Thursday, August 28, 2014

2014 Pre-Season Rankings



Hey football fans! I’m taking it down to the wire but I did want to post my pre-season ranking before the season kicks-off this evening.
                     
Below you can find information on my pre-season rankings. I rank teams using a Bayesian hierarchical model. Bayesian statistics requires that I specify a prior distribution, which is essentially a best-guess as to the relative strength of each team before any game results are observed. My pre-season rankings are this ‘best-guess’ and are an aggregate ranking from various pre-season rankings that can be found on the internet. Throughout the season, my model will update these rankings as game results are observed, but, unlike the polls, I take a formal statistical approach to updating the rankings, rather than the haphazard approach taken by the pollsters. The take away message from all of this is that I only present the pre-season rankings to give an idea of where we’re starting from and that these shouldn’t be taken too seriously because they will change dramatically over the first month of the season as game results are observed.

Here are the top 10 teams from the pre-season rankings along with their pre-season rating (points above average), two measures of their strength of schedule, their expected final record and probability of going undefeated:


Rank

Team

Rating

Record

SOS1

SOS2

E(FR)

P(WARG)
1
Florida State
25.0
 0-0
2.9
0.191
10.5 – 1.5
0.147
2
Alabama
24.6
 0-0
3.1
0.106
10.1 – 1.9
0.057
3
Oregon
23.8
 0-0
4.1
0.097
9.9 – 2.1
0.045
4
UCLA
23.3
 0-0
8.0
0.083
9.6 – 2.4
0.026
5
Oklahoma
22.5
 0-0
3.9
0.204
10.2 – 1.8
0.123
6
Michigan State
22.4
 0-0
0.6
0.134
10.0 – 2.0
0.047
7
Ohio State
22.4
 0-0
3.5
0.186
10.2 – 1.8
0.075
8
Auburn
22.2
 0-0
7.6
0.026
8.6 – 3.4
0.005
9
South Carolina
21.9
 0-0
5.9
0.073
9.3 – 2.7
0.016
10
Georgia
21.6
 0-0
4.2
0.105
9.6 – 2.4
0.022

Florida State is the unanimous number 1 and returns several key starts, most notably Jameis Winston, from their national title team a year ago. I conservatively set a unanimous number 1 ranked team to be 25 points above average at the beginning of the season but, in reality, Florida State is likely much better than that. For example, the number 1 ranked team has been at least 30 points above average at the end of the last four seasons and Florida State finished the season 37 points above average a year ago. Outside of Florida State, the top 10 includes four SEC teams (Alabama, Auburn, South Carolina and Georgia), two each from the Big Ten (Michigan State and Ohio State) and Pac12 (Oregon and UCLA) and Oklahoma from the Big 12. I should also note that these rankings are based on pre-season polls that didn’t know about Braxton Miller’s injury and Ohio State may therefore be overrated but my guess is that they’ll be a top 20 team without him.


Another way to look at the pre-season rankings is to consider the top 10 teams with the best chance of finishing undefeated:


P(WARG) Rank

Rank

Team

P(WARG)
1
42
Marshall
0.489
2
1
Florida State
0.147
3
5
Oklahoma
0.123
4
7
Ohio State
0.075
5
12
Baylor
0.075
6
2
Alabama
0.057
7
6
Michigan State
0.047
8
3
Oregon
0.045
9
4
UCLA
0.026
10
38
Boise State
0.026

Finishing the season undefeated was of the utmost importance in the BCS era but it remains to be seen what the new playoff selection committee will emphasize when picking the four playoff participants. They claim that strength of schedule will be very important but my guess is that finishing undefeated will still be the best way to assure participation in the playoff. Now, to immediately contradict myself, Marshall has the best chance of finishing undefeated but I don’t think they can be considered real playoff contenders. In this sense, strength of schedule will likely matter but Marshall wouldn’t have been a contender in the BCS era either. Amongst the contenders, Florida State has the best chance of finishing undefeated and may actually be more likely to finish undefeated than the numbers suggest. The only other point worth mentioning is that Baylor is more likely to finish undefeated than their ranking would suggest. This illustrates the advantage of not having to play a conference championship game, which results in one less tough game than will be played by teams in the SEC, Big Ten ACC and Pac 12. Oklahoma also benefits from the lack of a Big 12 title game and are almost 3 times more likely to finish undefeated than Oregon and UCLA even though they are ranked below both teams.

As always, full rankings can be found at my faculty webpage:


Enjoy the first weekend of the season!

Wednesday, January 1, 2014

Happy New Year!



Happy New Year everyone! 

We’re about two thirds of the way through the bowl season and I wanted to update my predictions for the remaining games. We don’t have any new information about these teams, in particular, but we do have more information about the relative strength of each conference due to the bowl results, so far. This is important because of the lack of common opponents for most of the bowl games. Not much has changed but the predicted margin of victory moved by about a half-a-point to a point in a couple of cases. 

My predictions for the remaining bowl games can be found below. Enjoy!

Bowl
Favorite
Underdog
Predicted
MOV
Upset Probability
Date
BCS National Championship Game
 Florida State
 Auburn
18.55
0.093
1/6/14
Orange Bowl
 Ohio State
 Clemson
2.44
0.431
1/3/14
Sugar Bowl
 Alabama
 Oklahoma
12.23
0.192
1/2/14
Fiesta Bowl
 Baylor
 Central Florida
22.71
0.053
1/1/14
Rose Bowl
 Stanford
 Michigan State
5.97
0.335
1/1/14
Cotton Bowl
 Oklahoma State
 Missouri
1.94
0.445
1/3/14
Gator Bowl
 Georgia
 Nebraska
11.34
0.209
1/1/14
Capital One Bowl
 Wisconsin
 South Carolina
1.46
0.459
1/1/14
Outback Bowl
 LSU
 Iowa
10.09
0.236
1/1/14
GoDaddy.com Bowl
 Ball State
 Arkansas State
7.31
0.301
1/6/14
BBVA Compass Bowl
 Vanderbilt
 Houston
0.28
0.492
1/4/14
Heart of Dallas Bowl
 North Texas
 UNLV
5.87
0.338
1/1/14