SOS is arbitrary, so it can be defined in a whole bunch of ways. But just so the wrong impression isn't left, I would like to point out that while it's true that the old SOS was an average SOS, it did not use rankings. The only input into the old SOS formula were wins and losses of your opponents and your opponent's opponents. Some of the models may use rankings, but not all of them. I think at least one uses the old SOS. But your overall point is correct, avg SOS is not a good measure of overall risk. This is what did in AU last year. Among the 3 undefeateds, AU played and beat better good teams, but the others played and beat better bad teams. Unfortunately, AU's bad teams were really bad and accounted for most of the difference in SOS. When avg SOS is so important, then the key is not to schedule really bad teams. There are plenty of mid range teams that aren't much of threat to a good team. The Big 12 have been masters at this kind of scheduling over the last few years.
