Matthews Grid Ratings (a GPI Component)

[BigApp]

For those who don't know, it's similar in setup in that it lists all Division I football teams together, like Sagarin does:


San Diego one spot below Georgia Tech...:eek:
Massachusetts
Northern Iowa
Appalachian
Montana
North Dakota State
Illinois State
James Madison
Youngstown
Maine


http://homepages.cae.wisc.edu/~dwilson/rsfc/rate/matthews.txt

[89Hen]

Information from Herman Matthews site on his #1 team...

Team Name: San Diego Bulldogs
Home Town: Birmingham, AL
Conference: Pioneer League

:lmao: : retard : : retard : Reality 1 - Credibility 0

http://www.expertpicks.com/colteaminfomain.asp?ID=176&Submit=Go%21

[*****]

Did either of you email him to ask what was up? I just did so maybe you should too instead of just taking potshots.

[BigApp]

Nope, not taking potshots. Clicked on the link provided at the bottom of the GPI page, http://www.i-aa.org/article.asp?articleid=81685, scroll down to the bottom and you'll see it.

"Journalists" question things. I am questioning the legitimacy of the GPI, not being a cheerleader.

BTW, the link to the Self Ratings takes you to a 2005 poll...xcoffeex

[GOTOREROS]

Maybe something got messed up on his website. I doubt the guy thinks San Diego is from "Birmingham, AL".......just my opinion.

He has "Air Force" as:

Team Name: Air Force Falcons
Home Town: Maxwell AFB, AL
Conference: Mountain West

GOTOREROS

[JoshUCA]

He has Central Arkansas listed as I-A Indepedent!!xlolx

[*****]

... "Journalists" question things...I think you'll find the pedigree of the computer rankings the GPI uses this year to be exemplary. :nod: Notice which is ranked first here: http://www.masseyratings.com/cf/compare1aa.htm

[PantherRob82]

also very questionable, althouh UNI sounds good to me. :hurray:

[*****]

Matthews final in 2005:
1. Appalachian St
2. Texas St
3. N Iowa
4. Furman
5. New Hampshire
6. Illinois St
7. Youngstown St
8. Ga Southern
9. S Illinois
10. Richmond
11. Brown
12. Cal Poly
13. James Madison
14. Montana St
15. Massachusetts
16. Montana
17. Hofstra
18. Nicholls St
19. E Washington
20. Portland St
21. N Dakota St
22. UC Davis
23. E Illinois
24. Hampton
25. Grambling

[YoUDeeMan]

Nope, not taking potshots. Clicked on the link provided at the bottom of the GPI page, http://www.i-aa.org/article.asp?articleid=81685, scroll down to the bottom and you'll see it.

"Journalists" question things. I am questioning the legitimacy of the GPI, not being a cheerleader.
BTW, the link to the Self Ratings takes you to a 2005 poll...xcoffeex

BigApp - Sorry, but you can't question the legitimacy of the GPI. Only Delaware fans can do that. xlolx xlolx

[Tod]

Maybe something got messed up on his website. I doubt the guy thinks San Diego is from "Birmingham, AL".......just my opinion.

He has "Air Force" as:

Team Name: Air Force Falcons
Home Town: Maxwell AFB, AL
Conference: Mountain West

GOTOREROS

Just FYI (Because of course Maxwell AFB is wrong), but there is a reason for that screw-up. Maxwell is the home of the Air Education and Training Command, and the Air Force Academy falls under it.

Again, just FYI...xcoffeex

[89Hen]

I doubt the guy thinks San Diego is from "Birmingham, AL".......just my opinion.
No, he obviously has them morphed with the Samford Bulldogs of the OVC who are located in Birmingham. I just hate to see mistakes like this on something that promotes accurate rankings. I am an admitted computer basher, but IMO for good reason. :nod:

[Lehigh Football Nation]

I am no statistician, and overall I am a fan of the GPI, but what I see with this index troubles me. I really simply cannot understand what possible formula that one can use to rate San Diego over TCU (I-A).

TCU:
* beat #24 ranked Texas Tech at home
* beat UC-Davis at home
* are 4-2, with their only losses coming on the road to Utah (4-4) and at home to BYU (5-2)

I mean, how is this possible??? How can they be ranked behind a team who has only beaten Yale? Do they use margin of victory, no matter how lowly the opponent?

