2014-11-02

NCAA Football Power Rankings, Week 10

Mississippi State is #1 for the second week in a row and the third time in 4 weeks. My top two are in line with the AP poll and the CFB rankings, although those haven't been released for week 10 yet.

Missouri climbed from #19 to #16 more spots with a win over Kentucky, dropping them from #41 to #49. In a 4-round NCAA Football tournament, I'd have them on the bubble right now.

Kent State and SMU both took the week off, but they swapped positions due to their opponents outcomes this week.

Last week I had a very fair question posed about my methodology, because 6-2 Missouri were ranked ahead of 6-1 Georgia, despite a dominant 34-0 win by Georgia over Missouri. I had a few general ideas of why this could happen, including that in my system, it doesn't matter who you beat and who beats you, it only matter what your record and strength of schedule are. For example, going 1-1 against Oregon and Kent State will result in the same ranking regardless of which team was the win, and which was the loss. I'd like to point out that Georgia lost to a team Missouri beat (Florida) this week, and Missouri won, making it a 7-2 team above a 6-2 team. But I will also point out that that in no way justifies the ranking last week. These are not predictive rankings; they are supposed to be based entirely on past performance.

The short answer to Georgia's low rank can be summed up in 1 word: Troy. Georgia played Troy, and it annihilates their strength of schedule each week.

I only show the ranks for the FBS teams, but I actually keep a ranking all teams in any games shown on the scores pages on espn.com, which was 718 teams as of last week. That's every division 1, 2, and 3 team, as well as any lower division opponents they may play, such as NAIA teams like William Penn, who lost to South Dakota, but played no other NCAA opponents. Those one-off lower division teams cause me to have teams like William Penn listed at 0-1 in my rankings, and I have considered different ways of handling those. For now, they are simply treated as a team that has played only one "real" game. Circling back to Troy, with all 718 teams considered, they have been ranked #673 for the past 2 weeks. When calculating strength of schedule, 673 is the number that counts against Georgia, not the 125 it would be if I was only considering the FBS rankings. I don't have a way to quickly check which method would have penalized Georgia less, but I have a feeling 125 out of a possible 127 would still hit them pretty hard.

The method I use is fairly naive, and the penalty for each opponent is equal to their rank. So, playing the #2 team weakens your schedule twice as bad as playing the #1 team, but playing #101 is only 1% worse than playing #100. There are probably better curves to draw that would more accurately reflect the falloff in difficulty from #1 to #718, but it's hard to determine that dynamically in a closed system. The other half of the ranking equation is record, with a win worth 718, and a loss worth -718. I then add a factor for the difficulty of the team, which ranges from 717 for beating the #1 team, to 0 for beating the #718 team. For losses, a 717 points are added so that a loss to #1 is only a -1 on net, and losing to #718 costs you the full 718 points. If you shift some equations around, the simpler formula is that your score is 2*wins*N + sum(ranks of opponents), which is where the 6-1 vs 6-2 issue feels tricky, but really isn't.

I initially sort the teams by record - actually by wins then losses, so 2-6 is above 1-0, for example - then repeat this ranking procedure over and over until it stabilizes. Every time an opponent moves up or down, their opponents are affected, as are theirs, etc. Sometimes it stabilizes into a loop of 2 ranking sets, so I have to take an average of them, and that was the case this week and last, so these numbers aren't dead on. But, they give you a ballpark figure.

Last week, Georgia's score was:
4308 for a 6-1 record
-27 Clemson
-222 South Carolina
-673 Troy
-315 Tennessee
-476 Vanderbilt
-56 Missouri
-255 Arkansas
=2284

Missouri's was:
4308 for a 6-2 record
-144 South Dakota State
-169 Toledo
-137 UCF
-330 Indiana
-222 South Carolina
-63 Georgia
-308 Florida
-476 Vanderbilt
=2459

Overall, I think these rankings work pretty well. I may play around with the scoring system in the offseason, to see what sorts of changes would put me more in line with other computer rankings, if nothing else for my own curiosity. I certainly don't want to delve into the territory of choosing a system that favors my favorite team, but I'm always open to exploring the new possibilities.

