• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Dullard's College Football - the real Week 7 (Oct 8)

dullard

dullard

Elite Member
Still early on in the season, so computers tend to shuffle teams around quite a bit.

Big movers up: Georgia, Oregon, Minnesota, FLorida, Colorado, South Florida, Louisville, and Boston College

Big movers down: Notre Dame, Ohio State, Wisconsin, Michigan St, Arizona St, Michigan, North Carolina St, California, and Southern Miss

The rest of the top 30 or so are basically unchanged. I'll be back tomorrow for some comments. But since I haven't watched many games this year, I won't have much to say.

Place : Win Rating / Score Rating ( W , L ) Team name
01 : 66.3 / 66.5 ( 5 , 0 ) Southern Cal
02 : 65.8 / 59.8 ( 6 , 0 ) Penn State
03 : 63.9 / 60.8 ( 6 , 0 ) Virginia Tech
04 : 62.9 / 56.4 ( 5 , 0 ) Alabama
05 : 62.3 / 61.2 ( 5 , 0 ) Texas
06 : 61.7 / 59.3 ( 4 , 1 ) Miami FL
07 : 59.3 / 53.5 ( 5 , 0 ) Florida St
08 : 59.0 / 50.8 ( 5 , 0 ) Georgia
09 : 56.5 / 52.2 ( 5 , 1 ) Oregon
10 : 56.4 / 52.4 ( 5 , 1 ) Minnesota
11 : 55.7 / 52.2 ( 3 , 1 ) LSU
12 : 54.9 / 48.6 ( 5 , 1 ) Florida
13 : 54.5 / 51.1 ( 4 , 1 ) Colorado
14 : 53.5 / 50.9 ( 5 , 0 ) UCLA
15 : 52.3 / 53.3 ( 4 , 1 ) Notre Dame
16 : 52.2 / 54.2 ( 3 , 2 ) Ohio State
17 : 51.8 / 49.7 ( 4 , 1 ) Auburn
18 : 51.8 / 49.9 ( 3 , 2 ) South Florida
19 : 51.8 / 48.8 ( 4 , 1 ) Louisville
20 : 50.9 / 45.3 ( 5 , 1 ) Wisconsin
21 : 49.6 / 51.4 ( 4 , 1 ) Michigan St
22 : 49.4 / 46.0 ( 5 , 1 ) Boston College
23 : 49.1 / 48.9 ( 3 , 3 ) Arizona St
24 : 48.8 / 51.1 ( 3 , 1 ) Fresno St
25 : 47.9 / 43.9 ( 3 , 2 ) Georgia Tech
26 : 47.3 / 40.3 ( 3 , 2 ) Tennessee
27 : 45.8 / 47.0 ( 3 , 3 ) Michigan
28 : 45.5 / 45.4 ( 2 , 2 ) North Carolina St
29 : 45.4 / 47.5 ( 5 , 1 ) California
30 : 45.1 / 38.5 ( 5 , 1 ) West Virginia
31 : 44.8 / 40.8 ( 5 , 0 ) Texas Tech
32 : 44.2 / 38.0 ( 3 , 2 ) Northwestern
33 : 43.6 / 36.0 ( 5 , 1 ) TCU
34 : 41.8 / 38.0 ( 4 , 1 ) Nebraska
35 : 41.4 / 38.7 ( 3 , 2 ) Texas A&M
36 : 41.2 / 35.6 ( 3 , 3 ) South Carolina
37 : 40.6 / 37.2 ( 3 , 2 ) Colorado St
38 : 40.5 / 36.1 ( 2 , 3 ) Oklahoma
39 : 40.4 / 38.3 ( 2 , 2 ) Southern Miss
40 : 40.2 / 38.0 ( 2 , 3 ) Clemson
41 : 39.7 / 31.3 ( 4 , 1 ) Kansas St
42 : 39.5 / 35.2 ( 4 , 2 ) Maryland
43 : 39.2 / 38.5 ( 4 , 2 ) Iowa
44 : 38.7 / 33.0 ( 4 , 1 ) Indiana
45 : 38.7 / 34.3 ( 3 , 2 ) Missouri
46 : 38.6 / 34.3 ( 3 , 2 ) Kansas
47 : 38.4 / 35.0 ( 2 , 3 ) North Carolina
48 : 38.1 / 32.9 ( 4 , 2 ) Wyoming
49 : 37.5 / 32.2 ( 3 , 2 ) Oregon St
