A fast Stratego engine, players that learn by playing themselves, and the research behind deciding well when most of the board is hidden.
- Board
- One byte per cell
- Engine
- Millions of moves a second
- Rules
- Tournament anti-chase rules
The ladder
All results- ExpectimaxTop1535Elo343-7-110 in 460 games
- Greedy1393Elo303-8-229 in 540 games
- Rollout1283Elo19-0-81 in 100 games
- Ataraxos smoke1090Elo18-26-36 in 80 games
- Random1000Elo127-219-354 in 700 games
- N-tuple leaf0Elo0-0-0 in 0 games
Games, wins, draws and losses over every registered game of September 18, 2026 (the 16-world depth-2 cell of the sweep counts for the expectimax rung) plus a 60-game random-expectimax pairing; Elo is a logistic fit over the pairwise effective scores with random anchored at 1000; rungs with zero games are unmeasured; built by scripts/build_second_pass_assets.py.
Hidden information
Each side sees only its own ranks. Play means acting on beliefs about forty face-down pieces and updating them as they move and fight.
Search at play time
The ladder climbs from greedy rules to expectimax over sampled worlds, then to networks trained by self-play and refined by search.
An engine you can measure
The rules run allocation-free on one byte per cell, so millions of positions a second feed the arena and the training loop.