Nature paper reports an AI called Ataraxos beating the most decorated human Stratego player
1 October 2026
A paper published in Nature on 30 September 2026 introduces Ataraxos, an artificial intelligence for the board wargame Stratego. The authors describe Stratego as prized as a challenge problem for the scale of its hidden information, and say that top-human-level play had remained beyond the reach of AI despite industrial research efforts costing millions of dollars. In a 20-game series, Ataraxos defeated Pim Niemeijer, described in the paper as the most decorated Stratego player of all time, with 15 wins, 1 loss and 4 draws. The authors say this is, to their knowledge, the first superhuman result in the game's history, and that Ataraxos cost only a few thousand dollars to train. Using the same techniques for self-play reinforcement learning and test-time search, the team also built a superhuman AI for Barrage Stratego and state-of-the-art AIs for Hanabi and dou dizhu. The paper appears in Nature volume 658, pages 55-59, and its authors are at Carnegie Mellon University, New York University, Stanford University and the Massachusetts Institute of Technology.
In plain words
Imperfect information Imperfect information — A term for interactions in which some players hold information that others do not, unlike chess or Go where the whole position is visible.
Self-play reinforcement learning Self-play reinforcement learning — Training in which a program improves by repeatedly playing against versions of itself rather than against recorded human games.