In my ongoing attempt to get the "Pilsch Index" added to the BCS statistical calculations, I am once again spending Sunday mornings arguing with my computer's BCS ranking program. Generally, my program is incapable of producing any coherent results during the first three weeks of the season (for instance, it had FIU as the #5 team in the nation last week). Unlike all of the actual BCS computers (and the BCS human polls, for that matter), my computer only uses on-field data to determine rankings. The AP Poll, USA Today Coaches Poll, and the BCS computers all use preseason rankings (in some fashion) as initial seeds to their ranking systems. In the case of the AP and USA Today polls, this seeding takes the form of the very scientific formula of "hype + historic reputation + rumor mill," all very useful factors in any mathematical calculation.
My program, instead, chooses to provide a snapshot of college football as it is currently being played. It rewards teams playing strong schedules and doing well against them, regardless of reputation or down-the-road strength of schedule. This is why during the first few weeks of the season, the rankings can be a bit skewed.
The Notre Dame Problem
I have made one small change to the way rankings are calculated. This is part of my on-going quest to eliminate outliers. Two years ago, the computer program had a problem with ranking undefeated Idaho higher than some major conference teams, even after they had played good teams (in other words, Idaho was ranked higher than SEC teams who had played other SEC teams; I refuse to acknowledge the fact that other conferences play quality football (I'm (mostly) kidding)). To fix the Idaho problem, I now use the overall quality of opponents as a weight in quality calculation (the old calculation was based merely on raw quality differential; originally, if the #1 team beat the #4 team, this counted the same as if the #31 team beat the #34 team). This season the problem was Notre Dame, as the Fighting Irish so often are when it comes to college football rankings (especially ones that rely on hype and reputation based seedings). I could talk about how there is no less-deserving team in all of college football and how the specter of Ara Parseghian and Knut Rockne almost guarantee the Irish an appearance in the pre-season Top 25, but I'll spare you that rant.
The problem Notre Dame posed to my poll is that, unlike almost every other team in the country, Notre Dame has played four quality opponents and not played an FCS cupcake. They were also helped by the fact that Michigan is really well-thought-of by my computer. The final factor that really caused the jump is that Michigan State, who the Irish beat, is the 10th team in terms of offensive/defensive quality on my rankings, and the Irish's win against them counted as an upset (the Irish's offensive/defensive quality ranking is only #43). So, despite being 2-2, Notre Dame was ranked at #18 (last week, at 1-2, they were ranked in the top 15), because they had played a bunch of good teams and beaten two of them. This condition occurs because their early schedule of quality opponents gave them a strength of schedule factor that was 3X higher than the average strength of schedule factor amongst FBS teams. It was so high that it overcame the other factors I was using and gave them a massive overall quality ranking.
What I have done to correct this, and please tell me in the comments if this doesn't make sense, is I now adjust the strength of schedule ranking by win percentage. I used to adjust the overall quality ranking by win percentage, but this meant that there was no way a one loss team could ever get ranked ahead of no loss teams (even if the one loss was to a good team, a la WVU losing to LSU last night). I think adjusting strength of schedule by win percentage is a much more manageable way of factoring in the overall record, especially as strength of schedule is only one factor in my overall quality ranks.
Duck, Duck, Duck, Goose
I've noticed that there is one team my computer consistently hates, and I've also noticed that this hatred is actually starting to spill over into my own thinking about college football. Oregon consistently has the biggest drift between my computer rankings and hype-based ranking systems like the AP poll. For instance, Oregon doesn't even make it into my Top 25, despite their tenth place ranking in the AP poll.
This differential is a result of the fact that, in my rankings, Oregon has lost to the 8th best team (in terms of offensive/defensive quality; LSU is #1 overall, as you can see below) and beaten the 67th, 98th, and 120th teams (all FCS schools are calculated as the worst team in the FBS; which is why teams that lose to an FBS school are really hurt in the rankings). This kind of schedule is fairly standard for Oregon. They usually play a really weak non-conference schedule and, outside of Stanford and USC, play in a terrible conference. Oregon's ranking will rise if they can beat the other two good teams in the Pac-12 (beating Washington won't hurt either), but, right now, I see no evidence that Oregon is any good.
R.I.P. Mumme Poll
This all used to be tied to the Mumme Poll, but apparently 3SiB isn't doing the poll anymore. I found references to the poll on other blogs, but I can't find a place to register this year. Sad.
Rankings for Week 4
Here's my computer's top 25. Based solely on on-field play, I agree with these rankings. If you asked me to introduce some futurological speculation into my rankings, I would say that Michigan and South Carolina are probably too shaky to last as Top 5 teams, but, so far, they've played really well against good competition.
- Louisiana State 4-0 (373.632575757576)
- South Carolina 4-0 (360.673106060606)
- Oklahoma 3-0 (351.436363636364)
- Michigan 4-0 (346.0)
- Alabama 4-0 (316.793939393939)
- Boise State 3-0 (306.907575757576)
- Oklahoma State 4-0 (301.724104683196)
- Illinois 4-0 (299.375)
- South Florida 4-0 (286.970075757576)
- Florida 4-0 (270.427272727273)
- Texas 3-0 (262.668181818182)
- Clemson 4-0 (260.85)
- Georgia Tech 4-0 (260.256818181818)
- Stanford 3-0 (259.318181818182)
- Nebraska 4-0 (251.256818181818)
- Virginia Tech 4-0 (243.665151515152)
- Florida International 3-1 (238.466818181818)
- Baylor 3-0 (234.15)
- Houston 4-0 (230.369318181818)
- Arizona State 3-1 (212.90625)
- Southern California 3-1 (205.60625)
- Auburn 3-1 (205.410416666667)
- Wisconsin 4-0 (198.133333333333)
- Ball State 3-1 (195.189583333333)
- Kansas State 3-0 (194.065151515152)