Observed arrival · 2026-09-21
Armies: Stratego as a Research Arena
A browser-playable Stratego engine where visitors can face research players, inspect replays, and compare agents on a competitive ladder.
- For
- Stratego players and reinforcement-learning researchers
- Worth noticing
- The ladder compares expectimax, greedy, rollout, random, and other agents with recorded games and Elo estimates.
Field notes
Armies exposes the mechanics of its experiment rather than presenting AI as a sealed opponent. The page describes a ladder spanning greedy rules, rollouts, expectimax over sampled worlds, and self-play systems, with Elo estimates fitted from pairwise results. It also specifies a compact one-byte-per-cell representation and reports millions of positions processed per second. The visible research lineage points to a 2025 arXiv paper on Stratego, self-play reinforcement learning, and test-time search.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue