11-18-2004, 12:32 PM | #1 | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
College Starter
Join Date: Dec 2001
|
My Computer Ratings (NCAA and NFL)
It was my goal at first to devise computer ratings that accounted for more than just points, such as yards gained and allowed. However, that quickly changed when I decided that I would want to be able to apply my system to the high school ranks and publish an open-source ratings set for Texas high school football.
With that in mind, I decided to combine aspects of both the Massey and Colley formulas, as explained below: I've gone with a matrix system using the mathematical basis used by Colley's rankings as explained in this PDF file. This is basically a method to solve the system of equations created by attempting to solve for each team's rating based on each game played. However, Colley uses a simple game scoring system of 1 and -1. For a win you get 1 and a loss you get -1. He's 100% opposed to using margin of victory in any way, and is actually quite feisty about it. I don't agree and am including it because it clearly indicates team strength when Team A beats Team C by 40 points while Team B beats them by 3 points. So how to score games? There certainly should be a difference between winning by 7 points and winning by 21 points. But the bonus for winning by 54 instead of 40 shouldn't be quite so high. My game scoring method is based on the Game Outcome Function used by Massey and explained here: Quote:
That was my attempt to mimic Massey's GOF. Now, for my ratings I modified the function so that the minimum GOF for a winning team is 2/3 . While that's a popular number around here, I'm actually using it because I wanted a certain level of credit for winning the game. I'm actually using 3/4 for the NFL and the straight GOF above for high school. It's my opinion that the higher the level, the more confidence you can have that the winning team is the better team. I took the GOF and multiplied it by 2 and subtracted 1 to arrive at the game score for each team in each game. This results in an infinite blowout being a score of 1 for the winning team and a score of -1 for the losing team. Here are the results of this rating system for the season-to-date in college football:
Initial review shows that this system is heavily points dependent, as it should be based on the GOF used. A team that wins by a single point gets a game score of ~.333 while the losing team gets a game score of ~-.333 as compared to the 1 and -1 from a system that doesn't use points at all. Logically this can be taken to mean that, based on this game, this GOF is saying that we can be 66.7% sure that the winning team is better than the losing team whereas the no-points system states that, based only on this game, we are 100% sure that the winning team is a better team. Anyway, here's the NFL results for this year so far:
Texas high school results will be forthcoming. But that's a lot of data to enter. Thoughts?
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
11-18-2004, 12:44 PM | #2 | |
Head Coach
Join Date: Oct 2000
Location: North Carolina
|
Quote:
It passes the smell test in that the final rankings approximate how I think that the teams should be ordered in my subjective opinion. Personally, I don't agree with using margin of victory simply because we tell teams that winning is the only thing, and then rate them on how much they win by. But that's just a difference of opinion. Overall, I like the results. |
|
11-18-2004, 01:03 PM | #3 |
Hall Of Famer
Join Date: Jan 2001
Location: Decatur, GA
|
Not a bad ranking system at all. There are a few quibbles, such as albionmoonlight refered to, that margin of victory can be deceptive sometimes. Some teams will run up the score, while others won't and that does have an effect, no matter how slight.
__________________
"A prayer for the wild at heart, kept in cages" -Tennessee Williams |
11-18-2004, 01:44 PM | #4 | |
Coordinator
Join Date: Nov 2003
Location: The Great Northwest
|
Quote:
such as it starts at .333 and it for the first 10 points and works down from there or something. |
|
11-18-2004, 02:06 PM | #5 |
College Starter
Join Date: Dec 2001
|
The diminishing returns shown above is the same thing as a cap. And because it doesn't use direct margin of victory, the cap you speak of can't be implemented.
