Last year, I applied the accuracy scoring metric utilized by the Stassen Preseason Poll (preseason.stassen.com) to my own conference predictions. My projections had the same point total as the first place finisher, the Compughter Ratings, putting me ahead of all the preseason magazine publishers for the 2011 season, including Phil Steele, Athlon, Lindy’s, etc.
So what did I do after having such a good year? I went to work revamping all my prediction techniques and developed some innovative, unique modeling methods along the way. I logged countless hours, collected new types of data, simulated historical seasons, and significantly improved my procedures.
My 2012 conference predictions are presented below. Estimated probabilities of winning the conference or reaching the conference championship game (where applicable) are included, which I believe is a fairly unique feature, certainly not available in the preseason magazines. It allows you to see which conference and divisional races figure to be tight (e.g. SEC East, SBC) or where one team has a huge advantage (e.g. ACC Atlantic, Big 12).
Note: I’ve projected divisional finishes for Ohio State and North Carolina, but both are ineligible from their respective conference championship games, so I noted that in the corresponding tables. The projections for the Indepedents are based on expected final record, since they don’t have a conference affiliation.
Wtf? Houston has a college football team? Who outside Texas knew that?
I can certainly attest that jjmcillece did *a lot* of work on this, having discussed what he does at some length with him.
I can’t speak as a diehard NCAA football fan, but only as a data analyst. Odd things happen with data, and sometimes you make a prediction that seems to contradict all sense because of the available data. Why? Sometimes you have a team undergoing big changes in its starting lineup. Those changes can be so different from the vast majority of your observed data, and the best you can do is assume that the team will behave the same way as everything else. It’s like saying that if a 20-year old can run faster than a 10-year old, then a 60-year old can run much faster than the 20-year old. A linear assumption like that is usually dangerous, sometimes it’s clearly stupid, and sometimes it’s the best you can do when you’re a one-man operation working with data in 10+ dimensions.
The prediction for Houston might be a result of a bad assumption. Only time will tell for sure. Anybody who analyzes data won’t strongly criticize you for it.
Thanks Scott. Houston will likely cost me, as I probably should’ve looked more closely at the diagnostics. They were an extreme outlier in several dimensions, and I should’ve watched out for extrapolation. Anyway, I’m sure I’ll continue to hear about it, and the reaction is understandable. Looking forward to the season and learning some lessons the hard way. 🙂
Good luck now that you’re “under the microscope”. My model finished #4 in my first year. My site still has a lot to be added, but you can see some of the predictions now. You might want to start here:
http://fortheresy.webs.com/preseason.htm
I have to agree with the others about Houston. Yes, they aren’t the same, and I do have them losing to UCLA and La Tech. But they don’t play UCF or USM, they have Tulsa at home, and the rest of the East pretty much blows (I’m mystified that so many sites think highly of SMU).
Enjoy the season.
Thanks for the comment. You make good points. Houston had such a tremendous passing offense last year that I saw the losses of Sumlin, Keenum and their top receivers as devastating to the offensive side. Perhaps I overestimated those losses.
My CUSA West predictions clearly riled some people up. If they prove to be awful, I’ll acknowledge it after the season. Time will tell, and good luck to the Cougars.
Houston has not finished last in 10 years in CUSA. They have had the best recruiting class in 2 of the last 4 years and have the Freshman of the year starting at QB. He is no Case (who is???), but will still lead the league in yards and TD’s. We have the best back in the conference also returning and got one of the best wide receiver recruits, Greenberry (a 5 star in Scout). Four of the 5 Offensive linemen are back and the majority of an improving Defense.
I am not saying we will win the west, but everyone else predicts us to be 1, 2, or 3. There is NO WAY we finish last in the West. NO WAY. You had better check you advanced methodology and ‘tweek’ it quite a bit.