Play-style similarity

How alike do two bots play? For every pair this compares who they beat and lose to, after subtracting what ELO already predicts — so it measures style, not strength. Two bots of equal rating score near zero unless they actually over- and under-perform against the same opponents. How it works.

Read this as similarity, not as an accusation. A high score has plenty of innocent explanations: a shared open-source template, the same standard build order, or simply two bots by the same author. There is only one way to 12-pool, and two bots that both do it will look alike here.

Matrix

Bots are ordered by ELO, strongest first. Only same-race pairs are compared (Random bots appear in every matrix, since they play all three). A cell compares how the two bots fared against the opponents they both faced — whether they ever played each other makes no difference, and most scored pairs never have. Blank cells are pairs with fewer than 12 opponents in common. Hover any cell for the detail.

Style correlation, different authors 0 → +1 Same author 0 → +1

Most similar pairs

Ranked by how far the correlation sits above 0.5 relative to its spread, so a pair rises only when the correlation is large and the evidence behind it is solid. A striking number resting on a dozen shared opponents has a wide spread and sinks. Sort any column to reorder, and the # column to come back.