basesWorkers16x16A (final standings and head-to-head matrix), (ii) one recorded game per starting position (P0 and P1) for each of the 18 opponents in the tournament, and (iii) detailed game-theoretic metrics (Nash, Alpha-Rank, Copeland, robustness) behind the ranking.
Contents
Tournament summary
Overall ranking of our agent against every opponent on the final 16×16 single-map tournament. The two figures below give the context that is referenced throughout the game recordings. More detailed game-theoretic analyses are provided at the bottom of the page.
Example games
Recordings below show our single-map agent (UECD-Best) playing one game per starting position against each opponent of the tournament, on the basesWorkers16x16A map. All games are played with deterministic actions (argmax over the policy), matching the tournament protocol. Left column: our agent starts in position P0 (top-left corner). Right column: our agent starts in position P1 (bottom-right corner). Opponents are listed in the tournament order of strength.
vs RAISocketAI source ↗
vs UtsImass source ↗
vs TMA source ↗
vs UECD-TopFeats source ↗
vs ObiBotKenobi source ↗
vs UECD-AllFeats source ↗
vs UECD-Rushed source ↗
vs CoacAI source ↗
vs WorkerRush source ↗
vs Mayari source ↗
vs MixedBot source ↗
vs Droplet source ↗
vs GridNet source ↗
vs Tiamat source ↗
vs StrategyTactics source ↗
vs NaiveMCTS source ↗
vs LightRush source ↗
vs RandomBiasedAI source ↗
Detailed tournament analysis
Game-theoretic metrics behind the final standings. These complement the head-to-head matrix shown at the top of the page.