1Mississippi State8-0
2Florida State8-0
3Oregon8-1
4Nebraska8-1
5Auburn7-1
6Ole Miss7-2
7Colorado State8-1
8LSU7-2
9UCLA7-2
10Alabama7-1
11TCU7-1
12Kansas State7-1
13Michigan State7-1
14Notre Dame7-1
15Georgia Tech7-2
16Missouri7-2
17Arizona State7-1
18Marshall8-0
19Ohio State7-1
20Clemson6-2
21Baylor7-1
22Duke7-1
23Boise State6-2
24Miami (FL)6-3
25West Virginia6-3
26Utah6-2
27Oklahoma6-2
28Maryland6-3
29Arizona6-2
30Wisconsin6-2
31Texas A&M6-3
32Georgia6-2
33Minnesota6-2
34USC6-3
35Louisville6-3
36Washington6-3
37Boston College6-3
38Iowa6-2
39Utah State6-3
40Nevada6-3
41Louisiana Tech6-3
42Georgia Southern7-2
43East Carolina6-2
44Air Force6-2
45North Carolina State5-4
46Louisiana-Lafayette5-3
47Northern Illinois6-2
48Central Michigan6-4
49Kentucky5-4
50BYU5-4
51UCF5-3
52Western Michigan6-3
53Toledo5-3
54Stanford5-4
55Oklahoma State5-4
56California5-4
57Houston5-3
58Arkansas State5-3
59Florida4-3
60Memphis5-3
61Tennessee4-5
62Rutgers5-4
63Cincinnati5-3
64North Carolina4-5
65Virginia Tech4-5
66Middle Tennessee5-4
67Illinois4-5
68UTEP5-3
69Temple5-3
70Rice5-3
71Virginia4-5
72Bowling Green5-3
73South Carolina4-5
74Arkansas4-5
75South Alabama5-3
76Pittsburgh4-5
77UAB5-4
78Texas State5-3
79Wyoming4-5
80Michigan4-5
81Penn State4-4
82Texas4-5
83Navy4-5
84Oregon State4-4
85Purdue3-6
86Syracuse3-6
87Northwestern3-5
88San Diego State4-4
89Ohio4-5
90Indiana3-5
91Akron4-4
92San Jose State3-5
93Texas Tech3-6
94Southern Miss3-6
95Ball State3-5
96Vanderbilt3-6
97South Florida3-6
98Iowa State2-6
99Florida International3-6
100Western Kentucky3-5
101Fresno State3-6
102Louisiana-Monroe3-5
103Florida Atlantic3-6
104Buffalo3-5
105New Mexico3-5
106Washington State2-7
107Appalachian State3-5
108Kansas2-6
109Colorado2-7
110Tulane2-6
111Wake Forest2-6
112Connecticut2-6
113UTSA2-6
114Hawaii2-7
115UNLV2-7
116Army2-6
117Eastern Michigan2-7
118North Texas2-6
119Miami (OH)2-8
120New Mexico State2-7
121Massachusetts2-7
122Tulsa1-7
123Georgia State1-8
124Idaho1-7
125Troy1-8
126SMU0-7
127Kent State1-7


2014 History and #1s
Week 1 Everybody who was 1-0
Week 2 LSU
Week 3 Oklahoma
Week 4 North Carolina State
Week 5 Nebraska
Week 6 Auburn
Week 7 Mississippi State
Week 8 Florida State
Week 9 Mississippi State

2 comments:

  1. In fairness, Georgia deserved to be ranked far lower based on how the game against Florida went.

    Forgive me for my error in asking!

    ReplyDelete
    Replies
    1. The way I see it, my system should hold up under scrutiny, so please always feel free to point out things that look wrong or at least misguided. I tend to focus my attention on who's #1, where Missouri is, and just for fun, who's at the bottom. Then, if there was a huge upset that week, I might note how those teams were affected. But I can easily miss an incongruous-looking pair of teams like Missouri and Georgia were last week.

      Delete