «эвристики» следует признать весьма эффективными (Silver et al., 2018).”
Deep Reinforcement Learning in Smart Grid: Progress and Prospects
Akagić, A., & Džafić, I. (2022) · International Conference on Information, Communication and Automation Technologies (ICAT) · doi:10.1109/icat54566.2022.9811131 · Cited by 4
“…with human-level control in Atari games [31], AlphaGo which defeated the top human experts in the Go game [49], AlphaZero which defeated a world champion in the game of chess, shogi, and Go [50], and MuZero which masters Go, chess, shogi and Atari without needing to be told the rules [51].”
BackgroundModel-Based Reinforcement Learning
Plaat, A. (2022) · Springer eBooks · doi:10.1007/978-981-19-0638-1_5 · Cited by 3
Playing Good-Quality Games with Weak Players by Combining Programs with Different Roles
Hsueh, C.-H., & Ikeda, K. (2022) · IEEE Conference on Games (CoG) · doi:10.1109/cog51982.2022.9893698 · Cited by 2
“A well-known example is AlphaZero [1] for chess, shogi, and Go.”
InfluentialAlphaBluff: An AI-Powered Heads-Up No-Limit Texas Hold’em Poker Video Game
Lazaridis, A., Perchanidis, C., Chovardas, D., & Vlahavas, I. P. (2022) · International Conference on Innovations in Intelligent Systems and Applications (INISTA) · doi:10.1109/inista55318.2022.9894244 · Cited by 2
“4) AlphaZero with Infoset MCTS: AlphaZero [19] is a general RL algorithm that combines RL and Search.”
MethodMonte Carlo Tree Search: A Survey of Theories and Applications
Ou, T., Lu, Y., Wu, X., & Cao, J. (2022) · International Conference on Big Data, Artificial Intelligence and Internet of Things Engineering (ICBAIE) · doi:10.1109/icbaie56435.2022.9985899 · Cited by 8
“The success of AlphaZero breaks the phenomenon that, for at least four decades, Chess programs using traditional MCTS are weaker than those using 𝛼 - 𝛽 search [50].”
MethodLearning to Play Koi-Koi Hanafuda Card Games With Transformers
Guan, S., Wang, J., Zhu, R., Qian, J., & Wei, Z. (2023) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2023.3240674 · Cited by 2
“After the series of successes of AlphaGo, AlphaGo Zero, and AlphaZero [31], [32] in the field of board games including Go , chess , and shogi , deep RL is increasingly adopted in card game AI.”
BackgroundA Local-Pattern Related Look-Up Table
Shih, C.-C., Wei, T.-H., Wu, T.-R., et al. (2023) · IEEE Transactions on Games · doi:10.1109/tg.2023.3263536 · Cited by 2
“The solver program is an AlphaZero MCTS solver [25], [26] that is modified from the RZS program [10], [27], which in turn is based on the Go player CGI [28].”
MethodIntroduction of machine learning for astronomy (hands-on workshop)
Wang, Y., Moradi, R., Haghighi, M. H. Z., & Rastegarnia, F. (2022) · Astronomical and Astrophysical Transactions · doi:10.17184/eac.7535 · Cited by 3
Artificial Intelligence With Graph Neural Networks Applied to a Risk -Like Board Game
Bauer, A. (2023) · IEEE Transactions on Games · doi:10.1109/tg.2023.3270162 · Cited by 2
“Blomqvist [12] combined Monte Carlo tree search (MCTS) with a neural network to play Lux , in a manner similar to AlphaZero [13].”
MethodSchlicht, P. (2023) · Springer eBooks · doi:10.1007/978-3-658-40232-7_29 · Cited by 2
Intelligence at any price? A criterion for defining AI
Nadin, M. (2023) · AI & Society · doi:10.1007/s00146-023-01695-0 · Cited by 2
Planning and Learning with Adaptive Lookahead
Rosenberg, A., Hallak, A., Mannor, S., Chechik, G., & Dalal, G. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i8.26149 · Cited by 3
“Notably, in AlphaGo (Silver et al. 2018) and MuZero (Schrittwieser et al.”
BackgroundPolicy-Independent Behavioral Metric-Based Representation for Deep Reinforcement Learning
Liao, W., Zhang, Z., & Yu, Y. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i7.26052 · Cited by 12
“Deep reinforcement learning (RL) has made a series of major breakthroughs in recent years (Mnih et al. 2015; Silver et al. 2018; Vinyals et al. 2019).”
