Week 8 and Prediction Model 2.11 - Results

This weekend lived up to the season's hype with top 10 teams ceding to lower and unranked teams. Vanderbilt put up the most surprising performance of the weekend. That game convinced me that, while Kentucky is a really good team (more discussion below), South Carolina is just posing. I watch Kentucky and I am impressed with the talent they have on the field. I watch South Carolina and I am impressed that they are not losing more often.

Kentucky is relatively unique in the SEC in that they can consistently put up some points against the SEC defenses. They have only been held under 40 twice, once at South Carolina (23) when Woodson struggled under pressure and Saturday against Florida (37). They were able to use that offensive prowess to comebacks against LSU and Florida, falling short in the Swamp. For all its worth in the national polls, going on the road after the win they had last week playing against a team that has had the week off, to keep it as tight as they did convinces me that Kentucky would be the superior team on most Saturdays.

Before anyone gets any wrong ideas, UCLA is not a good football team. They were heroically trashed by Utah and completely outplayed by BYU. UCLA lost to Notre Dame! Badly! The PAC 10 just isn't that good this year. They have potential and talent, but so does Notre Dame. They have three key victories this season against non-conference foes: USC beating Nebraska (my little sister could beat Nebraska), Cal beating Tennessee (at home, not impressed), and Oregon running over Michigan--who had lost the week before to a 1-AA opponent. I would be a little more impressed with Oregon, but I had the privilege last year of watching them get totally obliterated by BYU in their bowl game and I see no reason they should be that much better.

I would also like to announce that I have officially jumped off the Illinois band wagon and jumped on the Kansas wagon. Losing to Iowa, struggling against a Wisconsin team that is not good, and getting defeated soundly by a Michigan team against which they match up well is not a good way to win my confidence. Kansas struggled against Colorado, and hasn't played any real competition, but they did beat the same Colorado team in the same stadium where Oklahoma fell. Most impressive, though, was the Kansas drive after Colorado had scored to take the lead. They stormed down the field and looked absolutely unstoppable. The team is surprisingly fast and smart, and every bit as legitimate as Kentucky.

Now, to PM 2.11. Week 8 was not a good week. It barely topped 50% on its picks and was 40% against the spread. A blind man throwing darts should average 50% against the spread, but PM 2.11 in 8 weeks has never exceeded 45%, so I did a little investigation.

It seems that the model is too conservative. It does relatively well in close games, but it always picks the under dog when one team is clearly better than the other. This inkling tested out when I found that there was a meaningful negative correlation between magnitude of the spread and probability that PM 2.11 beats the spread. In later models I will have to adjust. This week I hope to unveil PM 2.5, which will include some corrections for home field advantage and 1-AA opponents and will attempt to adjust for match ups in the running and passing games.

(To Most Recent)