50 : 37.4 / 34.0 ( 3 , 2 ) Houston
51 : 37.3 / 35.7 ( 4 , 1 ) Toledo
52 : 37.1 / 32.1 ( 3 , 2 ) Central Florida
53 : 37.1 / 39.1 ( 4 , 1 ) Connecticut
54 : 36.7 / 30.5 ( 3 , 1 ) UTEP
55 : 35.6 / 31.6 ( 3 , 3 ) Tulsa
56 : 35.4 / 28.4 ( 3 , 2 ) Boise St
57 : 35.4 / 31.1 ( 4 , 1 ) Baylor
58 : 35.0 / 33.1 ( 2 , 3 ) Arkansas
59 : 34.9 / 28.1 ( 4 , 2 ) Vanderbilt
60 : 34.6 / 31.8 ( 3 , 2 ) Virginia
61 : 33.9 / 32.3 ( 3 , 2 ) Iowa St
62 : 33.7 / 33.3 ( 2 , 3 ) Purdue
63 : 33.1 / 26.4 ( 3 , 2 ) Oklahoma St
64 : 32.6 / 26.3 ( 3 , 2 ) Bowling Green
65 : 31.5 / 27.3 ( 2 , 2 ) Louisiana Tech
66 : 30.6 / 26.7 ( 2 , 3 ) Marshall
67 : 30.3 / 28.7 ( 3 , 3 ) Utah
68 : 30.2 / 27.1 ( 2 , 2 ) Stanford
69 : 30.2 / 31.4 ( 3 , 2 ) Washington St
70 : 29.6 / 28.1 ( 2 , 2 ) Navy
71 : 29.5 / 25.2 ( 2 , 4 ) Wake Forest
72 : 28.9 / 24.8 ( 3 , 2 ) Alabama-Birmingham
73 : 27.9 / 22.6 ( 3 , 3 ) Central Michigan
74 : 27.5 / 26.1 ( 3 , 2 ) Rutgers
75 : 27.4 / 23.2 ( 3 , 3 ) New Mexico
76 : 27.3 / 22.2 ( 2 , 3 ) Mississippi
77 : 27.1 / 23.8 ( 2 , 3 ) Memphis
78 : 26.8 / 28.2 ( 2 , 4 ) Pittsburgh
79 : 26.5 / 21.3 ( 2 , 4 ) Mississippi St
80 : 26.4 / 27.5 ( 2 , 3 ) Northern Illinois
81 : 26.3 / 25.0 ( 2 , 3 ) Brigham Young
82 : 26.1 / 27.7 ( 2 , 3 ) Miami OH
83 : 25.0 / 26.7 ( 2 , 4 ) Air Force
84 : 23.8 / 19.9 ( 2 , 3 ) East Carolina
85 : 23.7 / 25.1 ( 1 , 4 ) Syracuse
86 : 23.4 / 18.0 ( 3 , 2 ) Akron
87 : 23.1 / 20.8 ( 2 , 3 ) Arkansas St
88 : 23.1 / 26.5 ( 1 , 4 ) Washington
89 : 23.1 / 25.6 ( 2 , 4 ) San Diego St
90 : 23.0 / 15.9 ( 2 , 3 ) Ohio U.
91 : 22.8 / 20.1 ( 2 , 2 ) Tulane
92 : 22.7 / 18.1 ( 2 , 4 ) SMU
93 : 22.3 / 17.8 ( 1 , 4 ) Kentucky
94 : 21.9 / 17.1 ( 3 , 3 ) Eastern Michigan
95 : 21.8 / 16.3 ( 1 , 3 ) Middle Tennessee St
96 : 21.0 / 23.9 ( 1 , 4 ) Arizona
97 : 20.9 / 12.9 ( 2 , 3 ) Troy
98 : 20.2 / 17.4 ( 2 , 4 ) Illinois
99 : 20.0 / 16.9 ( 3 , 2 ) Nevada
100 : 20.0 / 14.3 ( 2 , 3 ) Cincinnati
101 : 19.1 / 16.9 ( 1 , 4 ) Hawai`i
102 : 16.3 / 14.2 ( 1 , 5 ) Duke
103 : 15.4 / 7.0 ( 2 , 4 ) Louisiana-Monroe
104 : 15.1 / 8.0 ( 1 , 4 ) Ball St
105 : 14.2 / 8.9 ( 3 , 3 ) Western Michigan
106 : 14.2 / 10.5 ( 1 , 4 ) Louisiana-Lafayette
107 : 13.5 / 4.9 ( 1 , 3 ) North Texas
108 : 12.9 / 9.1 ( 0 , 4 ) Rice
109 : 11.5 / 10.0 ( 2 , 4 ) UNLV
110 : 11.5 / 11.3 ( 0 , 5 ) Army
111 : 10.3 / 7.6 ( 0 , 6 ) New Mexico St
112 : 9.6 / 6.2 ( 2 , 2 ) Utah St
113 : 6.7 / 7.0 ( 1 , 4 ) Kent St
114 : 5.7 / 0.1 ( 0 , 6 ) Temple
115 : 2.6 / 3.0 ( 1 , 4 ) San Jose St
116 : 0.7 / 0.1 ( 1 , 5 ) Idaho
117 : 0.0 / 0.0 ( 0 , 5 ) Buffalo
 
