Rating teams based on All-time poll rankings

[Diego-Bucks]

My apologies, I saw this

NC (#1) in either poll = 100 points

and assumed that he was rewarding the National Title an additional 100 points. But looking back at the full post I was wrong.

College football rankings from any given year are based on individual voters perception of the relative strength of the teams. Trying to compare teams strength between years would be insane.

Isn't the concept of this analysis meant to serve to compare teams' strengths between years, in fact over the majority of college football history?

I think your suggestions take away from the simplicity of the ranking system and add other problems.

Certainly mine add some problems, however my suggestions are to better BB's model of analysis. Myself, I would have made different, and I feel simpler model. I'm working on the database right now and feel that I should have something punched up in a day or two, depending on my work schedule. I'll see how my ideas shape up but I'm planning on using the AP poll's top-10 from 1936 onwards and the coaches poll from 1950 onwards, rewarding teams 1-10 points on final rankings. I'll probably do combo score, but I felt that if in one poll Ohio State were #10 and in the other they were outside the top 10, then at least they'd garner a point for their appearance. I'm going to look at the top-10 simply due to the AP did top-10 only for a decade or so before reverting back to top-25 and top-20 polls (ideally, we'd have all 120 teams ranked through the years but... oh well). I'm going to try and log all the losses and ties per team and possibly (possibly not) penalize each team per loss. I feel that an undefeated champ is more 'prestigious' than a defeated one and will try to find a way to fully reward that team.

I'm not entirely certain on the methods yet, but I felt that if I were critiquing BB's with suggestions then I should be courageous enough to offer up my own flawed model. One thing I really want to see are certain teams relative rankings as they've changed over the last 75 years. The fall of Navy, Army, Minnesota and the likes vs the rise of the Florida programs and what not.

 

[BB73]

My apologies, I saw this and assumed that he was rewarding the National Title an additional 100 points. But looking back at the full post I was wrong.

Isn't the concept of this analysis meant to serve to compare teams' strengths between years, in fact over the majority of college football history?

Certainly mine add some problems, however my suggestions are to better BB's model of analysis. Myself, I would have made different, and I feel simpler model. I'm working on the database right now and feel that I should have something punched up in a day or two, depending on my work schedule. I'll see how my ideas shape up but I'm planning on using the AP poll's top-10 from 1936 onwards and the coaches poll from 1950 onwards, rewarding teams 1-10 points on final rankings. I'll probably do combo score, but I felt that if in one poll Ohio State were #10 and in the other they were outside the top 10, then at least they'd garner a point for their appearance. I'm going to look at the top-10 simply due to the AP did top-10 only for a decade or so before reverting back to top-25 and top-20 polls (ideally, we'd have all 120 teams ranked through the years but... oh well). I'm going to try and log all the losses and ties per team and possibly (possibly not) penalize each team per loss. I feel that an undefeated champ is more 'prestigious' than a defeated one and will try to find a way to fully reward that team.

I'm not entirely certain on the methods yet, but I felt that if I were critiquing BB's with suggestions then I should be courageous enough to offer up my own flawed model. One thing I really want to see are certain teams relative rankings as they've changed over the last 75 years. The fall of Navy, Army, Minnesota and the likes vs the rise of the Florida programs and what not.

Just giving 1-10 points for the top-10 teams would give a #5 team 6 points and the #10 team only 1 point.

That's just an example, but would you really want to assign 6 times as many points to a #5 team as the #10 team that year? And assign the same number of points to the #11 team as a team that had a losing record?

Those are the types of questions that I asked myself before assigning the point values as I did. They may be arbitrary, but it was done with some thought about the relative values of each spot in the final poll.

 

[Diego-Bucks]

Just giving 1-10 points for the top-10 teams would give a #5 team 6 points and the #10 team only 1 point.

That's just an example, but would you really want to assign 6 times as many points to a #5 team as the #10 team that year? And assign the same number of points to the #11 team as a team that had a losing record?

Those are the types of questions that I asked myself before assigning the point values as I did. They may be arbitrary, but it was done with some thought about the relative values of each spot in the final poll.

A valid point, but the quantitative difference, as opposed to percent difference, isn't that large between 1 and 6 so over 75 years, that would average out.

I suppose an interesting, albeit subjective question, would be is it better to have 10 seasons ranked at #10, or 1 season as the #1 team in the nation?

The biggest reason I don't want to include teams beyond the top-10 would be the lack of data from the AP poll when they went to only the top-10 from... I think 1962-70 or something like that. I want to restrict my own data-set to the poll that is the most restricted itself. That being said, I don't think its all that crazy to only look at the top-10 since those teams are going to be the ones most contending for a title during that season. The top-10 teams distinguished themselves from the rest of the competition that arguably the other ranked teams just didn't. So I suppose that Those would be equivalent with the method I proposed, and I don't think that its a bad thought, ya know? And of course, if I have a team with sustained excellence, for example the Florida State years in the 90s to 00s, then they'd be rewarded immensely. I have a few things to ponder...

