A thousand games.
One strength test.

Ratings start at 1500. Each match compares all six pairs using final engine scores (Elo, K=32 divided across three opponents). Tied scores split the result. Ratings remain provisional until a robot has played 30 games. Four-player results measure performance in these conditions; they do not guarantee the same order in a duel.

* Tied first-place finishes share one win. The strongest final score wins at elimination or the 300-turn cap. All unique selectable robot names use the same variants as the public server; duplicate roster entries are counted once. Player ratings are kept separate.