{"schemaVersion":"1.0","generatedFrom":"https://brightaifuture.com/discoveries/ataraxos-stratego-hidden-information","record":{"id":"ataraxos-stratego-hidden-information","headline":"Ataraxos AI beat a Stratego champion by planning around hidden information","canonicalUrl":"https://brightaifuture.com/discoveries/ataraxos-stratego-hidden-information","datePublished":"2026-10-10","dateModified":null,"sourcePublicationDate":"2026-09-30","author":null,"publisher":{"name":"Bright AI Future","url":"https://brightaifuture.com/"},"topics":[],"summary":"Ataraxos won 15 of 20 games against a Stratego champion. Its hidden-information planning and estimated training compute below $8,000, explained.","evidenceState":"Emerging","keyFacts":[{"label":"AI’s role","value":"Ataraxos combines self-play reinforcement learning with decision-time planning using plausible models of hidden piece identities."},{"label":"Documented result","value":"The headline test was a 20-game series against Pim Niemeijer, a four-time world champion. Ataraxos won 15 games, lost one and drew four."},{"label":"Important limitation","value":"That is a compute estimate, rather than an accounting of salaries, development and every experiment behind the project."}],"limitations":["That is a compute estimate, rather than an accounting of salaries, development and every experiment behind the project.","A broader picture would include more repeated series against different leading players.","the two systems did not play a direct match.","Public code helps make a result inspectable; successful independent reproduction still has to be established.","Bright has not run the software.","Real-world usefulness remains to be demonstrated."],"evidenceLinks":[{"title":"Scalable decision-making for games of imperfect information · Nature, September 30, 2026","url":"https://www.nature.com/articles/s41586-026-11036-y","type":"paper"},{"title":"Scalable Decision Making for Games of Imperfect Information · author manuscript v2, October 4, 2026","url":"https://arxiv.org/html/2511.07312v2","type":"paper"},{"title":"Author manuscript version history · first posted November 10, 2025","url":"https://arxiv.org/abs/2511.07312","type":"paper"},{"title":"CMU researchers develop AI that tackles hidden information in Stratego · October 8, 2026","url":"https://www.cs.cmu.edu/news/2026/ai-tackles-stratego","type":"institution"},{"title":"Game-playing AI brings a new champ to Stratego · MIT, September 30, 2026","url":"https://news.mit.edu/2026/game-playing-ai-stratego-new-champ-0930","type":"institution"},{"title":"Ataraxos vs. Pim Niemeijer · public 20-game Stratego archive","url":"https://ataraxosai.github.io/","type":"dataset"},{"title":"Ataraxos Stratego repository · training instructions, MIT license and pretrained files","url":"https://github.com/AtaraxosAI/stratego","type":"repository"},{"title":"Mastering Stratego, the classic game of imperfect information · DeepMind, December 1, 2022","url":"https://deepmind.google/blog/mastering-stratego-the-classic-game-of-imperfect-information/","type":"institution"}],"evidencePackUrl":"https://brightaifuture.com/evidence-pack/ataraxos-stratego-hidden-information","embedUrl":"https://brightaifuture.com/embed/story/ataraxos-stratego-hidden-information","attribution":{"credit":"Bright AI Future","requirements":["Link to the canonical Bright record.","Keep material limitations with the claim they qualify.","Link to the original evidence when repeating a substantive claim.","Do not describe a source check or organization-reported result as independent verification."],"sourceRights":"Linked source material, quotations, trademarks and media remain subject to their owners’ terms. No reuse right is granted for third-party media."}},"claim":{"humanProblem":"Making a decision can require weighing information another player has concealed, while accounting for what your own actions reveal.","priorConstraint":"Stratego combines many possible hidden piece identities with strategic behavior, making straightforward exhaustive planning difficult.","aiRole":"Ataraxos combines self-play reinforcement learning with decision-time planning using plausible models of hidden piece identities.","documentedResult":"The headline test was a 20-game series against Pim Niemeijer, a four-time world champion. Ataraxos won 15 games, lost one and drew four.","whyItMayMatter":"Planning around plausible hidden information offers a controlled research result other groups can investigate. Lower training-compute requirements could broaden academic participation.","unresolvedQuestions":["How would repeated matches against other leading players change the assessment?","Can independent research groups reproduce the results with the public code and pretrained files?","What additional models, failure detection and auditable explanations would be needed for useful real-world decision tools?"]},"evidenceAssessment":{"state":"Emerging","claimConfidence":"unassessed","reviewState":"approved","reviewMethod":"ai-assisted","reviewNote":"Owner explicitly approved Bright publication and branded distribution after the source-checked explainer was reviewed. Final Nature publication metadata and indexed text were checked; direct final publisher PDF was not read. Detailed numerical claims use the accessible author manuscript v2 and primary match archive. September 30 Nature publication, October 8 CMU account and November 2025 original preprint are kept distinct. Under-$8,000 is an estimate for a specific training run at 2025 rental prices, excluding broader research and development. No direct DeepNash match, independent reproduction by Bright or demonstrated real-world deployment is claimed. Original AI-generated conceptual hero reviewed for rank symbols, concealed identities and non-documentary labeling.","lastSourceReview":"2026-10-10","independentVerification":"not-established-by-this-source-review"},"sources":[{"id":"source:ataraxos-nature","title":"Scalable decision-making for games of imperfect information · Nature, September 30, 2026","type":"paper","url":"https://www.nature.com/articles/s41586-026-11036-y"},{"id":"source:ataraxos-author-v2","title":"Scalable Decision Making for Games of Imperfect Information · author manuscript v2, October 4, 2026","type":"paper","url":"https://arxiv.org/html/2511.07312v2"},{"id":"source:ataraxos-history","title":"Author manuscript version history · first posted November 10, 2025","type":"paper","url":"https://arxiv.org/abs/2511.07312"},{"id":"source:ataraxos-cmu","title":"CMU researchers develop AI that tackles hidden information in Stratego · October 8, 2026","type":"institution","url":"https://www.cs.cmu.edu/news/2026/ai-tackles-stratego"},{"id":"source:ataraxos-mit","title":"Game-playing AI brings a new champ to Stratego · MIT, September 30, 2026","type":"institution","url":"https://news.mit.edu/2026/game-playing-ai-stratego-new-champ-0930"},{"id":"source:ataraxos-games","title":"Ataraxos vs. Pim Niemeijer · public 20-game Stratego archive","type":"dataset","url":"https://ataraxosai.github.io/"},{"id":"source:ataraxos-stratego-code","title":"Ataraxos Stratego repository · training instructions, MIT license and pretrained files","type":"repository","url":"https://github.com/AtaraxosAI/stratego"},{"id":"source:ataraxos-deepnash-prior","title":"Mastering Stratego, the classic game of imperfect information · DeepMind, December 1, 2022","type":"institution","url":"https://deepmind.google/blog/mastering-stratego-the-classic-game-of-imperfect-information/"}],"revisions":[{"id":"revision:ataraxos-first-publication-20261010","recordedAt":"2026-10-10T22:43:40.128Z","sourceIds":["source:ataraxos-nature","source:ataraxos-author-v2","source:ataraxos-history","source:ataraxos-cmu","source:ataraxos-mit","source:ataraxos-games","source:ataraxos-stratego-code","source:ataraxos-deepnash-prior"],"summary":"Published a substantial source-checked research explainer with controlled-game evidence, cost scope and comparison limits, rather than presenting the older research as breaking news."}],"corrections":[]}