Replay-engine records only: two equal human teams of at least two known public-keyed players, at least 3 minutes, no crash or cheat flag. One-on-one, incomplete, and excluded matches do not change this rating.

Everyone starts at 1500. Team-average pre-match ratings set the expected win chance. Result points = K × (actual − expected); K starts at 12 and tapers to 8 after 20 rated games.

All six (or more) players are ranked on replay score 25%, unit K/D 30%, kills 15%, structures destroyed 10%, completed research topics 15%, and units built 5%. Ties share rank. The weighted score is compared with the player’s expected rank from pre-match ratings, multiplied by 36, and capped at ±10 personal points. Hover a match’s personal points to inspect its metric ranks. Research is the published topic count, not the analyzer’s lab-activity percentage.

Total change is capped at ±15. A strong losing player may gain points; a poorly performing winner may lose points. If any participant lacks trusted counters, the entire match uses team-result points only. Match rank measures these six metrics, not every kind of contribution; no cheating verdict or public ELO change follows from it.