I'm no expert, but I think beating the #24 team in the nation is deserving of being rated above San Diego, don't you?

Isn't it very likely that even if TCU goes undefeated the rest of the way, they still won't catch San Diego? San Diego is likely to end the regular season undefeated with their Pioneer schedule - probably winning games by double-digit margins (I'm being kind).

If this last statement is true, how can the index then ever "correct itself" unless a team loses?

[89Hen]

Do they use margin of victory, no matter how lowly the opponent?
Yes. If you read his website it says the reason they dropped out of the BCS computations was because the BCS stopped using margin of victory and Matthews likes margin and refuses to change his method.

[BigApp]

BigApp - Sorry, but you can't question the legitimacy of the GPI. Only Delaware fans can do that. xlolx xlolx

Sorry Cluckie, guess I gotta keep learnin' the rules the hard way! xlolx

[*****]

I am no statistician... how can the index then ever "correct itself" unless a team loses?I'm not a statistician either but the GPI uses a blend of W/L and MOV systems so all the eggs are not in one basket. As I linked, the systems always seem to improve as more data is fed into them. Massey explained that very well on WAVES a couple weeks ago.

[Lehigh Football Nation]

I'm not a statistician either but the GPI uses a blend of W/L and MOV systems so all the eggs are not in one basket. As I linked, the systems always seem to improve as more data is fed into them. Massey explained that very well on WAVES a couple weeks ago.

Right, the GPI not only spreads it out between human/computer pollsters, it throws out the highest and lowest computer rankings, so Matthews' #1 rating of San Diego gets thrown out of the GPI for this week.

It *does* raise the question about having too many indexes in the GPI that overly weigh margin of victory, though. The BCS changed their index in this way which is worth noting. There are two polls that put San Diego at #1 and #2, and only one can get thrown out.

I guess I am mystified mostly because San Diego is not in my personal Top 25, yet computer pollsters (average: #12) and human pollsters (average: #20) have them much much higher than I would put them.

[*****]

... It *does* raise the question about having too many indexes in the GPI that overly weigh margin of victory, though. The BCS changed their index in this way which is worth noting... I guess I am mystified mostly because San Diego is not in my personal Top 25, yet computer pollsters (average: #12) and human pollsters (average: #20) have them much much higher than I would put them.MOV systems typically outperform W/L systems as Massey pointed out on WAVES. He was not in favor of the BCS eliminating them and said it was a political move. As for San Diego, though it is too early to tell, it seems to be sorta like Hampton.
Now:
GPI 32 Hampton, Polls 20 18 17
GPI 14 San Diego Polls 21 21 20
Last year:
GPI 14T Hampton Polls 4 2 2

[89Hen]

it seems to be sorta like Hampton.
Last year:
GPI 14T Hampton Polls 4 2 2
:nono: :nono: :nono: :nonono2: :nonono2: :nonono2:

[bluehenbillk]

Big App you're the man. The GPI is fundametally flawed for over valuing computer rankings, which account for 2/3 of teams total rating. I'm with 89Hen, the AGS Poll is more accurate & should be proven so come 11/19.

[*****]

:nono: :nono: :nono: :nonono2: :nonono2: :nonono2:What I typed is completely correct.

[89Hen]

What I typed is completely correct.
You can't have it both ways Ralph. Either the GPI is just an indicator of the field and does not actually rank teams, or it does both. When I bring up that the AGS Poll had the missing playoff selection (Lafayette) more highly ranked than the GPI, you quickly change your tune.

Decide, is it a playoff indicator only, or does it actually rank teams too?

[GOTOREROS]

:nono: :nono: :nono: :nonono2: :nonono2: :nonono2:

No more Jim Harbaugh picture? You were a billboard for San Diego and our coach - please put his picture back up! :D

GOTOREROS

[89Hen]

No more Jim Harbaugh picture? You were a billboard for San Diego and our coach - please put his picture back up!
He's actually the guy in the middle covering his face. He actually did think there was an Old Guard and went around undercover trying to find them to plead your case to them.

[*****]

... Either the GPI is just an indicator of the field and does not actually rank teams, or it does both..."The Gridiron Power Index (GPI), the index ranking for I-AA and top indicator of at-large playoff selection ..." The GPI has always been a ranking and top indicator. That is what any ranking aspires to.

