Combined BB and FB ratings: UM #1 in nation

Submitted by michelin on January 8th, 2013 at 2:39 PM

UM is #1 in the nation in a combined BCS-like ranking for BB and FB together.  To compute such a rank, I took the mean* of the computer and human poll ranks (using Sagarin and the coach’s poll for both BB and FB).  

COMBINED UPI AND SAGARIN RANKINGS

 

fb sagarin

bb sagarin

fb coachs

bb coachs

geo mean rank

Michigan

20

2

26

2

7

Florida

7

5

10

9

8

Notre Dame

5

14

3

16

8

Louisville

27

3

13

4

8

Kansas State

8

38

11

23

17

Oklahoma State

17

19

15

28

19

Cincinnati

36

18

22

20

23

Arizona State

28

70

33

36

39

 

*While there are many ways to analyze such results---and limited time has prevented me from looking any many of them-----I wanted measures that  are more stable and less sensitive to outliers.  So, I used the geometric rather than the arithmetic means.

Note also that, if we used the sagarin computer ratings only, we could rank more teams (ie those that do not receive votes in the coach’s poll).   In these computer-only polls,  UM is in a virtual tie with FLA for #1.  Ohio—even if we forget their NCAA probation and make them “eligible” for this ranking--still is only ranked at #8 overall.  They are even lower if we use only the most accurate Sagarin PREDICTOR rankings (not shown).  In fact, Ohio is actually behind UM not only in BB but also in FB according to the Sagarin PREDICTOR rankings!   To quiet skeptics and make the BB computer rankings more similar to the BB rankings, however, I have  used ELO as well as PREDICTOR rankings from Sagarin throughout.

 

COMPUTER ONLY RANKINGS

Florida

6

Michigan

6

Oregon

8

Notre Dame

8

Duke

9

Louisville

9

Alabama

9

Ohio State

14

Syracuse

16

Texas A&M

16

Kansas St

17

Oklahoma St

18

Arizona

20

Indiana

21

Baylor

22

Oklahoma

22

Stanford

23

Minnesota

23

Cincinnati

25

LSU

26

Missouri

27

Michigan St

27

Florida St

28

Mississippi

29

Wisconsin

30

Miami-Fl

32

Boise State

34

San DiegoSt

36

Pittsburgh

37

Texas

37

UCLA

37

Oregon St

40

North Dk St

41

NC State

41

Clemson

41

Utah State

42

Arizona St

44

Northwestern

45

Iowa State

48

North Car 

50

Iowa

59

Tennessee

62

LA tech   

63

Rutgers

65

Arkansas

68

Washington

70

California

71

Central Fl

72

Utah

72

Ohio

82

Purdue

92

http://usatoday30.usatoday.com/sports/sagarin.htm

http://espn.go.com/college-football/rankings

http://espn.go.com/mens-college-basketball/rankings

Comments

michelin

January 8th, 2013 at 3:01 PM ^

but---if we really want to measure the performance of real "student athletes,"  then it might be interesting to weight the athletic performance by a real "strength of shedule"---ie the strength of a student's academic schedule.  That would quickly eliminate a lot of visible athletic universities which are not providing real university educations for athletes--and whose athletes often do not belong in college to begin with

MGoBender

January 8th, 2013 at 4:13 PM ^

Let's not pretend like we're leaps and bounds better than schools.

A little better, sure. However, the idea that Michigan holds extrememly higher standards for their scholarship athletes is false. It was debunked by an admissions and/or athletic employee on these boards a bit ago.

Basically, he said that the University gives the football team a certain number of admits where the only standards for admission are the NCAA requirements (so, any school). After that number, then the football team can make the case for students that don't quite live up to the University's standards for scholarship athletes (which are still much lower than general admission standards).

michelin

January 8th, 2013 at 5:02 PM ^

but the pretense that we are watching real student athletes compete at many schools.  (eg one Ohio qb remarked this year that the need to take classes was a joke--that the Ohio athletes were there only to play FB)

To be sure, your point about UM is well taken.  Almost every school wishing to compete athletically is forced to compromise its standards, but I do believe that some schools make far bigger compromises than others. 

