Why does `reduce_impact` not account for both teams' Phi values?

[ReagentX]

Why not add this on line 79?

phi = math.sqrt(team_1.phi ** 2 + team_2.phi ** 2)

[drdarina]

Why do you think that's reasonable? He's following the original algorithm as published by Mark Glickman, so g(phi) is a function only of the opponent's rating deviation.