[89Hen]

"The Gridiron Power Index (GPI), the index ranking for I-AA and top indicator of at-large playoff selection ..." The GPI has always been a ranking and top indicator. That is what any ranking aspires to.
OK, then when I say that the AGS Poll had the missing playoff pick last year ranked higher than the GPI did, and therefore was more accurate, stop saying the GPI isn't a ranking.

[Lehigh Football Nation]

MOV systems typically outperform W/L systems as Massey pointed out on WAVES. He was not in favor of the BCS eliminating them and said it was a political move. As for San Diego, though it is too early to tell, it seems to be sorta like Hampton.
Now:
GPI 32 Hampton, Polls 20 18 17
GPI 14 San Diego Polls 21 21 20
Last year:
GPI 14T Hampton Polls 4 2 2

You'll note that Hampton massively corrected this year on the GPI simply because they lost to SC State, which (arguably) put them back to where they "really belong".

What was Hampton last week, before they lost to SC St? I bet the computers put them higher than the human polls (maybe #8-10?) and now they've lost, they're well below (#32)

Point being, that the computers work better when they have a data point which defines a "loss". Otherwise, the computer only has "blowout wins" to work with.

[*****]

OK, then when I say that the AGS Poll had the missing playoff pick last year ranked higher than the GPI did, and therefore was more accurate, stop saying the GPI isn't a ranking.Yes, the best poll around had Lafayette in their top 25 though the GPI had them ranked far outside it. Lafayette lost in the first round. Are you saying the AGS Poll was more correct? Are you saying YSU was not correct as the GPI indicated?

[*****]

... What was Hampton last week, before they lost to SC St? I bet the computers put them higher than the human polls (maybe #8-10?) and now they've lost, they're well below (#32)...The prior week:
GPI 20T Hampton Polls 12 11 11 Computers 26 23 20 37 18 35

[cosmo here]

Yes, the best poll around had Lafayette in their top 25 though the GPI had them ranked far outside it. Lafayette lost in the first round. Are you saying the AGS Poll was more correct? Are you saying YSU was not correct as the GPI indicated?

Lafayette lost in the first round at the eventual national champions, and held a 17-10 halftime lead on Appalachian State. The game was 20-20 after three quarters. Don't try to take that away from us since we weren't highly ranked in a BCS-type formula.

[89Hen]

Are you saying the AGS Poll was more correct? Are you saying YSU was not correct as the GPI indicated?
Yes, more correct. No, YSU was correct. LC losing in the first round?... that's what the AGS would have expected, by a margin of about what it was. The GPI would have had you believe ASU would blow out LC with a 40 spot spread in ranking. :p

[*****]

Lafayette lost in the first round at the eventual national champions, and was the only team in the playoffs last season to hold a halftime lead on Appalachian State (17-10). The game was 20-20 after three quarters. Don't try to take that away from us since we weren't highly ranked in a BCS-type formula.No, no. Not taking anything away from the 'Pards as I've stated many times including when Coach Tavani has been on WAVES. BTW, the GPI is only BCS-like in that it mixes polls and computers. The polls and computers are very different in nearly every way.

[*****]

... The GPI would have had you believe ASU would blow out LC with a 40 spot spread in ranking.No, the GPI is not predictive. The disparity might lead you to believe that ASU would win, they did, but that is all AGS.

[Cincy App]

Lafayette lost in the first round at the eventual national champions, and held a 17-10 halftime lead on Appalachian State. The game was 20-20 after three quarters. Don't try to take that away from us since we weren't highly ranked in a BCS-type formula.

The Pards were very impressive in Boone last year. They also played against ASU when Richie Williams was healthy.

[89Hen]

No, the GPI is not predictive. The disparity might lead you to believe that ASU would win, they did, but that is all AGS.
Hogwash. If a system ranks things in order of superiority it is predictive. Otherwise the rankings are worthless. xcoffeex

[DocMike]

89Hen and Bluehenbillk complaining about the GPI. And the sun rose. Some things that can be counted on.

The GPI does rank (index) teams so that when the autobids are removed the next 8 are indicated as playoff selections. But the way ranking is used by 89Hen would indicate that the GPI is to also be used as indicator of the seeding. That is not the case here, so the complaint about "ranking" in the GPI is not valid. The hogwash is that you are imposing your understanding on the system. There are systems out there who do say that being predictive is what the system seeks to accomplish. The GPI does not have that nor does it claim to do so.

Though with 89Hen and Bluehenbillk, the GPI use of computer systems (or the use of a select group of computer systems) will never meet their standards. 89Hen wants to look at each individual part and question why that one is used. Bluehenbillk preaches against the sacrilege of computer usage in the process advocating for their elimination and reliance on polls only.

I would offer the challenge to both of you to develop your own systems and methods and put them here for public scrutiny. It is much more work than just sitting there with your repetitive posts. So if you continue with your current methods, you will only show that you have nothing to add.

[89Hen]

But the way ranking is used by 89Hen would indicate that the GPI is to also be used as indicator of the seeding. That is not the case here, so the complaint about "ranking" in the GPI is not valid.
WRONG, WRONG, WRONG... it is RALPH that is now claiming the GPI is a ranking of teams. A ranking of teams means that the system thinks 1 is better than 2 which is better than 3 and so on. That is what a ranking is. If the GPI's sole function is picking the field, then there is no reason to publish beyond the 8th at-large team that should be selected. Furthermore, saying the GPI picked ASU #1 last year while the Polls didn't (Ralph's point, not mine) is moot if the GPI does not indicate seeding or ranking. It merely indicates that ASU would be a correct selection (I think they actually had the auto last year).

[*****]

89Hen and Bluehenbillk complaining about the GPI. And the sun rose. Some things that can be counted on.

The GPI does rank (index) teams so that when the autobids are removed the next 8 are indicated as playoff selections. But the way ranking is used by 89Hen would indicate that the GPI is to also be used as indicator of the seeding. That is not the case here, so the complaint about "ranking" in the GPI is not valid. The hogwash is that you are imposing your understanding on the system. There are systems out there who do say that being predictive is what the system seeks to accomplish. The GPI does not have that nor does it claim to do so.

Though with 89Hen and Bluehenbillk, the GPI use of computer systems (or the use of a select group of computer systems) will never meet their standards. 89Hen wants to look at each individual part and question why that one is used. Bluehenbillk preaches against the sacrilege of computer usage in the process advocating for their elimination and reliance on polls only.

I would offer the challenge to both of you to develop your own systems and methods and put them here for public scrutiny. It is much more work than just sitting there with your repetitive posts. So if you continue with your current methods, you will only show that you have nothing to add.:bow: :bow: :bow: :bow: :bow: :bow: :bow: ALL HAIL DOC MIKE :bow: :bow: :bow: :bow: :bow: :bow: :bow:

[89Hen]

I would offer the challenge to both of you to develop your own systems and methods and put them here for public scrutiny. It is much more work than just sitting there with your repetitive posts. So if you continue with your current methods, you will only show that you have nothing to add.
IT IS RIGHT HERE.... http://www.i-aa.org/article.asp?articleid=80344

[usdtoreros]

IT IS RIGHT HERE.... http://www.i-aa.org/article.asp?articleid=80344

You're really arguing that one human poll is better than a mix of several human and computer polls? So I-A should just pick one of the human polls for the BCS? Being an applied math major, I'd really like to see some defense of this.

[*****]

... it is RALPH that is now claiming the GPI is a ranking of teams. A ranking of teams means that the system thinks 1 is better than 2 which is better than 3 and so on...Now claiming???? When has the GPI been anything else? Rankings can be predictive and not. The GPI is not. AGS. AGS. AGS. AGS. AGS. AGS.

The GPI measures what a team did, not what it is going to do.

[89Hen]

You're really arguing that one human poll is better than a mix of several human and computer polls? So I-A should just pick one of the human polls for the BCS? Being an applied math major, I'd really like to see some defense of this.
I don't give a flying F what the BCS does. And yes, I'm arguing that one human poll is better than a mix of several human and computer polls. The proof is in the pudding. The AGS was 7 of 8 in predicting the field last year and had the lone miss ranked higher than the GPI which also had 7 of 8. IOW, the GPI takes the AGS and makes it less accurate. Defend that.

[89Hen]

Rankings can be predictive and not. The GPI is not. AGS. AGS. AGS. AGS. AGS. AGS.