TruBluMich

January 8th, 2013 at 6:11 PM ^

This is a classic example of using math to come up with an equation that favors your point of view.  Both Florida and Notre Dame could use far easier equations to come up with a different result, that differ from yours.  Also by not using the AP poll you conviently left out Ohio State, which could only lead me to believe you wanted your equation to have Michigan at #1.  By no means discrediting your work, just saying that your equations and picking polls that favor your preferred outcome leave room for debate.

justingoblue

January 8th, 2013 at 6:22 PM ^

of cherry-picking is using Sagarin (although I have no idea if that favors M or not) and not other computer rankings. The better method might be using the BCS computer average and Kenpom/Sagarin average or something.

I think the Coaches Poll is fine to use, since he was looking for a "BCS like" ranking system, and that's the poll they like. I do agree something should have been done to include OSU in the analysis, though.

michelin

January 8th, 2013 at 7:44 PM ^

For that reason, I did use the coach's poll, which does not rank Ohio.  For the reader's interest, I also did show how Ohio fared in another bcs component: a computers only poll---which put Ohio at #8---far off the pace set by UM,FLA,ND,OR.  Frankly, I do not think that Ohio even deserved that mention.*  Whatever they achieved this year, they achieved essentially with players recruited by a liar who tried to conceal his players'illegal benefits.

I do agree that it would be interesting to see how other computer polls might provide different results.  Possibly UM would not fare as well, but possibly UM would fare even better.  I honestly do not know.  Unfortunately, my time and abilities are far too limited to summarize all the BCS computer polls.  Sorry.

justingoblue

January 9th, 2013 at 12:49 AM ^

especially since I wasn't trying to be critical of your work (and definitely not needed for me or any of your other readers in the first place). My post was trying to defend a suggestion of cherry picking, and I said I didn't think that's what you were trying to do, and brought up a couple things I thought were at least thinking about for improving this formula. I did enjoy your work and though it was interesting; if you care to check I was the first to upvote the diary.

Also, just as an explanation for the OSU stuff, I missed your line in the original diary in my desire to get to the charts.

ZooWolverine

January 8th, 2013 at 10:53 PM ^

It's a fun exercise, but using that data and getting a result that puts us above Florida and Notre Dame means the method doesn't pass a sanity-check in my opinion. Our football rankings are not outliers, they should have roughly equal weight as the basketball rankings, but clearly do not.

Looking at ND, the difference between 14/16 and 20/26, while not huge, is definitely bigger than the difference between 2/2 and 3/5, so they really ought to be above us.

michelin

January 9th, 2013 at 3:52 PM ^

(1) you seem to think of "sanity" as including only subtraction (to find differences) but most students in 3rd grade also learn to find differences though division.   So what is a greater difference between two pairs of numbers, like 1-3 and 3-6?  In the first pair, the additive difference (3-1=2 vs 6-3=3) is less.  But the multiplicative difference (ratio, or 3/1=3 vs 6/3=2)) is greater.  Siimilarly, by the latter method, UM's advantage over ND in BB is greater than its disadvantage in FB.   I really do not see this perspective as any less sane; unless one assumes that a third grade education leads to insanity .

(2) it is not true that the fb and bb ratings have unequal weights in my calculation of geometric means.  

(3) using geometric means make sense logically.  eg if more games were played, very high bb ratings would likely come down and very low fb ratings likely go up "on average".   So, it does not make sense to let the extremes take precedence.

(4) geometic means are used by many knowledgable people.  At Harvard, average course ratings are calculated using similiar measures (either geometric means or medians), which, in effect, place less weight on the extremes.

(5) Although I do think that geometric means are always better than arithmetic ones, it is true that different measures may give us slightly different results.  But UM will still rank much more highly than most people realize.  I wrote this diary to publicize that fact---since we often focus unduly on a single sport in evaluating teams eg saying how great the SEC is, or how great ALA is, without realizing that this applies only to FB.  The BB team of ALA is really quite mediocre.  LIkewise, partisans will tell us how great ND is because they played in the NC game and finished with a high rank in FB.  But how great do you really think they would look if they had to replay their entire season?   Was getting clobbered by ALA an anomaly or was their prior undefeated record the result of collosal luck?   If the latter, then we should not really place so much weight on their extremely high (and likely spurious) ranking.

