I guess I struggle to understand why under estimation is worse than over estimation, when the final result is a ranking. It seems like they're equally likely to

I guess I struggle to understand why under estimation is worse than over estimation, when the final result is a ranking. It seems like they're equally likely to produce an incorrect ranking!