Originally posted by: tfinch2
How your rankings can be so fubared
Click to expand...
What are your predictions of all of next week's scores?

Maybe, the problem stems from this. Damn I spend an hour trying to diagnose the problem yesterday and eventually just threw out one of the games. I guess they both should count.
 
Originally posted by: thesurge
switch up FSU and Miami (even though UM is my favorite team), and its all good.
Click to expand...
Do I have to do that with all teams that lost a game? Or just this one? What do I do when FSU loses? Switch FSU with whatever team beats them? Does Miami have to remain below FSU at all times? What if Miami is 10-1 and FSU is 5-6, does Miami have to stay below FSU?

Team X beats Y. Team Y beats Z. Team Z beats X. Do they all have to be below the others? How is that accomplished?
 
Originally posted by: TStep
Interesting. Are you willing to enlighten us on the ranking formula?
Click to expand...
Sure. I just wrote a PM a week or two ago describing it. Here it is:


It was a fun activity to write the code and tweak the formulas. So I highly recommend it. I always looked at the code in four parts.
1) Data gathering. Collecting the team names, scores, home field, etc. To use more data than that costs a lot of money as no place has it in convenient form for free. At least, no place that I found.
2) Team ranking. More on that later.
3) Predictions of next week's scores.
4) Prediction checking/optimizing.

1) Data gathering. This step is pretty straightforward. I'm sure anyone with some coding experience can do it easilly. There is a lot of error checking to incorporate, depending on your source. Sometimes one site doesn't update all the scores (Division II and below are a bitch sometimes). So you might need multiple sources. And of course each source names the teams differently. Is it Texas State, Texas St., or Texas St? Heck, even sometimes I've seen one website change spellings mid-year. The easiest method is just to create a big look-up table of all known name spellings and give a number to each team. Also include the division and maybe even conference (if you want to rate conferences against each other).

One lesson I learned is that Division II and below games have no impact on Division IA team rankings. So to simplify you can ignore any game that included those teams. Division IAA games do impact the IA rankings a bit. So I wouldn't drop them. Specifically, what I do is rate all IA and IAA teams. Then I rate all II/III teams that have played 2+ games against higher division teams.

2) Team Ranking. This is the fun part. You can play with anything you want. My formula basically looks like this:

[*]R = A*W + (B-1)*L + sum(C(SD)*SD) + sum(D(R)*R)

where

[*]R is the rating for a team.
[*]A is a weighting factor for winning.
[*]W is the number of wins that the team had.
[*]B is a weighting factor for losing.
[*]L is the number of losses that the team had.
[*]C(SD) is a weighting factor for the score difference. I like to make this a function of the score difference. Use any formula you like for this function. This is where you can be quite creative. Or just use a constant. If score is meaningless to you, then C(SD) = 0.
[*]SD is the score difference in each game. If Ohio St. beat Iowa by 31 to 6, then for Ohio St. use a SD of 31-6 = 25, for Iowa use 6-31 = -25.
[*]D(R) is a weighting factor for the schedule difficulty. For simplicity, start with D(R) = 1.
[*]R is the rating for the teams you played.

