@xPetu How are you aggregating from individual WPA's to the team level? Do you have a linear model or are you using something like log-odds? Also wouldn't the various teams differing winrates warp the WPA numbers? (i.e. outsized positive effect for a team with a very low winrate)