Well, either force equal players or find a better way to balance players slot for slot than making sure team's average elo is the same. But how many times was this point raised?

I feel like there almost needs to be a secondary metric for "activity" for low elo players for balancing in sharp situations. There's a huge difference between a 1400 who builds stuff and just has terrible micro, fight management, etc., and a 1400 who builds a bertha at the start of the game.

I don't think it would matter for anyone above like 2k elo, but having a way to measure e.g. tendency towards frontlining would help a lot I think.

I notice a concentration of high variance players on Team 1 here, with Team 1 having a (current, unfortunately post hoc) average +- of 253 compared to Team 2 having an average +- of 190.

There's no mathematical basis for balancing teams for even whr uncertainty; the errors average out regardless of which teams the high variance players are on. But from a perspective of sharing the frustration fairly, it is perhaps worth balancing for even variance between teams albeit weighted an order of magnitude less than even rating.

As far as balancing for frontlining tendencies goes, this game transcends those. The decisive period of this game (~2:30 - 5:30) had six players on team east (KS, DF, WF, TB, SB, Br) focusing on three players from team west (Rh, BP, ktdm) while team west had two players who were on the frontline but disengaged and porcing up well short of the midline without any visible enemies in front of them. (GO and JB) Meanwhile, on the backlines, Steel_Blue was using planes to support the concentrated attack on Rallehop while VexCollective's planes rendered only negligible support on that front - backliners are not interchangeable in terms of front line impact.