Thursday, July 23, 2009

Recruiting Based Ranking – How It Held Up

We based my preseason BlogPoll in 2008 entirely on a statistical analysis I did employing recruiting rankings tweaked for 2007 seasonal performance.

We took some ribbing from BlogPoll maestro Brian at MGoBlog for our methods, which was then defended by Henry as relatively accurate, especially when compared to the BlogPoll itself for 2007.

So how did my tweaked method hold up last year, at least as compared to the preseason BlogPoll when using the final Coaches’ Poll as a success measure?

First I knew certain teams in our method weren’t likely to hold up, or even seemed absurd, such as Michigan at 8th, Miami at 15th or (especially) Notre Dame at 16th. However FSU, over ranked at 10th in my poll, did end the season ranked 23rd, and was nowhere to be found in the BlogPoll.

Here were our 2008 preseason rankings –

1 Southern Cal 4.856
2 LSU 4.555
3 Ohio State 4.519
4 Florida 4.440
5 Georgia 4.439
6 Texas 4.404
7 Oklahoma 4.376
8 Michigan 4.227
9 Tennessee 4.067
10 Florida State 4.046
11 Auburn 3.980
12 Clemson 3.972
13 Alabama 3.948
14 Penn State 3.897
15 Miami-FL 3.847
16 Notre Dame 3.785
17 Virginia Tech 3.763
18 California 3.721
19 Oregon 3.717
20 UCLA 3.717
21 Missouri 3.695
22 Kansas 3.683
23 Arizona State 3.679
24 South Carolina 3.660
25 Nebraska 3.639

And here was the 2008 preseason BlogPoll –

1 Georgia (24)
2 Ohio State (17)
3 Southern Cal (15)
4 Oklahoma (6)
5 Florida (14)
6 Missouri
7 LSU (2)
8 Clemson
9 West Virginia
10 Auburn
11 Texas
12 Texas Tech
13 Wisconsin
14 Virginia Tech
15 Kansas
16 Arizona State
17 Tennessee
18 Oregon
19 South Florida
20 Brigham Young
21 Penn State
22 Illinois
23 Wake Forest
24 Alabama
25 Utah

By final tally we got 13 of the eventual top 25 correct per the Coaches’ Poll. The BlogPoll did one better with 14 correct.

But our method was even better for the top 10 than the BlogPoll’s collective wisdom. Only 3 of my top 10 teams ended up unranked at all (LSU, Michigan and Tennessee) while 4 of the BlogPoll teams were unranked by season end (LSU, Clemson, West Virginia and Auburn).

Moreover fully half (5 of 10) of our top 10 teams were final top ten teams (USC, Florida, Georgia, Texas and Oklahoma) while the BlogPoll got 4 right in this category (Georgia, USC, Oklahoma and Florida).

Our initial poll versus the BlogPoll on several notables is below -

We were about the same as the BlogPoll on Florida and USC. However our method found far more strength in eventual number 3 Texas than the BlogPoll, and was more accurate in forecasting 2008 “surprises” Alabama and Penn State. Our method, being relatively unbiased, also missed the groupthink that overrated Georgia and Missouri.

So two years after implementing a recruiting based preseason polling method we can say that it is at least as good, if not better, than the traditional preseason polling methods. When it came to forecasting the top ten last year it was better.

We will use it again. Our preseason top 25 tomorrow.

No comments: