BYU and Media Bias: Looking at year end polls - cougarguard.com — unofficial BYU Cougars / LDS sports, football, basketball forum and message board

pelagius · 12-08-2006, 07:07 PM

It seems like one of the things I hear quite frequently is that there is a media bias against BYU. That at least on a national level BYU is often ignored. Certainly, it seems to have intensified this year because of the TV contract. However, my sense is that many cougar fans believe BYU is often slighted. Do you think this is really true and how does it manifest itself? My first thought was that if there is a media bias against BYU that it should show up in the AP rankings. So I looked at year end AP rankings from 1976-2006 and compared them to year end Massey computer rankings. I defined bias as the following:

bias = Massey Computer Rank - AP Rank

Certainly some may argue that the computer rankings are biased against BYU as well. There is truth to this claim in the current BCS setup where the computer rankings can only use wins and loses and must throw out MOV. Thus BYU can't make up for poor SOS by winning by a large MOV. However, I am using a MOV a victory based ranking systems so it is not obvious that it is biased against BYU.

Here is what I found for years in which BYU was ranked in the year end AP poll:

Average Bias: -1.54
Median Bias: -1

Thus on average there was very little bias and in fact the average bias is actually negative. Meaning on average the AP has ranked BYU higher than the Massey. Furthermore, in every year where BYU was ranked in the top 20/25 by Massey they were also in the end of year AP poll.

Also, as a quick comparison I did the same thing for Ohio State:

Average Bias: -1.38
Median Bias: -1

BYU and Ohio State are almost identical. So do you think that griping about media bias is just one those "shared communal experiences" of fans everywhere or is there really a case for anti BYU bias? If there is how does it manifest itself?

Indy Coug · 12-08-2006, 07:14 PM

Is Massey's formula completely secret, or is there a detailed explanation of the methodology involved?

Archaea · 12-08-2006, 07:20 PM

I actually don't believe the media treats BYU unfairly in football. We seem to receive fair treatment in comparison to other nonBCS schools. Now if we were near the top, we might be questioned but otherwise we are fairly treated.

pelagius · 12-08-2006, 07:23 PM

Quote:

Originally Posted by Indy Coug

Is Massey's formula completely secret, or is there a detailed explanation of the methodology involved?

It is quasi-secret/quasi-public. He does describe most of the essential feature. It is pretty similar to the poll modeling I described the other day:

http://cougarguard.com/forum/showthread.php?t=5335

Differences: He translates the MOV into some sort of empirical probability so that all MOVs are translated into a number between 0 and 1. He does talk about the essential features of the function but doesn't disclose it in detail. Still, it doesn't appear that different than taking like the square root of the margin of victory. He also uses maximum likelihood to estimate but that shouldn't be that different than using something like a logistic regression once you translate your dependent variable to the 0-1 range

jay santos · 12-08-2006, 07:37 PM

A theory I have on this is that the human polls are SOS light, meaning they don't take SOS into consideration as much as they should (assuming avg computer polls get it right).

I've never tested this, but it would be an interesting analysis.

I think what you might find is that after you take out the SOS effect in the polls, BYU is slightly biased against.

For example, look at Boise St and Wisconsin in this year's polls.

Avg computer ranking

Boise 7
Wisconsin 8

AP

Boise 9
Wisconsin 6

My theory is that the SOS light nature of the AP poll would rank Boise 5, Wisconsin 6, but then the anti-non BCS/anti west nature of the AP poll brings Boise back down to 9.

pelagius · 12-08-2006, 08:32 PM

Quote:

Originally Posted by jay santos

For example, look at Boise St and Wisconsin in this year's polls.

Avg computer ranking

Boise 7
Wisconsin 8

AP

Boise 9
Wisconsin 6

My theory is that the SOS light nature of the AP poll would rank Boise 5, Wisconsin 6, but then the anti-non BCS/anti west nature of the AP poll brings Boise back down to 9.

I think Boise makes an interesting comparison to BYU. Obviously, the sample is small for Boise but here is what it looks like:

Code:

     AP   Massey
00   NR   24
02   15   9
03   16   14
04   12   12
06    9   6

The average bias for Boise St is 2.75, and they have never had an AP rank higher than their Massey ranking. Compare this to BYU were half the time their AP rank has been bigger than their Massey rank. One hypothesis that I have is this represents a time trend. Pollster in the early 80s were SOS light (as Jay puts it) but these days with the advent of the BCS, pollsters pay more attention to SOS than they used to.

jay santos · 12-08-2006, 09:16 PM

Some more data and a new theory.

Top 25 teams with strong SOS (top 25 by my model)

team, avg computer rating, AP, difference

FL 3 2 1
Mich 2 3 -1
USC 4 8 -4
Cal 15 20 -5
Tenn 13 17 -4
Ark 16 12 4
Ore St 21 24 -3
UCLA 23 35 -12

Top 25 teams with weak SOS (below top 50 by my model) BCS schools only

Ohio State 1 1 0
Wisconsin 8 6 2
VT 14 14 0
WF 20 15 5
BC 22 23 -1
TAM 25 21 4

Top 25ish non-BCS teams

Boise State 7 9 -2
BYU 19 19 0
Hawaii 31 28 3
TCU 26 25 1

Another data point: Utah was higher in the computers than in the AP in their great year.

New theory: AP poll is SOS light, including its affect on non-BCS schools, however, there is a glass ceiling effect around top 10 where non-BCS schools are punished in the AP poll.