The B-1 is a fudge factor to please Anandtech. They wanted me to weight losses more than I was. So I put in the -1 part. I get better predictions by just using B*L, but then it ranks a lot of 1 loss teams above 0 loss teams for example. People get pissed. I fudged for Anandtech and people were happier with the results. I guess all that matters really is that you have 0 or 1 loss in the season. Nothing else is important since we have a 1 game playoff for the championship.

There are many ways to proceed to solve this. You can guess all the team's values of R, then substitute and repeat until convergence if you like. Note: this does not always converge due to rounding, so you have to take some care in solving. Or you can do Gauss elimination, LU decomposition, etc.

Clearly there are ways to expand my simple formula. Add in factors for home team, yards gained by rushing or passing, defense statistics, etc. However, I haven't yet found a reliable source for this data without big $$$.

There are issues of how often to update. Do you update this calculation with each game? Does a game on Thursday affect your Saturday predictions? Oddly in my experience, using that Thursday game has WORSENED the score predictions. So I do it week by week. But of course, you then have some teams that play twice in a week (Monday and Saturday are common two game weeks especially when hurricanes mess up the schedule). So you need some more error checking.

3) Prediction of next weeks scores. Back to the data gathering. You also need to read in future games, who is the home team etc. My prediction formula is simple.

Score prediction = R(i) - R(j) + H. I subtract the rating for each team to get the score prediction, then add in a weighting factor H to the home team. Note: you can also put home team advantage in the team ratings themselves, but I couldn't morally justify this. Specifially harming teams that play more home games seemed wrong. Who knows, maybe I could do much better predictions if I did this though.

Note: H~=3 is usually a very good starting point. Home teams on average do 3 points better than predictions. If you wanted to be really good let H vary by team. Some teams (Hawai'i and Nebraska for example) tend to to about 7 points better at home and 7 points worse when away. I never went into that much detail again for moral reasons. Can you really harm a team for having great fans? Or maybe, they poison the other team's water supply. 🙂

4) Prediction checking. This is where I think I differ from a lot of other programs. Each week I compare my predictions to what really happened. Then I vary all the weighting factors (A, B, C, D, H) to get the optimal prediction for that week. Those new values of A, B, C, D, and H are used for the rating that week and for next week's predictions. The variable factors let me capture how teams are doing at the moment. Suppose at the end of the season many star players are injured and scores are lower than they were in the beginning of the year. Then my predictions for next week include a lower weighting for scores to reflect this. Note: I drop the -1 portion in the (B-1)*L part of the formula when optimizing.

Note: these optimization problems are quite difficult (Newton's method or similar). The local minimum is not well defined. There could be a wide range of values for a certain constant that have no measurable effect on the predictions. Thus I try to find the range which has no major effect, then I use the value of the constant within that range that was closest to last week's value. This tends to smooth out week to week variations.

I personally do the optimization twice. Once to get the most accurate prediction of this week's winners. Then again to find the most accurate prediction of this week's scores. The results are subtly different. If one team is predicted to win but have lower points than another, you know it'll be a fun game to watch.
 
How would North Carolina State move down? They should have been way down already after losing to UNC, but after beating Georgia Tech you'd think they would move up?


Oh well, I'm pretty happy with #28 🙂
 
Originally posted by: Skiguy411
How would North Carolina State move down? They should have been way down already after losing to UNC, but after beating Georgia Tech you'd think they would move up?
Click to expand...
With 4 or fewer games played, the ratings just aren't that accurate. Teams move around quite a bit the first few weeks. Looking at things, the biggest factor could be NC's blowout loss (NC dropped 19 positions). When NC got pulled down, it probably pulled NC St. down with it. Oh, and I did have all pre-Oct 8th games included in the last poll. That included the mid-week Georgia Tech win.

 
Missed the preview thread. Pretty interesting. My teams going to need a little more offense to stay at No. 2 though.
 
Originally posted by: dullard
Originally posted by: tfinch2
How your rankings can be so fubared
Click to expand...
What are your predictions of all of next week's scores?

Maybe, the problem stems from this. Damn I spend an hour trying to diagnose the problem yesterday and eventually just threw out one of the games. I guess they both should count.
Click to expand...

WTF?!?!

Matt Hill must have never played football in his life, unless he was a kicker.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top