Ataraxos artifacts and availability boundaries were verified from primary sources.
Ataraxos beats Stratego champions with damped self-play and search
2026-09-30Frontier watchWATCH
Ataraxos combines dynamically damped self-play reinforcement learning with decision-time search for hidden-information games. It beat Pim Niemeijer 15–1 with four draws and went 38–2 at the 2025 Stratego World Championship. Training and belief-model code are public, but pretrained checkpoints are not.
What this means for you
For researchers building simulator-backed planning agents, Ataraxos is evidence that training dynamics and search can reduce the data needed for large hidden-information games. Without checkpoints or an independent reproduction, the result is not yet a practical trial target.
FRONTIER WATCH
Public artifacts and validated results are different.
The recorded fact is the public work and its artifacts. Author-reported outcomes do not establish independent reproduction. Next checks include usable code, task conditions and third-party validation.
Evidence & related material
Technical details & scores
| Method | dynamically damped self-play RL plus decision-time search |
|---|---|
| Human Match | 15 wins, 1 loss, 4 draws |
| World Championship | 38-2 at 2025 Stratego World Championship |
| Efficiency Claim | <1/100 DeepNash training examples; <1/30 self-play games |
| Artifacts | MIT-licensed training and belief-model code; no pretrained checkpoint |
Technical impact 4
Engineering value 2
Adoption signal 1
Maturity 1
Verification cost 5
Risk 2