I've done some simple analysis on AP vs computers before and as I recall, the AP is generally SOS light in the past as well.

jay santos · 12-08-2006, 09:20 PM

I have a lot of problems with human polls. My list includes (but is not limited to):

biased against teams in the West
too SOS light
biased against teams with losses late in the season compared to early
biased against non-BCS--maybe only as they approach the top 5

These are mostly just gut feelings not a result of careful analysis, although this little discussion today helps me cement a couple of them.

pelagius · 12-08-2006, 09:27 PM

Quote:

Originally Posted by jay santos

I have a lot of problems with human polls. My list includes (but is not limited to):

biased against teams in the West

Jay, I think that one is interesting. I think at some point I would like to do a careful test of that hypothesis.

12-08-2006, 07:07 PM	#1
pelagius Senior Member Join Date: Nov 2006 Posts: 1,431	BYU and Media Bias: Looking at year end polls It seems like one of the things I hear quite frequently is that there is a media bias against BYU. That at least on a national level BYU is often ignored. Certainly, it seems to have intensified this year because of the TV contract. However, my sense is that many cougar fans believe BYU is often slighted. Do you think this is really true and how does it manifest itself? My first thought was that if there is a media bias against BYU that it should show up in the AP rankings. So I looked at year end AP rankings from 1976-2006 and compared them to year end Massey computer rankings. I defined bias as the following: bias = Massey Computer Rank - AP Rank Certainly some may argue that the computer rankings are biased against BYU as well. There is truth to this claim in the current BCS setup where the computer rankings can only use wins and loses and must throw out MOV. Thus BYU can't make up for poor SOS by winning by a large MOV. However, I am using a MOV a victory based ranking systems so it is not obvious that it is biased against BYU. Here is what I found for years in which BYU was ranked in the year end AP poll: Average Bias: -1.54 Median Bias: -1 Thus on average there was very little bias and in fact the average bias is actually negative. Meaning on average the AP has ranked BYU higher than the Massey. Furthermore, in every year where BYU was ranked in the top 20/25 by Massey they were also in the end of year AP poll. Also, as a quick comparison I did the same thing for Ohio State: Average Bias: -1.38 Median Bias: -1 BYU and Ohio State are almost identical. So do you think that griping about media bias is just one those "shared communal experiences" of fans everywhere or is there really a case for anti BYU bias? If there is how does it manifest itself? Last edited by pelagius; 12-08-2006 at 09:39 PM.

12-08-2006, 07:20 PM	#3
Archaea Assistant to the Regional Manager Join Date: Aug 2005 Location: The Orgasmatron Posts: 24,338	I actually don't believe the media treats BYU unfairly in football. We seem to receive fair treatment in comparison to other nonBCS schools. Now if we were near the top, we might be questioned but otherwise we are fairly treated. __________________ Ἓν οἶδα ὅτι οὐδὲν οἶδα

12-08-2006, 07:14 PM	#2
Indy Coug Senior Member Join Date: Jan 2006 Location: Between Iraq and a hard place Posts: 7,569	Is Massey's formula completely secret, or is there a detailed explanation of the methodology involved?

12-08-2006, 07:37 PM	#5
jay santos Senior Member Join Date: Jan 2006 Posts: 6,177	A theory I have on this is that the human polls are SOS light, meaning they don't take SOS into consideration as much as they should (assuming avg computer polls get it right). I've never tested this, but it would be an interesting analysis. I think what you might find is that after you take out the SOS effect in the polls, BYU is slightly biased against. For example, look at Boise St and Wisconsin in this year's polls. Avg computer ranking Boise 7 Wisconsin 8 AP Boise 9 Wisconsin 6 My theory is that the SOS light nature of the AP poll would rank Boise 5, Wisconsin 6, but then the anti-non BCS/anti west nature of the AP poll brings Boise back down to 9.

12-08-2006, 09:16 PM	#7
jay santos Senior Member Join Date: Jan 2006 Posts: 6,177	Some more data and a new theory. Top 25 teams with strong SOS (top 25 by my model) team, avg computer rating, AP, difference FL 3 2 1 Mich 2 3 -1 USC 4 8 -4 Cal 15 20 -5 Tenn 13 17 -4 Ark 16 12 4 Ore St 21 24 -3 UCLA 23 35 -12 Top 25 teams with weak SOS (below top 50 by my model) BCS schools only Ohio State 1 1 0 Wisconsin 8 6 2 VT 14 14 0 WF 20 15 5 BC 22 23 -1 TAM 25 21 4 Top 25ish non-BCS teams Boise State 7 9 -2 BYU 19 19 0 Hawaii 31 28 3 TCU 26 25 1 Another data point: Utah was higher in the computers than in the AP in their great year. New theory: AP poll is SOS light, including its affect on non-BCS schools, however, there is a glass ceiling effect around top 10 where non-BCS schools are punished in the AP poll. I've done some simple analysis on AP vs computers before and as I recall, the AP is generally SOS light in the past as well.

12-08-2006, 09:20 PM	#8
jay santos Senior Member Join Date: Jan 2006 Posts: 6,177	I have a lot of problems with human polls. My list includes (but is not limited to): biased against teams in the West too SOS light biased against teams with losses late in the season compared to early biased against non-BCS--maybe only as they approach the top 5 These are mostly just gut feelings not a result of careful analysis, although this little discussion today helps me cement a couple of them.