For the third year in a row, I’m posting my computer rankings here. On top of that, I have added a conference ranking, based upon the average ranking for each conference. First, though, some details about the rankings:
My implementation takes the following into account:
1.) Wins – I don’t take the number of losses into account, only wins. I did this for two reasons. One, is that if Team A accumulates more wins than Team B, we can fairly safely assume that Team A also lost fewer games. It does not make sense to count losses twice, once directly and once indirectly. The second reason is that I did not want to have a system that would punish the loser of a conference championship game relative to a team that did not even make it to the conference championship game (though, I admit conferences that don’t have such a game are hurt somewhat – that’s a tradeoff that I decided to accept during design and is worthy of criticism). So, total wins is considered. Wins are determined somewhat by who the opponents are. A win against an FBS team (or Division IA) counts as a full win. A win against a non-FBS team by at least a factor of four counts as a full win. A win against a non-FBS team by at least a factor of two but less than a factor of four counts as a half win. Nothing else counts as a win.
2.) Strength of schedule – This is calculated by taking into account opponents’ wins and opponents’ opponents’ wins, each weighted such that each level cumulatively counts less than its parent level (with both counting less than total wins).
3.) Average margin of victory – This tends to get the most criticism from people. So, check out the explanatory details below.
The total point value for each team is roughly determined by the following equation (up to this point, I have chosen to not widely distribute the exact algorithm):
Total Points = wins + (strength of schedule / games played) + average margin of victory
Teams are then ordered by descending point totals.
So, onto the average margin of victory issue…
Yes, I take margin of victory into account. Why? Because I just can’t get this hypothetical scenario out of my head. Consider Team A and Team B (again). Let’s say that they have played out schedules of similar difficulty and have both gone 12-0 through a regular season. Team A has won those twelve games by a total of 12 points (thus, we can conclude that they must have won each game by exactly one point). Team B, on the other hand, has won its twelve games by an average of 30 points. There is just no way that I can look at that and consider these two teams equal. So, I need some way to distinguish their respective performances through those twelve games.
Most often, I have been presented with the single-game scenario as a counter to this. “If Team A beats Team B by 30 points and Team C beats Team B by two points, you can’t draw any conclusion about who is better between Team A and Team C,” I have been told. And, I have to say that I agree. But, this isn’t a single-game scenario (or two-game scenario, either, especially once we’re past week 2). I have thus far come to the conclusion that over the course of twelve (and for some teams, thirteen) games, an average margin of victory is a completely reasonable metric of team performance. If a team consistently wins by large margins, shouldn’t that be an indicator that it is better than a team that consistently just squeaks by against similar competition?
The other argument against that I have heard is that we don’t want a system that rewards running up the score. I’m not concerned about that for three reasons. One, only you all who have received this e-mail are aware of this system. So, I’m not concerned that coaches and players are going to score huge amounts of points against vastly inferior opponents solely to get higher scores in my rankings (and since the actual BCS computer rankings cannot take margin of victory into account, again there is no motivation for it). Secondly, margin of victory is evaluated on a base-10 logarithmic scale in my system. Therefore, for a win to count twice as much as another, it must win by a margin ten times greater than the other (more importantly, the difference between a 40-point win and a 50-point win is miniscule). So, ultimately, you can think of this as the fine adjustment.
The biggest reason, though, that I have no problem with including margin of victory in the calculation is that the current BCS already uses it. “What?” you say. Didn’t I already say that the BCS computer rankings can’t take margin of victory into account? I did say that. But, if you can point out to me a single voter in the coaches poll who isn’t more impressed with Oklahoma beating Missouri by 40 points than Ohio State beating Iowa by four, then I’ll retract that statement. Might as well have a system that will apply the same margin of victory evaluation evenly across all teams.
All of that said, the three components can be thought of this way:
1.) Wins – most important part of the rankings (coarse adjustment)
2.) Strength of schedule – moderately important (medium adjustment)
3.) Margin of victory – least important (fine adjustment)
Lastly, a couple of caveats regarding the first week’s rankings:
1.) There is a reason why the BCS chooses to not release its rankings until sometime in October. The computer rankings take a few weeks to start finding an equilibrium. Right now, the rankings are based on a relatively small sample size of data. So, you’ll see some results that are intuitively inaccurate. Fear not! Things will right themselves in due time.
2.) A small explanation for the week 1 rankings… In the first week only, the rankings are determined by winning and margin of victory. Strength of schedule has no meaning right now due to a quirk that is special to the first week. Right now, all of the teams that are 1-0 share the worst strength of schedule because all of their opponents are winless. Conversely, all of the teams that are 0-1 share the strongest strength of schedule because all of their opponents are unbeaten. As you can imagine, this effect will diminish as the season goes on.
