Following up on our posts last year on a new approach to ranking college football teams, we provide our first rankings this season for the 138 FBS teams. We again considered only games played between FBS teams and not games against FCS teams. And we made one small adjustment to our methodology. In our final rankings last season, we weighted cells below the diagonal by the distance below the diagonal divided by the number of teams, we gave equal weight to all cells above the diagonal, and we assigned 20 percent of the weight to cells above the diagonal and 80 percent to cells below the diagonal. We noted last year that our results were not very sensitive to that weighting. In fact, the cells above the diagonal had little effect on the model. So, we decided it was preferable not to give them any weight.
So, in a ranking of ten teams, the cell weights are now as shown in the table below. Of course, most cells below the diagonal still have 0’s, because there are relatively few games where a lower ranked team in our model defeated a higher ranked team. Indeed, the model is simply trying to find the ranking that minimizes the penalty for those “upsets.”
|
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
|
0.61 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
|
1.21 |
0.61 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
|
1.82 |
1.21 |
0.61 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
|
2.42 |
1.82 |
1.21 |
0.61 |
0 |
0 |
0 |
0 |
0 |
0 |
|
3.03 |
2.42 |
1.82 |
1.21 |
0.61 |
0 |
0 |
0 |
0 |
0 |
|
3.64 |
3.03 |
2.42 |
1.82 |
1.21 |
0.61 |
0 |
0 |
0 |
0 |
|
4.24 |
3.64 |
3.03 |
2.42 |
1.82 |
1.21 |
0.61 |
0 |
0 |
0 |
|
4.85 |
4.24 |
3.64 |
3.03 |
2.42 |
1.82 |
1.21 |
0.61 |
0 |
0 |
|
5.45 |
4.85 |
4.24 |
3.64 |
3.03 |
2.42 |
1.82 |
1.21 |
0.61 |
0 |
We ran only the model that does not adjust for when the game occurred during the season, but that does adjust for the margin of victory and the game location, using codes designed so a win by 17 points or more is worth twice as much as a 1-point (or overtime) win, with an average bonus of 3 points for road victories. Following is our model’s ranking of the 138 FBS teams.
|
Rank |
Team |
|
1 |
Utah |
|
2 |
Georgia |
|
3 |
Alabama |
|
4 |
Kentucky |
|
5 |
Tennessee |
|
6 |
Louisiana St. |
|
7 |
Auburn |
|
8 |
South Carolina |
|
9 |
Florida |
|
10 |
Mississippi |
|
11 |
Texas |
|
12 |
Missouri |
|
13 |
Mississippi St. |
|
14 |
Oklahoma |
|
15 |
Ohio St. |
|
16 |
Miami |
|
17 |
Wake Forest |
|
18 |
California Los Angeles |
|
19 |
Indiana |
|
20 |
Brigham Young |
|
21 |
Notre Dame |
|
22 |
Michigan |
|
23 |
Iowa |
|
24 |
Minnesota |
|
25 |
Purdue |
|
26 |
Washington |
|
27 |
Wisconsin |
|
28 |
Arizona |
|
29 |
Cincinnati |
|
30 |
Texas Tech |
|
31 |
Houston |
|
32 |
Iowa St. |
|
33 |
West Virginia |
|
34 |
Oklahoma St. |
|
35 |
Southern California |
|
36 |
Oregon |
|
37 |
Illinois |
|
38 |
Boise St. |
|
39 |
New Mexico |
|
40 |
Nebraska |
|
41 |
Northwestern |
|
42 |
Oregon St. |
|
43 |
Penn St. |
|
44 |
Texas San Antonio |
|
45 |
Michigan St. |
|
46 |
Central Florida |
|
47 |
Louisville |
|
48 |
Southern Methodist |
|
49 |
Duke |
|
50 |
North Carolina |
|
51 |
Clemson |
|
52 |
Texas Christian |
|
53 |
Pittsburgh |
|
54 |
Virginia Tech |
|
55 |
Boston College |
|
56 |
Stanford |
|
57 |
Georgia Tech |
|
58 |
Colorado |
|
59 |
Baylor |
|
60 |
Texas A&M |
|
61 |
Louisiana Tech |
|
62 |
Western Michigan |
|
63 |
Toledo |
|
64 |
Buffalo |
|
65 |
Texas St. |
|
66 |
Colorado St. |
|
67 |
San Diego St. |
|
68 |
James Madison |
|
69 |
Florida St. |
|
70 |
Virginia |
|
71 |
Syracuse |
|
72 |
California |
|
73 |
Arkansas |
|
74 |
Vanderbilt |
|
75 |
North Carolina St. |
|
76 |
North Texas |
|
77 |
Arizona St. |
|
78 |
Kansas St. |
|
79 |
Kansas |
|
80 |
Tulsa |
|
81 |
Army |
|
82 |
Tulane |
|
83 |
Florida Atlantic |
|
84 |
Memphis |
|
85 |
Nevada Las Vegas |
|
86 |
North Dakota St. |
|
87 |
South Florida |
|
88 |
Air Force |
|
89 |
Navy |
|
90 |
Florida International |
|
91 |
Jacksonville St. |
|
92 |
Middle Tennessee St. |
|
93 |
Wyoming |
|
94 |
Fresno St. |
|
95 |
Nevada |
|
96 |
Ohio |
|
97 |
Central Michigan |
|
98 |
South Alabama |
|