Any reason to subtract points for the losing seasons though? Aren't those teams already punished for the lack of accumulated points that season compared with their peers?

It may be that as I get my database set-up and my methods consistent that your model becomes that much of a better indicator than what I've nit-picked. I'll see what becomes of this.

 

[BB73]

A valid point, but the quantitative difference, as opposed to percent difference, isn't that large between 1 and 6 so over 75 years, that would average out.

I suppose an interesting, albeit subjective question, would be is it better to have 10 seasons ranked at #10, or 1 season as the #1 team in the nation?

Those are the subjective decisions that have to be made. My scoring system has a #1 season equal to a pair of #4 years, 5 #15 years, and 10 #20 years. That's not right or wrong, it's just what I thought was reasonable.

The biggest reason I don't want to include teams beyond the top-10 would be the lack of data from the AP poll when they went to only the top-10 from... I think 1962-70 or something like that.

Yep, from '62-'67 (inclusive), they only had 10 in the AP poll.

I want to restrict my own data-set to the poll that is the most restricted itself. That being said, I don't think its all that crazy to only look at the top-10 since those teams are going to be the ones most contending for a title during that season. The top-10 teams distinguished themselves from the rest of the competition that arguably the other ranked teams just didn't. So I suppose that Those would be equivalent with the method I proposed, and I don't think that its a bad thought, ya know? And of course, if I have a team with sustained excellence, for example the Florida State years in the 90s to 00s, then they'd be rewarded immensely. I have a few things to ponder...

Any reason to subtract points for the losing seasons though? Aren't those teams already punished for the lack of accumulated points that season compared with their peers?

I was rewarding consistency as well as those contending for a title. And penalizing teams for bad years compared to mediocre years, rather than treating all unranked teams the same. So over two years, a team that was #20 one year and went 5-7 the next year gets zero points, the same as a team that went 7-5 and unranked both years. Another subjective decision, but that was the rationale behind it.

I also decided ahead of time that I wouldn't change my model after the results came back, since that could be interpreted as changing the model to benefit tOSU's position. Thus, I gave it a fair amount of thought before crunching the numbers.

 

[DaddyBigBucks]

Not really a part of this discussion, so I will leave the GPA as my only commentary on it.

Those are the subjective decisions that have to be made. My scoring system has a #1 season equal to a pair of #4 years, 5 #15 years, and 10 #20 years. That's not right or wrong, it's just what I thought was reasonable.



Yep, from '62-'67 (inclusive), they only had 10 in the AP poll.



I was rewarding consistency as well as those contending for a title. And penalizing teams for bad years compared to mediocre years, rather than treating all unranked teams the same. So over two years, a team that was #20 one year and went 5-7 the next year gets zero points, the same as a team that went 7-5 and unranked both years. Another subjective decision, but that was the rationale behind it.

I also decided ahead of time that I wouldn't change my model after the results came back, since that could be interpreted as changing the model to benefit tOSU's position. Thus, I gave it a fair amount of thought before crunching the numbers.

 

[zincfinger]

Not really a part of this discussion, so I will leave the GPA as my only commentary on it.

Fine, but if you really wanted to contribute, you would apply full-on DSA to every college football team for every season going back to the inception of the sport. You've got the spreadsheet ready to go, so it really shouldn't take more than a couple years.

 

[Buckeyeskickbuttocks]

Oh dear.... you've challenged DBkB?

 

[DaddyBigBucks]

Fine, but if you really wanted to contribute, you would apply full-on DSA to every college football team for every season going back to the inception of the sport. You've got the spreadsheet ready to go, so it really shouldn't take more than a couple years.

Oh dear.... you've challenged DBkB?

I'll get to it just as soon as the pesky last vestiges of my will-to-live have been beaten out of me by my soon-to-be-EX wife. At that point, suicide by spreadsheet won't seem so bad.

But then everyone will have one less BkB to kick around.

 

[zincfinger]

Oh dear.... you've challenged DBkB?

Well sure, if you get the whole bkb alter ego network in on the act, you can work 24/7 and it shouldn't take more than a couple months.

 

[DaddyBigBucks]

Well sure, if you get the whole bkb alter ego network in on the act, you can work 24/7 and it shouldn't take more than a couple months.

So how much of this are you in for? How about you enter data for the years for which Alabama claims a National Championship?

I'll be glad to crunch the numbers as soon as you're done.

 

[zincfinger]

So how much of this are you in for? How about you enter data for the years for which Alabama claims a National Championship?

I'll be glad to crunch the numbers as soon as you're done.

I haven't been "invited" to join the bkborg as of yet. So all of you are on your own.

 

[Jaxbuck]

I have a couple of suggestions to make it more reflective of reality and it won't take a lot of work;

-Subtract an additional 100 points for any AP Top 10 team that loses to a I-AA/FCS team. Subtract another 100 if they lost to them at home.