ZooWolverine

January 9th, 2013 at 5:44 PM ^

(1) I don't mean sanity in a derogitory sense as in you can't be sane and come up with those results, the term sanity-checking is just refering to a process of stepping back and looking at results and saying "do these make sense" and I disagree that they do. I think you can make an argument that the difference between #1 and #3 is bigger than between #3 and #6 in football, but it seems like you chose the geometric mean for the opposite reason--in (3) you say you're trying to give less weight to extremes.

(2) I understand that you're not weighting them differently, but the geometric mean is taking 2 and (roughly) 23 and coming up with an average of 7--from a more classical view of average, that's giving the #2 ranking more weight. I guess you're arguing that a geometric mean is always better, and I should ditch my classical view, but I'm just not buying it.

(3) When there are four numbers, two of which are a totally different metric than the other two, I have a hard time talking about outliers. As I mentioned, I think your argument here is the opposite of your argument in (1)--in (1) you're saying that the difference between 1 and 3 could be bigger than the difference between 3 and 6; here I think you're saying the opposite, #1 maybe really isn't that good after all; there would be some regression to the mean if more games were played.

(4) I don't argue that geomtric means have their place, just that I'm not buying their use here.

(5) I'm in total agreement that Michigan has an outstanding combination of good football and excellent basketball this year, I'm just not willing to go so far as saying they have the best combination. 

michelin

January 9th, 2013 at 7:29 PM ^

the extremely low (#26) rank for UM in FB coaches poll is not allowed to unduly affect the results if one uses a geometic mean but it would do so if one uses an arithmetic mean.

I consider that rank an outlier because it is the lowest for UM among all the human polls as well as for the different computer measures cited above.  The good UM ranks in BB, both commputer and human, are not outliers in that they agree.

 

michelin

January 9th, 2013 at 6:57 PM ^

just compare the product of ranks for UM with that for ND (ie divide the two for each, equally weighted FB or BB measure)

ie Michigan/ ND ratio overall = (fb SAGARIN rating RATIO * BB SAGARIN rating RATIO

* BB COACH poll Ratio * fb COACH poll Ratio)
    

=(20/5)*(2/14)*(26/3)*(2/6)=4*(1/7)*8.7*(1/3)= (34.8)/42 <1

So the product of the ranks for UM is lower (better) than for ND

(about 15% better)

michelin

January 11th, 2013 at 7:32 PM ^

ALA: FB#1, BB#97.   Mid Tenn: FB #52, BB#48.

If we add then average the BB and FB ranks, the two teams are nearly tied (separated by only one place).*  Yet, the #1 team (ALA in FB) was light years ahead of the prior BCS #1.  It bludgeoned ND in the title game.

Accordingly, we obtain far more reasonable results using the geometric mean.

ie for the team with a national title: ALA Rank=10,

For the team that did not even get into a bowl game: MID TENN rank=50.

So your idea to look at extremes like this does support the use of geometric mean, IMO.

Inevitably: whenever we have such extreme examples, like ALA BB and FB, there is more uncertainty about any composite rank. I am sure that one could fabricate an example to make one or another average look bad due to this uncertainty.  But, when you have any #1 ranked team, like ALA,  there is no telling just how good they are.  No one is above them.   (also, there is no telling how bad Mid Tenn would be in a bowl, had they gone to one).   Thus, I believe that, if we were to consider all such cases,  the geometric mean would provide a greater number of plausible results.

 

*Arithmetic  mean Ala fb 1 bb 97  AVG RK 49

Middle tennessee fb 52 bb48  AVG RANK 50

BB  http://usatoday30.usatoday.com/sports/sagarin/bkt1213.htm

FB  UPI used for top 30, site below for unranked teams

http://www.covers.com/sports/power-ranking/ncaaf-power-ranking.aspx

jocular_jock

January 11th, 2013 at 10:07 AM ^

Just use the directors cup. We are second behind Standford right now.

1. Stan - 398

2. UM - 373

3. FSU - 358.5

4. ND - 353

5. UNC - 334

53. Tidenfreude - 100 pts

They all have equal wt, national champions of a given sport get 100 pts. Womens CC nat champ gets the name amount of points as Alabama.