Calling a 63-42 win the same thing as a 41-0 win is crazy, IMO. (A 21 point MOV cap)
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings Last edited by Huckleberry : 11-18-2004 at 04:14 PM. |
11-21-2004, 03:54 PM | #6 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
College Starter
Join Date: Dec 2001
|
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
11-21-2004, 04:05 PM | #7 |
College Starter
Join Date: Oct 2000
Location: Norman, OK
|
What about the differnence between a 63-42 win vs. a 21-0 win. I didn't read all too in-depth about your margin system, but woudln't you think that a 21-0 win should get more than a 63-42 win?
|
11-21-2004, 05:20 PM | #9 |
College Starter
Join Date: Dec 2001
|
sooner333 -
In this system, the 21-0 gets more credit.
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
11-23-2004, 10:54 AM | #10 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
College Starter
Join Date: Dec 2001
|
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
11-23-2004, 11:05 AM | #11 | |
Torchbearer
Join Date: Sep 2001
Location: On Lake Harriet
|
The rankings seem to be about as spot on as any I've seen.
However, I'm curious about this statement from your first post... Quote:
I think I'm probably of the opposite opinion. My thought would be that the higher the level, the closer the two teams probably are in talent, the more impact coaching and game planning can have and the greater chance of an upset. It probably doesn't change your college ratings because they certainly lie in between the NFL and high school, no matter which way you go, but I'm wondering how much it skews the NFL ratings if you were to go closer to the straight GOF. |
|
11-23-2004, 11:43 AM | #12 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
College Starter
Join Date: Dec 2001
|
First my logic. Professional athletes are more likely to perform at their ability level. They have more practice time, more experience, etc. High school kids, and to a lesser extent college players, will have more variation in their performance, therefore our confidence that they played at their ability level is lower for each game. Anyway, here are the rankings using the straight GOF (~.53 instead of .75 for a close win):
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
11-23-2004, 11:50 AM | #13 |
Torchbearer
Join Date: Sep 2001
Location: On Lake Harriet
|
Interesting. Thanks. I understand your logic (though still not sure if I agree with it), but I do think the ratings look better with the modified GOF.
|
11-28-2004, 08:10 AM | #14 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
College Starter
Join Date: Dec 2001
|
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
12-04-2004, 02:05 AM | #15 |
College Starter
Join Date: Dec 2001
|
I'm not sure if anyone still cares about this, but I've added automation to the process. Now all I have to do is type in the teams and scores for each game and let the program I wrote do the rest.
Here's a link to the output page from the program for college ball. In case you weren't aware, it's pretty clearly based on David Rothman's program. I had to change it up, but the code includes references to all source material. http://www.hornfans.com/recruiting/h...rry/D1RATE.TXT
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
12-04-2004, 10:28 AM | #16 |
College Starter
Join Date: Oct 2000
Location: Edmond, OK
|
I haven't said so yet, but I definitely care about this. Nice job, Huck.
|
12-04-2004, 11:13 AM | #17 | ||
Pro Starter
Join Date: Oct 2000
Location: Muskogee, OK USA
|
Pretty good for a Longhorns fan.
__________________
Quote:
Quote:
|
||
12-04-2004, 12:02 PM | #18 |
n00b
Join Date: Sep 2002
|
could you send me a copy of you program. I have dabled in trying different ranking systems and have even made an elaborate spreadsheet in excel. Can your matrix be input into excel? I have read massey's articles but, my pathetic math knowledge limits me in many ways.
|
12-04-2004, 12:02 PM | #19 |
n00b
Join Date: Sep 2002
|
could you send me a copy of your program. I have dabled in trying different ranking systems and have even made an elaborate spreadsheet in excel. Can your matrix be input into excel? I have read massey's articles but, my pathetic math knowledge limits me in many ways.
|
12-04-2004, 01:15 PM | #20 |
College Starter
Join Date: Dec 2001
|
Actually, my rankings set started out in Excel because it's matrix based. However, my Excel (Office 97) has issues inverting matrices larger than 52 x 52, so I had to get Quattro Pro to do that part.
The matrix I have is Colley's Matrix but the game scoring system is different. PM me with your e-mail address and I can send you what I have.
__________________
The one thing all your failed relationships have in common is you. The Barking Carnival (Longhorn-centered sports blog) College Football Adjusted Stats and Ratings |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
|
|