College Football Computer Rankings: January 2012

Tuesday, January 10, 2012

2011 Final Rankings

1 ALABAMA 21.057
2 LSU 20.531
3 OKLAHOMA ST 17.280
4 BOISE ST 14.526
5 OREGON 14.426
6 OKLAHOMA 12.276
7 STANFORD 12.136
8 MICHIGAN 11.183
9 WISCONSIN 10.507
10 SOUTH CAROLINA 10.214
11 ARKANSAS 10.059
12 SOUTHERN CAL 9.240
13 HOUSTON 8.797
14 BAYLOR 7.672
15 TCU 7.533
16 GEORGIA 6.808
17 SOUTHERN MISS 6.596
18 KANSAS ST 6.278
19 MICHIGAN ST 6.261
20 FLORIDA ST 6.056
21 TEXAS A&M 5.683
22 NOTRE DAME 5.301
23 NEBRASKA 5.054
24 VIRGINIA TECH 4.946
25 WEST VIRGINIA 4.884
26 TEXAS 4.810
27 MISSOURI 4.684
28 CLEMSON 3.996
29 CINCINNATI 3.993
30 TOLEDO 3.496
31 RUTGERS 3.092
32 PENN STATE 2.772
33 TULSA 2.595
34 FLORIDA 2.561
35 TEMPLE 2.435
36 BRIGHAM YOUNG 2.339
37 VANDERBILT 2.312
38 NORTHERN ILLINOIS 2.258
39 MISSISSIPPI ST 2.095
40 LOUISIANA TECH 1.922
41 AUBURN 1.900
42 ARIZONA ST 1.736
43 MIAMI FL 1.578
44 IOWA 1.488
45 OHIO STATE 1.431
46 CALIFORNIA 1.250
47 TENNESSEE 1.192
48 GEORGIA TECH 1.091
49 SOUTH FLORIDA 0.916
50 UTAH 0.885
51 PITTSBURGH 0.863
52 NORTH CAROLINA 0.794
53 LOUISVILLE 0.737
54 NORTH CAROLINA ST 0.566
55 WASHINGTON 0.538
56 NEVADA 0.377
57 SMU 0.138
58 ILLINOIS 0.019
59 ARKANSAS ST -0.013
60 OHIO U. -0.026
61 NORTHWESTERN -0.097
62 NAVY -0.187
63 SAN DIEGO ST -0.407
64 IOWA ST -0.486
65 WESTERN MICHIGAN -0.532
66 TEXAS TECH -0.718
67 AIR FORCE -0.839
68 CONNECTICUT -1.071
69 WESTERN KENTUCKY -1.105
70 PURDUE -1.256
71 VIRGINIA -1.281
72 LOUISIANA-LAFAYETTE -1.333
73 FLORIDA INT'L -1.452
74 WAKE FOREST -1.463
75 CENTRAL FLORIDA -1.479
76 UTAH ST -1.504
77 UCLA -1.515
78 KENTUCKY -1.578
79 WYOMING -2.261
80 MARSHALL -2.645
81 OREGON ST -2.810
82 SYRACUSE -2.835
83 ARIZONA -2.897
84 BALL ST -3.233
85 EAST CAROLINA -3.294
86 MINNESOTA -3.384
87 SAN JOSé ST -3.473
88 BOWLING GREEN -3.487
89 MIAMI OH -3.788
90 HAWAI`I -4.124
91 BOSTON COLLEGE -4.162
92 KENT ST -4.376
93 EASTERN MICHIGAN -4.636
94 DUKE -4.819
95 UTEP -4.968
96 WASHINGTON ST -5.027
97 FRESNO ST -5.162
98 NORTH TEXAS -5.471
99 LOUISIANA-MONROE -5.543
100 RICE -5.759
101 KANSAS -6.103
102 MISSISSIPPI -6.108
103 COLORADO -6.786
104 NEW MEXICO ST -7.640
105 ARMY -8.429
106 MARYLAND -9.400
107 CENTRAL MICHIGAN -10.246
108 BUFFALO -10.289
109 COLORADO ST -10.383
110 ALABAMA-BIRMINGHAM -10.595
111 TROY -10.619
112 UNLV -11.774
113 INDIANA -12.327
114 IDAHO -12.526
115 MIDDLE TENNESSEE ST -13.530
116 NEW MEXICO -15.452
117 MEMPHIS -16.465
118 TULANE -17.973
119 FLORIDA ATLANTIC -18.864
120 AKRON -19.895

