The first new addition to SMQ's sidebar, under the nerdy 'Research' category, is the information-dense site of former Auburn prof Auburn grad and current rural Mississippi resident Paul Kislanko, who crunches poll numbers in a variety of sophisticated ways you do not understand but can appreciate because the general results tell you what you know intuitively: the polls is wack.
Real World Example: Using four different statistical methods of counting (the Condocet order, Bucklin counting, Borda Count and the "traditional" point assignment system), Paul generated four different polls from the exact same votes. LSU, for example, is ranked 9th using Bucklin but 14th using Borda. Layman SMQ doesn't understand the methodology, exactly, but he trusts computers and statistics more than most football-focused folks seem to and thinks the results ought to be useful enough in generating top 25 polls and the like.
Also potentially useful is strength of schedule data, produced here, here and here, and a "Strength of Victory" rating (very, very similar to mulled, attempted and discarded versions obsessed over by SMQ), the results of which are here (much-needed explanation here and, for the baseball-focused theory from which they're derived, here).
In not so sophisticated fashion, Paul also corrected the record reported earlier today in games featuring a mid-major conference team against a BCS conference team (SMQ shorted the big guys five wins and the little guys one). Here is the proof.
Also linked is pretty much the greatest thing ever, Dave Wilson's list of every known college football rating system. There are easily over 100 there, almost all using insane algortihms that are definitely wrong. Or you can build your own insane rating system.