BackgroundDemidovskij, A. V., Tugaryov, A., Trutnev, A., Kazyulina, M., Salnikov, I., & Pavlov, S. V. (2023) · Mathematics · doi:10.3390/math11143120 · Cited by 3
Online 3D Packing Problem Based on Bi-Value Guidance
Qi, M., & Zhang, L. (2023) · Journal of Computer and Communications · doi:10.4236/jcc.2023.117010 · Cited by 2
Shi-jie, H., Li, S., An, B., Zhao, W., & Liu, P. (2023) · Autonomous Agents and Multi-Agent Systems · doi:10.1007/s10458-023-09620-x · Cited by 4
An Actor-Critic Framework for Online Control With Environment Stability Guarantee
Osinenko, P., Yaremenko, G., Malaniya, G., & Bolychev, A. (2023) · IEEE Access · doi:10.1109/access.2023.3306070 · Cited by 3
RBNets: A Reinforcement Learning Approach for Learning Bayesian Network Structure
Zheng, Z., Wang, C., Gao, X., et al. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43418-1_12 · Cited by 2
Optimisation of Matrix Production System Reconfiguration with Reinforcement Learning
Czarnetzki, L., Laflamme, C., Halbwidl, C., Günther, L. C., Sobottka, T., & Bachlechner, D. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-42608-7_2 · Cited by 2
He, Q., Zhou, T., Fang, M., & Maghsudi, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43421-1_34 · Cited by 5
Deep Learning - Recent Findings and Research
Domínguez-Morales, M., Civit-Masot, J., Munoz-Saavedra, L., & Damaševičius, R. (2023) · Artificial intelligence · doi:10.5772/intechopen.107725 · Cited by 5
Mixture of Public and Private Distributions in Imperfect Information Games
Arjonilla, J., Saffidine, A., & Cazenave, T. (2023) · IEEE Conference on Games (CoG) · doi:10.1109/cog57401.2023.10333169 · Cited by 2
Filevich, J. P., & Cancela, H. (2023) · doi:10.1109/clei60451.2023.10346191 · Cited by 2
“Throughout the last three decades, many games have been solved (at least in an approximate way); for example, chess [1], Go [2], Sh¯ogi and Atari [3], among others.”
BackgroundA Short Introduction to Artificial Intelligence: Methods, Success Stories, and Current Limitations
Heitzinger, C., & Woltran, S. (2023) · Springer eBooks · doi:10.1007/978-3-031-45304-5_9 · Cited by 4
Alexander, R., Ma, L., Dou, Z.-L., Cai, Z., & Huang, Y. (2023) · Journal of Cybersecurity Education Research and Practice · doi:10.32727/8.2023.34 · Cited by 5
Tian, Y., Zhang, K., Tedrake, R., & Sra, S. (2023) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc49753.2023.10383754 · Cited by 6
“Announced by DeepMind in 2019, MuZero extends the line of works including AlphaGo [19] and AlphaZero [20] by not requiring knowledge of the game rules.”
BackgroundCross-Modal Interaction via Reinforcement Feedback for Audio-Lyrics Retrieval
Zhou, D. M., Lei, F., Li, L., Zhou, Y., & Yang, A. (2024) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2024.3358048 · Cited by 3
“Recent years have witnessed breakthroughs of reinforcement learning in many fields such as games, robot control and recommender systems [31], [32].”
BackgroundHenshall, A., & Karaman, S. (2024) · AIAA SciTech Forum · doi:10.2514/6.2024-2762 · Cited by 3
AI Tool for Exploring How Economic Activities Impact Local Ecosystems
Strannegård, C., Engsner, N., Lindgren, R., Olsson, S., & Endler, J. A. (2024) · Lecture notes in networks and systems · doi:10.1007/978-3-031-47718-8_44 · Cited by 3
An Introduction to Reinforcement Learning and Its Application in Various Domains
Amin, S. (2024) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3693-1738-9.ch001 · Cited by 4
A Reinforcement-Learning Approach to Control Robotic Manipulator Based on Improved DDPG
Majumder, S., & Sahoo, S. R. (2023) · Indian Control Conference · doi:10.1109/icc61519.2023.10442503 · Cited by 2
“Robot manipulation can become as human-like as feasible by directly learning dexterous manipulation using RL algorithms [12].”
MethodInterpretability of rectangle packing solutions with Monte Carlo tree search
López, Y. G., García, C. G., García‐Díaz, V., Núñez‐Valdéz, E. R., & Gómez, A. G. (2024) · Journal of Heuristics · doi:10.1007/s10732-024-09525-2 · Cited by 3
“It is not however until Silver et al (2018) manage to employ it to beat the Go world champion that MCTS starts receiving great attention in literature.”