-Also, a -10 for a losing record is a start but it doesn't distinguish between a 5-6 record and say a 3-9 record. It should be an additional 10 point penalty for every game under .500.

 

[kn1f3party]

I have a couple of suggestions to make it more reflective of reality and it won't take a lot of work;

-Subtract an additional 100 points for any AP Top 10 team that loses to a I-AA/FCS team. Subtract another 100 if they lost to them at home.

-Also, a -10 for a losing record is a start but it doesn't distinguish between a 5-6 record and say a 3-9 record. It should be an additional 10 point penalty for every game under .500.

+1

 

[Diego-Bucks]

Eh... doesn't look like I'll get the time to start doing what I wanted for a while longer. I just got back from a month of field-work in the desert and now I gotta prepare for a new job opportunity elsewhere. I'd really love to re-approach this soon though, statistics and data-analysis are something I really love to do, especially when its about my favorite sport: college football

Didn't mean to fluster anybodies feathers around here on this discussion!

 

[BB73]

This has been updated after the final polls for the 2010 season.

OK, here's how this was calculated. I took each team's ranking in every year-end poll since the AP started in 1936. Once 2 polls were involved, I always used the higher ranking. Sliding scale points were awarded for every year that a team ended up ranked, and 10 points were deducted for each losing season. The scale was determined before seeing where teams ended up.

For each year since 1936, a team earns points based on these criteria:

NC (#1) in either poll = 100 points
02 -> 05 = 65, 55, 50, 45 points, respectively
06 -> 10 = 40, 37, 34, 32, 30 points
11 -> 20 = 28, 26, 24, 22, 20, 18, 16, 14, 12, 10 points
21 -> 25 = 08, 06, 04, 03, 02 points
non-ranked, but .500 or above = 0 points
losing record for the year = minus 10 points

Here are the all-time totals, updated after the 2010 final polls:

01. 2393 - Oklahoma
02. 2281 - Ohio State (passed ND this year)
03. 2255 - Notre Dame
04. 2180 - Alabama
05. 1960 - Michigan
06. 1952 - USC
07. 1779 - Texas (-10 for losing record in 2010)
08. 1703 - Nebraska
09. 1513 - Tennessee (-10 for losing record in 2010)
10. 1488 - Penn State
11. 1183 - Miami
12. 1130 - Florida State
13. 1094 - LSU
14. 1071 - Florida
15. 1044 - Auburn
16. 1013 - Georgia (-10 for losing season in 2010)
17. 0869 - UCLA (-10 for losing season in 2010)
18. 0715 - Arkansas
19. 0683 - Michigan State
20. 0569 - Washington
21. 0545 - Georgia Tech (-10 for losing season in 2010)
22. 0471 - Ole Miss (-10 for losing season in 2010)
23. 0429 - Pittsburgh
24. 0426 - Texas A&M
25. 0397 - Clemson (-10 for losing season in 2010)

Other schools: Colorado(383), Minnesota(369), Wisconsin(337), Syracuse(327), Army(307), Iowa(285), Purdue(115), Illinois (13)

Since 2007, I have created separate ratings by adding National Championship credit for those earned prior to 1936, on a sliding scale based on 12-year periods.

1869-1899 - 10 points for each MNC (no top teams here, almost all Ivy League)
1900-1911 - 25 points for each MNC
1912-1923 - 50 points for each MNC
1924-1935 - 75 points for each MNC

Here are the all-time totals, updated with the pre-1936 MNC points:

01. 2480 - Notre Dame (MNCs in '24, '29, '30 = 225)
02. 2455 - Alabama (MNCs in '25, '26, '34, 2/3 for '30 = 275)
03. 2393 - Oklahoma
04. 2281 - Ohio State
05. 2152 - USC (MNCs in '31, '32, 2/3 for '28 = 200)
06. 2135 - Michigan (MNCs in '01, '02, '23, '33 = 175)
07. 1779 - Texas
08. 1703 - Nebraska
09. 1513 - Tennessee
10. 1488 - Penn State
11. 1183 - Miami
12. 1130 - Florida State
13. 1094 - LSU
14. 1071 - Florida
15. 1044 - Auburn
16. 1013 - Georgia
17. 0869 - UCLA
18. 0715 - Arkansas
19. 0683 - Michigan State
20. 0670 - Georgia Tech (MNCs in '17, '28 = 125)
21. 0569 - Washington
22. 0554 - Pittsburgh (MNCs in '10, '16' '18 = 125)
23. 0519 - Minnesota (MNCs in '34, '35 = 150)
24. 0476 - Texas A&M (MNC in '19 = 50)
25. 0471 - Ole Miss
26. 0407 - Army (MNCs in '14, '16 = 100)
27. 0397 - Clemson
28. 0383 - Colorado
29. 0337 - Wisconsin
30. 0327 - Syracuse

Note - USC and Bama received 50 points, rather than 75, for disputed titles in '28 and '30, respectively

Note - Illinois, with MNCs in '14, '23, and '27 fails to make the top 30.