StarCraft AI / Research logLoading evidence…
StarCraft AI progress.
Native OpenBW matches, bot provenance and experiment decisions. Local playing strength is still uncalibrated.
Elo over time
Local league ratings
UncalibratedA separate Gaussian-prior Bradley–Terry league for our exact builds. Each connected component has an arbitrary 1000 reference; it is not a BASIL-scale rating.
Upstream BASIL observations
These dated ratings belong to the upstream bots. Our native ports use separately frozen builds and have no calibrated BASIL rating.
Matches
Bot progress
Follow one bot family across matches, replays, experiments and local rating evidence.
Experiment ledger
Hypotheses, evidence and conclusionsReplay archive
Play in the dgant OpenBW viewer, or download the replay. Failed matches remain in the log even when the engine could not emit a replay.