Football scheduling woes come from a bad ranking system?

So I've been reading threads like this interesting one over at BOTC, and one of the big complains is that schools have no incentive not to schedule cream puffs in order to get a winning schedule, and get into bowl games like KU did in 2007. This also chains into other worries, like the lame tie breakers between first place teams, and winners of the big 12 north getting screwed into worse bowl games because they face the predictable and meaningless indignity of getting beaten by the South champion (oh noes, they have an extra loss on their season! That means they MUST be worse than those other north teams they beat out in the regular season!!1!!one!!eleven!).

Heaven knows this stuff has all been debated to death, but it seems to me like possibly much of it has to do with a main underlying problem, which is this: When rankings are computed, strength of opponent is waaaay under-weighted, while straight win/loss total is highly over-weighted.

You can think of college rankings as a signal detection problem. We can't directly SEE how good a team is, we can only look at the small number of games they play and try to detect if there are signals in this performance that might give us clues. Unfortunately there are at least two sorts of noise in there. One is that there is game day randomness, so that a team which really is 20% better than another might still only win 80% of the time, depending on injuries, the weather, a key fumble, a single missed tackle, fifth downs, etc. The other is that you can't simply average across all of those because teams don't all play the same quality opposition. You have lower odds of fumbling the softer you get hit (though it's never zero), and you run further against worse tackling, etc.

People vary in how sophisticated they are in sorting through all this mess to detect the true underlying state of affairs, from your Bill C's on down to your basic ESPN unter-geniuses. The modal strategy , though, seems to be what psychologists call "anchor and adjustment". People anchor on the win/loss figure, and then adjust by adding or subtracting a bit based on strength of schedule, and even more if they like a few star players (I think the ESPN guys fixate waaay too much on the latter, which is a big part of their problem - makes good TV to type individual personalities rather than teams and systems, but it doesn't pay to start believing your own hype). The problem is, humans typically don't adjust enough in these situations. This is why we always underestimate how long things will take us ("uh, maybe 20 mins, 30 if things go wrong" ends up meaning "40-50 mins"). That means Kansas can schedule a season of complete cream puffs, and what sticks out to people is "lots of wins" more than the "yeah, against who though." And because this underweighting is predictable, schools try to take advantage of it by rushing to schedule a lot of lame teams, and not enough separation is created between teams with similar win/loss margins, so you end up with stupid tie-breakers deciding who goes to title games in pretty arbitrary ways.

So how do you get past this tendency? Computers can solve a bunch of it (see Bill's "plus" numbers that adjust stats for the strength of  competition they were acheived against), but so long as national rankings are going to continue to depend on a whole lot of people voting based on their cognitive biases, perhaps it would be helpful to start trying to give them some better rules of thumb. Like, perhaps we could fairly uncontroversially sort teams into rough strata of quality (i.e., not trying to force a 1 to 25 ranking, but noting there's maybe a clear 1-4, then another mush of 5-12 teams, and third much of eight more, etc), that covers everybody from the nation's number one to the cellar dwellers of the bottom conferences, you could then suggest something like a "2 level rule" to people. Beating a team more than 2 full levels below you is considered completely non-diagnostic and doesn't count towards your ranking, while losing to a team 2 levels above you is also considered meaningless and non-diagnostic. Neither  tells you anything about how well a team ranks against its neighbors, as ALL those neighbours should also have won / lost against that same team.

If this happened there would be no advantage to scheduling the U of North by NorthEast Staypuffedsville, as a win over them would be considered meaningless, while getting stomped by the visiting NFL all-star champions  would also mean little to nothing. This would give teams an incentive to try to schedule teams that were at most a step or two below them (so they actually get some credit for a win), or are higher (because beating a higher ranked team is always good). Schools could still schedule a lower team if they wanted - perhaps to tighten up their systems against - but it would be like trying an easy maneuver in gymnastics - it would get you almost no points towards getting ranked. That would actually create a lot more interesting strategy in terms of which teams you try to schedule. As it is, we have this ridiculous system where voters are still far too impressed by how good teams look when they're picking on 7 year olds, and that just encourages the behavior.

OR, you could argue that this would just hurt the weakest schools, because none of the rich schools would want to pay to stomp them, so they'd get less of the cash. I don't know, but it would seem like it would help solve some of this "race to the bottom" in terms of puffing out on your non-conference schedule.

Any thoughts?