The GPI measures what a team did, not what it is going to do.
Then it is worthless. We know what teams did. If you're not going to use a ranking to compare how Team A would do against Team B then what's the use??? If you say Montana State is 24 and Western Carolina is 30 but they never meet, so we can't possibly know how they did against each other, what is the point?!!>?:bang: :bang: :bang: :bang: :bang: :bang: :nonono2: :nonono2: :nonono2: :nonono2: :nonono2:

[usdtoreros]

I don't give a flying F what the BCS does. And yes, I'm arguing that one human poll is better than a mix of several human and computer polls. The proof is in the pudding. The AGS was 7 of 8 in predicting the field last year and had the lone miss ranked higher than the GPI which also had 7 of 8. IOW, the GPI takes the AGS and makes it less accurate. Defend that.

I bet one AGS voter's individual ranking predicted all 8 teams correctly, so by your argument, his or her ranking is the best and the AGS takes that ranking and makes it less accurate.

[89Hen]

I bet one AGS voter's individual ranking predicted all 8 teams correctly, so by your argument, his or her ranking is the best and the AGS takes that ranking and makes it less accurate.
Could very well be, we have some very smart voters around here.

[usdtoreros]

Then it is worthless. We know what teams did. If you're not going to use a ranking to compare how Team A would do against Team B then what's the use??? If you say Montana State is 24 and Western Carolina is 30 but they never meet, so we can't possibly know how they did against each other, what is the point?!!>?:bang: :bang: :bang: :bang: :bang: :bang: :nonono2: :nonono2: :nonono2: :nonono2: :nonono2:

Teams are awarded a playoff spot based on their past performance and then the playoffs determine who wins those matchups, giving us a National Champion. To me, that is exactly what the poll should be doing, putting the teams in that deserve to be in based on past performance.

[89Hen]

Teams are awarded a playoff spot based on their past performance and then the playoffs determine who wins those matchups, giving us a National Champion.
:nod: That's the beauty of I-AA, but we're not talking about playoffs. We're talking about rankings and polls.

[usdtoreros]

:nod: That's the beauty of I-AA, but we're not talking about playoffs. We're talking about rankings and polls.

Exactly. And I am saying the polls represent how deserving teams are based on past performance, not how one team would do against another in the future. Some polls, like Sagarin are made to be predictive, but most are not.

[*****]

Then it is worthless...Then you should be following predictive/fortune teller models and not the GPI. :p

What are the tea leafs saying this week?

[Torero Tradition]

The GPI has the Toreros in the Playoffs and going most likely to Montana... the tea leafs have the Toreros home field advantage throughout. xlolx

[89Hen]

Some polls, like Sagarin are made to be predictive, but most are not.


Then you should be following predictive/fortune teller models and not the GPI
:bang: :bang: ALL polls and rankings should be predictive in the fact that they tell us how teams are perceived relative to each other even if the two have never met. Polls and computer models take into account all the factors they can to arrive at the notion that one team has perfomed better than another even though they may have no opponents in common. Ralph, your tea leaves comment could just as easily be directed at a ranking that "measures what a team did". If two teams have never faced each other and haven't faced any common opponents, there is no way to "measure what a team did". If the computer models and human polls can make some measure of teams based on what they've done relative to one another even though they've never met, they they are judging which team has done better and would therefore expected (or predicted) to be better head to head.

Use an example of Team A and Team B who through five games have played the exact same five teams. Team A goes 5-0 against the five and Team B goes 3-2. Which team would be ranked higher and which team would be expected to win head-to-head? Obviously this is not a perfect system and when A and B meet, B might beat the stuffing out of A. But that doesn't mean there wasn't an expectation or that the ranking wasn't predictive.

Unfortunately, most teams in the rankings do not play each other and many have no common opponents. That's why we use rankings, records, home field, etc... to be able to compare how two teams have performed with their schedules. Once we compile all of the data, we rank who has performed the best. This ranking then becomes a way of predicting who the better team is. To say any ranking is non-predictive means that the ranking is worthless. You may as well just post conference standings and leave it at that.

[*****]

... To say any ranking is non-predictive means that the ranking is worthless. You may as well just post conference standings and leave it at that.xlolx xlolx xlolx No, friend. Just no. There is too much info out there, google some. Predictive systems vs. not. There is a difference and both have validity.

