[Gauss]

I'm surprised how incorrect the Sagarin rankings are.  Almost to the point of being offensive.  ??? 

It depends on what you are trying to measure.
Sagarin does a decent job (not the best out of all computer rankings, but good enough) at ranking the merit of each team's record. The teams Ohio State beat and lost to constitute about the 10th best record of this year.

You, like most humans, probably see the rankings a bit like "if this team played this other, who would win in a neutral field?".
My main problem with this (apart from being even more difficult to measure than merit of a record) is that is not what most people end up using.

If people are trying to answer that question, then they are assigning too much value to fluke events. For instance, if two teams are evenly matched (as seen by the flow of game) but one ball bounce or ref call determines that team A wins (see ASU-Wisconsin), that should still count as a tie for the "who's the best team" question.

In reality, I think humans want a hybrid system that kind of answers the "who's the best team" question, but also assigns much more value to a win by a single point than a loss by a single point.

TLDR: people will always complain about rankings because they don't define them precisely.

[Gauss]

I'm surprised how incorrect the Sagarin rankings are.  Almost to the point of being offensive.  ??? 

It depends on what you are trying to measure.
Sagarin does a decent job (not the best out of all computer rankings, but good enough) at ranking the merit of each team's record. The teams Ohio State beat and lost to constitute about the 10th best record of this year.

You, like most humans, probably see the rankings a bit like "if this team played this other, who would win in a neutral field?".
My main problem with this (apart from being even more difficult to measure than merit of a record) is that is not what most people end up using.

If people are trying to answer that question, then they are assigning too much value to fluke events. For instance, if two teams are evenly matched (as seen by the flow of game) but one ball bounce or ref call determines that team A wins (see ASU-Wisconsin), that should still count as a tie for the "who's the best team" question.

In reality, I think humans want a hybrid system that kind of answers the "who's the best team" question, but also assigns much more value to a win by a single point than a loss by a single point.

TLDR: people will always complain about rankings because they don't define them precisely.