3.) Early in the season, there may be teams with identical scores. I post them as sharing the same ranking.
And now, the entire 120-team rankings for the first week of the 2011 season…
1.) Central Florida
1.) Cincinnati
3.) Stanford
4.) Virginia Tech
5.) Rutgers
6.) Ohio State
7.) Mississippi State
8.) Arkansas
9.) Florida
10.) Alabama
11.) Texas Tech
12.) Wisconsin
13.) Virginia
14.) Temple
15.) Florida State
15.) Penn State
17.) Texas A&M
18.) Nebraska
18.) Oklahoma
20.) Connecticut
20.) Eastern Michigan
20.) North Carolina
23.) Arizona
24.) Florida International
25.) Iowa
25.) Oklahoma State
27.) Texas
28.) Northern Illinois
29.) Clemson
29.) Michigan
31.) Michigan State
32.) Washington State
33.) Georgia Tech
33.) West Virginia
35.) Ohio
36.) Illinois
37.) Pittsburgh
37.) South Carolina
39.) Toledo
40.) Bowling Green
40.) Hawaii
42.) Arizona State
43.) Vanderbilt
44.) California
45.) Boise State
45.) San Diego State
47.) LSU
47.) Tennessee
49.) Kentucky
49.) Missouri
51.) Navy
52.) North Carolina State
53.) Utah
54.) Maryland
55.) Northwestern
56.) Central Michigan
57.) Ball State
57.) Syracuse
59.) Louisville
60.) Kansas
61.) Colorado State
62.) Air Force
63.) Auburn
63.) Houston
65.) Tulane
66.) South Florida
67.) Purdue
68.) Baylor
68.) Southern California
68.) Southern Mississippi
71.) UTEP
72.) BYU
73.) Kansas State
73.) Washington
73.) Wyoming
76.) Arkansas State
76.) Army
76.) Boston College
76.) UCLA
76.) Colorado
76.) East Carolina
76.) Idaho
76.) Indiana
76.) Louisiana Tech
76.) Memphis
76.) Miami (FL)
76.) Mississippi
76.) New Mexico State
76.) North Texas
76.) Oregon
76.) SMU
76.) Troy
76.) Utah State
94.) Akron
94.) TCU
96.) Buffalo
96.) Florida Atlantic
96.) Fresno State
96.) Georgia
96.) Kent State
96.) Louisiana – Lafayette
96.) Louisiana – Monroe
96.) Marshall
96.) Miami (OH)
96.) Middle Tennessee State
96.) Minnesota
96.) UNLV
96.) New Mexico
96.) Notre Dame
96.) Rice
96.) San Jose State
96.) Tulsa
96.) Wake Forest
96.) Western Kentucky
96.) Western Michigan
116.) UAB
116.) Duke
116.) Iowa State
116.) Nevada
116.) Oregon State
Almost as important, the conference rankings (with average rank for member teams is:
1.) Big East (34.75)
2.) Big 12 (38.67)
3.) Ben Ten (38.83)
4.) SEC (41.00)
5.) ACC (48.67)
6.) MAC (55.92)
7.) Pac-12 (56.83)
8.) Mountain West (71.50)
9.) WAC (72.44)
10.) Independents (73.75)
11.) C-USA (75.00)
12.) Sun Belt (81.33)
October 9, 2011 at 8:54 am |
[...] As a recap of the ranking’s design, you can go here. [...]
October 16, 2011 at 9:00 am |
[...] As a recap of the ranking’s design, you can go here. [...]
October 23, 2011 at 8:34 am |
[...] Anyways, as a recap of the ranking’s design, you can go here. [...]
November 13, 2011 at 9:23 am |
[...] Even I don’t always agree with the results. But, I didn’t design these rankings to match my view of the college football landscape, but to provide an unbiased one. If I were to make changes every single time I see a pair of teams that are one spot off (by my estimation), I would never be done with the design AND it would constantly defeat the original purpose. Please try to remember those two points as you read through these rankings. As always, you can view the general design here. [...]
November 22, 2011 at 12:31 pm |
[...] delay… here are the rankings for week 12. For a review of the overall design, you can go here. This week, Houston’s position was actually strengthened relative to others because of the [...]
November 27, 2011 at 10:58 am |
[...] now… The week 13 team rankings (as always, you can go here to see the general design of the [...]
December 4, 2011 at 10:27 am |
[...] came up that produced results that I believe to be faulty, albeit with only one team. Recall from here that my system awards points for wins, strength-of-schedule, and margin of victory. So, every [...]
December 12, 2011 at 11:17 am |
[...] And so, with that, I choose to not leave out the minor updates to the rankings that result from this game in early December each year. As such, I give you my final computer rankings for the 2011 season (for a link to the general design, you can go here): [...]