99 |
Marshall |
|
100 |
San Jose St. |
|
101 |
Alabama Birmingham |
|
102 |
Louisiana |
|
103 |
Hawaii |
|
104 |
Georgia Southern |
|
105 |
Appalachian St. |
|
106 |
Charlotte |
|
107 |
Arkansas St. |
|
108 |
Georgia St. |
|
109 |
Liberty |
|
110 |
Coastal Carolina |
|
111 |
Delaware |
|
112 |
Miami-Ohio |
|
113 |
Temple |
|
114 |
Connecticut |
|
115 |
Maryland |
|
116 |
Old Dominion |
|
117 |
East Carolina |
|
118 |
Eastern Michigan |
|
119 |
Massachusetts |
|
120 |
Rutgers |
|
121 |
Bowling Green St. |
|
122 |
Kent St. |
|
123 |
Ball St. |
|
124 |
New Mexico St. |
|
125 |
Western Kentucky |
|
126 |
Utah St. |
|
127 |
Louisiana Monroe |
|
128 |
Troy |
|
129 |
Southern Mississippi |
|
130 |
Missouri St. |
|
131 |
Sam Houston St. |
|
132 |
Kennesaw St. |
|
133 |
Texas El Paso |
|
134 |
Northern Illinois |
|
135 |
Washington St. |
|
136 |
Rice |
|
137 |
Akron |
|
138 |
Sacramento St. |
As the discussion typically focuses on the teams at the top of the rankings, we provide a second table with the top 25 teams in our model and, for comparison, the 25 teams in the current AP rankings. Two general conclusions seem clear. First, our model loves the SEC, as 13 of its top 14 teams are from that conference. Second, our model is not as impressed with the Big Ten or the ACC or with Notre Dame.
|
Rank |
Our Model |
AP |
|
1 |
Utah |
Miami |
|
2 |
Georgia |
Ohio St. |
|
3 |
Alabama |
Georgia |
|
4 |
Kentucky |
Indiana |
|
5 |
Tennessee |
Notre Dame |
|
6 |
Louisiana St. |
Utah |
|
7 |
Auburn |
California Los Angeles |
|
8 |
South Carolina |
Brigham Young |
|
9 |
Florida |
Kentucky |
|
10 |
Mississippi |
Alabama |
|
11 |
Texas |
Texas |
|
12 |
Missouri |
Mississippi |
|
13 |
Mississippi St. |
Texas Tech |
|
14 |
Oklahoma |
Wake Forest |
|
15 |
Ohio St. |
Louisiana St. |
|
16 |
Miami |
Tennessee |
|
17 |
Wake Forest |
Florida |
|
18 |
California Los Angeles |
Houston |
|
19 |
Indiana |
Southern Methodist |
|
20 |
Brigham Young |
Boise St. |
|
21 |
Notre Dame |
Northwestern |
|
22 |
Michigan |
Arizona |
|
23 |
Iowa |
Missouri |
|
24 |
Minnesota |
Oklahoma St. |
|
25 |
Purdue |
Wisconsin |
The challenge, of course, is that top teams in different conferences play very few games against each other during the regular season. The only nonconference games so far this season between the top 25 teams in our model were Texas’s win over Ohio St., Michigan’s win over Oklahoma, Brigham Young’s win over Notre Dame, Mississippi St.’s win over Minnesota, Wake Forest’s win over Purdue, and Notre Dame’s win over Purdue. Miami, which is ranked first by the AP, is only number 16 in our model because its best win is over Wake Forest and Wake Forest’s best win is over Purdue; neither Miami nor Wake Forest has shown it can beat a good SEC or Big Ten team or even a middle-of-the-pack SEC team. Similarly, Ohio St., which is ranked second by the AP, is ranked number 15 in our model because it lost to a middle-of-the-pack SEC team and because its best win is over Indiana, whose best win is over Michigan; again, neither Ohio St. nor Indiana (nor California Los Angeles) has shown it can beat a good or middle-of-the-pack SEC team. Similarly, neither Brigham Young nor Notre Dame has shown it can beat a good SEC or Big Ten team or even a middle-of-the-pack SEC team.
Our model is kinder to Utah, ranking it first even though it hasn’t defeated any of the other top 25 teams in our model. Indeed, Miami has a better win than Utah does, so it isn’t clear why Utah is ranked above the grouping of SEC teams and Ohio St., while Miami is ranked below those teams. The Utah/Brigham Young and Miami/Notre Dame games on November 7 should help clarify the rankings of those teams. It will be challenging for any Big Ten teams to overtake at least the top several SEC teams in our model during the regular season, because the top Big Ten teams have no significant nonconference games remaining. Ohio St.’s loss to Texas, in particular, weighs heavily on the Big Ten teams in our model; the Big Ten’s best hope may be for Oklahoma to improve its standing in the SEC so that Michigan’s win over Oklahoma has more influence on the highest ranked teams in our model.