Apologies if my first post on this blog amounts to what ought to be a fanshot instead of a fanpost, but I am really interested in provoking discussion about the computer rankings, what they mean for the ducks, and how it is exactly that they even work in the first place. I know that the full BCS rankings will not appear for another week, but this morning is the beginning of when some rankings, like the Sagarin, begin to claim that they are no longer "unbiased." I am sure that many of you understand the mathematical mechanics of something like Sagarin much better than I do, and I have always been curious about some of the claims that these ranking experts peddle, and would love for there to be a way to discuss and understand all of this better.
Allow me to provide a link to this morning's new Sagarin rankings here, we can start with them since they are easy to find and relatively accessible. Now, a few important things to understand about the list here are that it is the "ELO_Chess" rankings that are used for the BCS, that he now claims that the rankings are unbiased by the human polls, and that the order that the teams are listed is based on the combination of the "ELO_Chess" and "Predictor" rankings. The "Predictor" ranking is partly based on margin of victory, while the "ELO_Chess" ranking is solely based on who you have beaten (and who they have beaten), as the BCS does not allow computer rankings to include margin of victory. The elo formula is based on formulas that were (and are?) used to rank chess players. Sagarin claims that his "Predictor" ranking provides a better indication of how good a team really is, but that the elo ranking is the "politically correct" way to rank teams according to the BCS. I suppose that the BCS does not want to encourage teams like Bama to run up the score when they play their traditional three cupcakes a year in order to gain themselves an advantage with the computers.
Okay, so, looking at the Sagarin rankings, I have some really simple questions that would seem to throw the entire system into question. First of all, Sagarin provides the strength of schedule played right in that list, and we can readily see the strength (or lack thereof) of Alabama's schedule, and yet, they are ranked #1 in the elo_chess ranking. How is this possible? I am not just a Duck homer wondering why the Ducks aren't ranked #1, mind you. Still, just by looking at Sagarin's list, it seems totally obvious to me that any unbiased mathematical ranking would put Florida at the top and not Alabama. Florida has now beaten LSU and TAMU. Alabama? They have beaten Arkansas and Michigan. It seems to me that the only possible reason that Alabama ends up #1 in the elo_chess is because they were ranked at #1 by humans to start. How is it that Sagarin can claim that this is unbiased? If elo_chess is solely based on who beat who, why don't Florida's wins count for more than Bama's? How does he even begin this whole process? In order to determine that one team's wins are more valuable than another's, doesn't he have to assign some kind of number to at least some of the teams before running the formula? These are the things that I do not understand and would love to engage more discussion and gain more understanding.
Bringing me to the Ducks in this process...I know that the Ducks' OOC schedule was really weak this year and that that itself accounts for the low schedule rank (86). But honestly, how does an unbiased mathematical formula spit out that Alabama's schedule (50) is so much stronger than the Ducks'? The Ducks' best wins are now against Arizona and Washington, whose best wins were against Oklahoma State and Stanford, respectively. Bama's best wins were against Arkansas and Michigan, whose best wins were against some really sorry seeming Auburn and Purdue teams. So, can anyone explain to me why 50 vs. 86? If one gazes at other teams' SOSs on that list, other anomalies seem to appear as well.
So, to sum up my questions/concerns:
1. Shouldn't ELO_Chess favor the teams with the most wins over the strongest schedules (or: why aren't Florida, South Carolina, Ohio State, Notre Dame, West Virginia, and Oregon State ALL ranked above Bama in ELO_Chess)?
2. Now, about those SOS rankings...shouldn't the Ducks' wins at this point count for more than Alabama's, based on who those teams have beaten?
I know that most comments will simply say, "yeah, computer rankings are stupid" and "you're just now figuring out that computer rankings are dumb?" Nevertheless, I would be really interested in having a substantive discussion about how this stuff works. I think it could help us all understand what the Ducks are up against this year considering their relatively weak OOC schedule. I would love to hear especially from folks who actually understand the math, or who can explain how exactly they can say these rankings are unbiased from a mathematical or philosophical standpoint.