Interpretable Policy Extraction with Neuro-Symbolic Reinforcement Learning
Dutta, R. K., Wang, Q., Singh, A., Kumarjiguda, D., Li, X., & Senthilnath, J. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10446037 · Cited by 7
“…gained popularity due to its effectiveness in finding optimal solutions to various sequential decision-making tasks ranging from two-player games [Silver et al.(2018)], indoor localization [Salimibeni and Mohammadi(2023)] to text semantic analysis [Gao et al.(2022)] and drug discovery [Popova et…”
BackgroundP2BPO: Permeable Penalty Barrier-Based Policy Optimization for Safe RL
Dey, S., Dasgupta, P., & Dey, S. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i19.30094 · Cited by 2
“Reinforcement Learning (RL) (Sutton and Barto 1998), a nature-inspired reward-based learning method, has shown potential for complex strategy-based tasks like playing Chess, Go (Silver et al. 2018), Atari games (Mnih et al. 2013) etc.”
BackgroundNativism and empiricism in artificial intelligence
Long, R. (2024) · Philosophical Studies · doi:10.1007/s11098-024-02122-w · Cited by 7
An investigation of belief-free DRL and MCTS for inspection and maintenance planning
Koutas, D., Bismut, E., & Straub, D. (2024) · Journal of Infrastructure Preservation and Resilience · doi:10.1186/s43065-024-00098-9 · Cited by 2
“Although MCTS was originally formulated for fully observable domains with great success [34, 35], it has also been applied to POMDPs [36, 37].”
Michishita, Y. (2024) · Journal of the Physical Society of Japan · doi:10.7566/jpsj.93.074005 · Cited by 6
Neural Network-Based Information Set Weighting for Playing Reconnaissance Blind Chess
Bertram, T., Fürnkranz, J., & Müller, M. (2024) · IEEE Transactions on Games · doi:10.1109/tg.2024.3425803 · Cited by 3
“Blüml et al. [5] showed that proficient gameplay can be achieved in an imperfect information games with an approach akin to AlphaZero [33], in which they fuse policies across different states.”
Deep Learning: Implications for Human Learning and Memory
McClelland, J. L., & Botvinick, M. (2024) · Oxford University Press eBooks · doi:10.1093/oxfordhb/9780190917982.013.8 · Cited by 3
Wang, C., Deng, J., Zhao, H., et al. (2024) · Applied Mathematics and Computation · doi:10.1016/j.amc.2024.128956 · Cited by 2
NovPhy: A physical reasoning benchmark for open-world AI systems
Pinto, V., Gamage, C., Xue, C. Q., Zhang, P., Nikonova, E., Stephenson, M., & Renz, J. (2024) · Artificial Intelligence · doi:10.1016/j.artint.2024.104198 · Cited by 8
Solving Complex Optimisation Problems by Machine Learning
Prestwich, S. (2024) · AppliedMath · doi:10.3390/appliedmath4030049 · Cited by 2
From the neural networks miracle – to AGI
Zhuravlev, D. V., & Smolin, V. S. (2024) · Proektirovanie buduŝego. Problemy cifrovoj realʹnosti · doi:10.20948/future-2024-2-2 · Cited by 2
A Simple, Solid, and Reproducible Baseline for Bridge Bidding AI
Kita, H., Koyamada, S., Yamaguchi, Y., & Shin, I. (2024) · IEEE Conference on Games (CoG) · doi:10.1109/cog60054.2024.10645547 · Cited by 2
“Notable examples include backgammon [1], chess [2], Go [2]– [4], poker [5]–[7], mahjong [8], and Atari 2600 [9].”
BackgroundMulti-DMC: Deep Monte-Carlo with Multi-Stage Learning in the Card Game UNO
Yang, X., Liu, X., & Lin, W. (2024) · IEEE Conference on Games (CoG) · doi:10.1109/cog60054.2024.10645662 · Cited by 2
“From AlphaGo Series (AlphaGo [1], AlphaGo Zero [2], Al-phaZero [3]) acquired famous achievements in 2016, various games have been developed, such as the board game Hex [4], the fighting game Blade & Soul [5], the stochastic puzzle game 2048 [6], and card games Mahjong [7], DouDizhu [8] [9], Texas…”
BackgroundRoot values and root skills: a new model for intelligence education
Gaspard, J. J. S., & Pili, G. (2024) · Intelligence & National Security · doi:10.1080/02684527.2024.2390762 · Cited by 2
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.