Description of Ratings System

My goal was to devise a system that relied exclusively on simple mathematical calculations without any built-in biases. This ranking system is based solely on the resulting score of the games played to date of the current season and the teams win-loss record. I will briefly cover some of my thoughts and philosophies while developing the system in the following.
After analyzing game statistics I concluded that the final score of a game is an all encompassing summary of the statistics. For example, if a team outgained an opponent by 200 yards of offense but still lost the game, the loss was usually due to turnovers or penalties. I started by developing a system that would record and determine the average margin of victory (AMV) for a team. A system based on this alone has obvious flaws due to teams winning by large margins over lesser opponents. After crunching data with various limitations on margin of victories, I decided to limit the margin of victory to a maximum of 21 points. However, this still allowed teams that play weak opponents to remain high in the rankings. To penalize teams that play weak opponents and give credit to teams that play strong opponents, I then added the team’s opponents’ average margin of victory (OAMV) and their opponent’s opponents’ average margin of victory (OOAMV) to their AMV. This process is carried out 5 times (OOOOOAMV) which mathematically relates every team to each other by the end of the season. This effectively lowered the teams rankings that had a high AMV due to playing weak opponents. However, this allowed teams with high AMVs and multiple losses to be ranked higher than no-loss or one-loss teams at the end of the season. I factored in the win-loss record by lowering the final result of a team with a positive ranking value by the ratio of the team’s losses over the total games played and raising the final result of a team with negative ranking value by the ratio of the team’s wins over the total games played.
To illustrate this more clearly, I will use a couple of examples.
As of week 6 of the 2000 season, TCU had the highest margin of victory of the 115 Division 1-A teams at 26.750. However, their opponents AMV was -12.400 indicating that TCU has played weak teams but have outscored these teams by a larger margin than the other opponents of these teams. Summing the AMV and the OAMV reduces their ranking value to 14.350. Carrying this process out 5 times (OOOOOAMV) results in a ranking value of 21.623. TCU is currently undefeated (4-0) so 21.623 is their final ranking value. This places TCU 4th out of 115 teams.
As of week 6 of the 2000 season, Texas A&M’s AMV is 15.000 and their opponents’ AMV was 0.438. Summing these two numbers increases their ranking value to 15.438. Carrying this process out 5 times (OOOOOAMV) results in a ranking value of 16.284. Texas A&M has played four 1-A opponents and their win-loss record is 3-1. Since Texas A&M has a positive ranking value they are penalized for the game they lost by reducing their result by the total number of games they have lost divided by the games played, or 25%. This result is 12.213 which places them 13th out of 115 teams. If Texas A&M wins their game in week 7 their record will then be 4-1 and their final result will be reduced by only 20%. This will enable them to possibly move up in the rankings by continuing to win games, similar to what would be seen in the AP or Coaches poll.
Additional comments:
The results from previous years have no effect on this years rankings. I feel that the two years are independent from each other. This causes some anomalies early in the season but everything settles out after week 6 or 7. I believe this is also why the BCS does not publish their rankings until late October although some ranking systems carry over from the previous year.
This system only uses the 120 Division 1-A teams. Any team that plays a team from another division receives no credit for the win. There can be a slight penalty for playing a team outside of Division 1-A. For instance, a team that has a victory over a 1-AA opponent does not receive credit for that victory in the win column. If their actual record is 7-2 this would be recorded as 6-2, therefore the penalty for the 2 losses would be 25% instead of 22.2 %. Likewise, losing to a 1-AA opponent would not be recorded in the loss column but this rarely happens.
This system is simple compared to some of the other systems being used but the results similar. I would welcome any comments or questions from anyone. I can be contacted at smaurer@kc.rr.com
Scott

College Football Computer Rankings

Tuesday, January 10, 2012

2011 Final Rankings

Description of Ratings System

Blog Archive

About Me