Here are the first rankings of the college football season.
Well done TCU and LSU who would be in the national chanpionship game if the season ended today and my rankings were used.
1 TCU
2 LSU
3 Oklahoma
4 Michigan st
5 Oregon
6 Auburn
7 Boise st
8 Stanford
9 Nebraska
10 Missouri
11 Arizona
12 South Carolina
13 Alabama
14 Kansas st
15 Nevada
16 Air Force
17 Florida st
18 Oregon st
19 Ohio state
20 Iowa
21 Michigan
22 Oklahoma st
23 Illinois
24 California
25 Wisconsin
26 Arkansas
27 Mississippi st
28 Iowa st
29 West Virginia
30 Utah
31 Arizona St
32 Maryland
33 Hawai`i
34 Temple
35 Penn state
36 Colorado
37 Navy
38 Florida
39 Northern Illinois
40 UCLA
41 Notre Dame
42 Houston
43 Miami FL
44 Southern Cal
45 Northwestern
46 Baylor
47 Purdue
48 Syracuse
49 Toledo
50 Texas A&M
51 Texas
52 Wyoming
53 Texas Tech
54 UTEP
55 SMU
56 Washington
57 Indiana
58 Idaho
59 San Diego st
60 South Florida
61 Pittsburgh
62 Tennessee
63 Army
64 Western Michigan
65 Georgia
66 Connecticut
67 Brigham Young
68 Southern Miss
69 Louisiana Tech
70 Clemson
71 Troy
72 Washington st
73 Louisiana-Monroe
74 Vanderbilt
75 Ohio U.
76 Central Florida
77 Colorado st
78 North Carolina
79 Buffalo
80 Louisville
81 Miami OH
82 UNLV
83 Duke
84 Tulsa
85 Virginia
86 Rice
87 North Carolina St
88 Rutgers
89 Utah st
90 Wake Forest
91 Marshall
92 Florida Int'l
93 Arkansas St
94 Alabama-Birmingham
95 Florida Atlantic
96 Cincinnati
97 Kent st
98 Bowling Green
99 Louisiana-Lafayette
100 Fresno st
101 Georgia Tech
102 East Carolina
103 Tulane
104 Memphis
105 Western Kentucky
106 New Mexico
107 North Texas
108 Kentucky
109 Boston College
110 Middle Tennessee St
111 New Mexico st
112 Virginia Tech
113 Eastern Michigan
114 Mississippi
115 Central Michigan
116 Ball St
117 Kansas
118 San José St
119 Minnesota
120 aa
121 Akron
How are these ratings calculated? Because to be honest they are a load of crap. Virginia Tech being ranked 112!!! Really!! Behind three teams that they beat convincingly. I know they lost to a D1-AA school, but it was by less than any of their wins. (BTW, only one other system listed in Massey's Comparison puts them below 45).
ReplyDeleteMississippi is behind all three teams they beat, all of whom have worse records against not as good teams. There are other blatant flaws that indicate a lack of credibility, but these are two that are immediately obvious.
Dear Josh
ReplyDeleteThank you for the feedback.
In terms of how the system works the google doc on this blogsite and a blogpost explains it thoroughly but in brief each team is compared to every other team and chains of victories or defeats that connect the two teams are calculated. e.g. Michigan state beat Michigan who beat Notre Dame who beat Boston College. So Michigan State has a primary win over Michigan a second level win over Notre Dame and a third level win over Boston College. Michigan State also beat Notre Dame so they also have a primary win over the Irish. These types of chains are calculated and then added up with long chains counting for less. The probability of Michigan State being "better" than Notre Dame is then calculated. The final rankings are the averages of these probabilities against all other 119 FBS teams and the combined FCS teams ("aa" in the rankings).
Margin of victory and location of a game are not taken into account as per the BCS computer ranking rules.
You actually touched on the reason so many teams in my rankings are much lower than the consensus rankings and that is because they lost to 1-AA teams or lost to teams that lost to 1-AA teams which is punished quite severely by my system.
However on reflection I think I need to adjust how I deal with 1-AA teams, currently they are all grouped together as one "superteam". However this results in a loss to a 1-AA team being much worse than a loss to a bad FBS team, as currently around 70 teams have beaten 1-AA opposition and so gain second level victories over teams such as Virginia Tech or Ole Miss who lost to 1-AA opposition. Whereas losing to a 1-11 team would mean you would only have 11 second level losses as a result of that game.
Therefore for next week (week 7) I will adjust the FCS superteam record by dividing each victory by 7.5 (the FCS superteam plays around 90-95 games a year compared with the normal 12 and 90/12 = ~7.5) so that the number of secondary losses Virginia Tech suffer is proportionate to having lost to a bad team. So as Indiana beat Towson they will now only gain c~13% of a secondary victory over Virginia Tech rather than 100% of a full victory at present.
Dealing with 1-AA teams is a difficulty all computer ranking systems have but hopefully next week the rankings should be fairer to teams who lose to 1-AA opposition.
My ranking has faired very poorly against the Massey site's consensus probably due to the over-penalising of 1-AA teams but it has done well looking at ranking violation% among the top 25 teams so it will be interesting to see how the adjustment affects things next week.
James
I still think your system will never mean much if you lump all 1AA teams together. JMU and ND State are better than VMI and Towson. Counting wins against Kansas and VT for beating teams at the bottom of the FCS makes your poll lose all credibility. You might as well not even count FCS games at all if you are going to lump them all together.
ReplyDeleteAnonymous
ReplyDeleteI think ignoring FCS games is completely wrong (Virginia Tech's loss to James Madison should be ignored?) but how to deal with them is a tricky issue.
Of the 6 BCS approved computer programs I cannot determine how Billingley or A and H deal with the issue. Wolfe, Massey and Sagarin actually rank every team at every level. So Sagarin currently has Delaware at No.18 Massey has them at 39 with Jacksonville at 41 but I assume these are omitted from the rankings on the comparison site. Colley originally ignored FCS teams but he now groups them into groups of teams i.e. FCS 1 FCS 2 etc with FCS group 10 currently at No.95. I remember reading a detailed description of how he did this but I can't now find it on his website.
The Mease system which is a similar MLE to Massey does group all FCS teams as one superteam so I dont think it is without precedent or outrageous.
Ideally I would also rank every team in the country but that is beyond my computer abilities as I am struggling to rank 120+1 hence the need for my fudge. If the correction I discussed above corrects the wildness of some of my rankings does it matter?
As an aside while there are c.109 rating systems for FBS on the Massey site there are only c.40 for FCS so I suspect many of the other ranking systems deal with this issue somehow and I would like to know how other systems deal with this.
JMC