[89Hen]

No, friend. Just no. There is too much info out there, google some. Predictive systems vs. not. There is a difference and both have validity.
Haven't found it yet, but one interesting thing I did find was Beck's info on predictions in I-AA by the computers models. Right now the highest percentage correct goes to the Born Power Index who is 256-105 (.709) even up. In the GoHens pool I am currently 132-36 (.785). :smiley_wi

[*****]

Haven't found it yet, but one interesting thing I did find was Beck's info on predictions in I-AA by the computers models. Right now the highest percentage correct goes to the Born Power Index who is 256-105 (.709) even up. In the GoHens pool I am currently 132-36 (.785). :smiley_wiI am just not into predictive stuff, too much like betting/gambling to me.

[BigApp]

So now polls are based on what a team did and aren't meant to be predictive?:confused: :confused: :confused:

Then why do we have pre-season polls?

:nonono2: :nonono2: :nonono2: :nonono2: :nonono2:

[89Hen]

I am just not into predictive stuff, too much like betting/gambling to me.
I don't wager on sports but I like them and I really believe all are predictive since they in fact rank the teams. To say they're non-predictive is a cop out. The teams have a number next to them. To say it's a crapshoot each week is ridiculous. If it were really a crap shoot, over time, all models would be around .500.

[*****]

I don't wager on sports but I like them and I really believe all are predictive since they in fact rank the teams. To say they're non-predictive is a cop out. The teams have a number next to them. To say it's a crapshoot each week is ridiculous. If it were really a crap shoot, over time, all models would be around .500.89H, predictive models use different criteria along with what non-predictive systems use. Predictive means they are trying to predict, adding factors like homefield advantage, etc. Non-predictive means they are not trying to predict but more... I guess interpret what has happened. The factors are so varied for interpretation is why (as Massey explained on WAVES) there are different results. He said you want a broad scope to capture the essense of each a little in the final result. All the systems are sound in their formula though, even if they look wacky at times to us. It's fun to watch them, ranking all the teams even.

[*****]

Haven't found it yet, but one interesting thing I did find was Beck's info on predictions in I-AA by the computers models. Right now the highest percentage correct goes to the Born Power Index who is 256-105 (.709) even up. In the GoHens pool I am currently 132-36 (.785). :smiley_wiRemember when I got him to set that up? That was directly from AGS discussions (with JSO too IIRC)! :thumbsup:
http://tbeck.freeshell.org/fb/prediaa.html

[DocMike]

ALL polls and rankings should be predictive in the fact that they tell us how teams are perceived relative to each other even if the two have never met. Polls and computer models take into account all the factors they can to arrive at the notion that one team has perfomed better than another even though they may have no opponents in common. Ralph, your tea leaves comment could just as easily be directed at a ranking that "measures what a team did". If two teams have never faced each other and haven't faced any common opponents, there is no way to "measure what a team did". If the computer models and human polls can make some measure of teams based on what they've done relative to one another even though they've never met, they they are judging which team has done better and would therefore expected (or predicted) to be better head to head.

All you need to do is take the mirror test to see what the problem with the GPI really is. The mirror test requires you to go into your bathroom look into the mirror and there will be the problem. The reason I can tell you this is your statement that all polls and rankings "should be predictive". Your prejudice creates your own problem. You also reveal another basis as to why the mirror test is perfect for you.

"Polls and computer models take into account all the factors they can to arrive at the notion that one team has perfomed better than another even though they may have no opponents in common."

No computer model or poll takes into account all of the factors they can. The computer model is based on analysis of factors that are believed to be important. Polls are human based and I am very sure that the voters don't take in all the factor that they can. Just like the computer models they look at a limited range of factors to arrive at their choices. The only one that shows its work and is consistent is the mathematical model. So, 89Hen you have seen the enemy and it is you. Your desire to force your limited viewpoints on what polls and computer models should be will always prove you wrong as you by your own admission not using them for what they are designed you are also admitting you don't really understand or know how they are constructed.

Though your petulant attacks are quite amusing. I am sure that you won't let facts that you yourself have presented stop you from your erroneous beliefs. Look forward to hearing more.

[89Hen]

Though your petulant attacks are quite amusing. I am sure that you won't let facts that you yourself have presented stop you from your erroneous beliefs. Look forward to hearing more.
Edited. I won't bother lowering myself to you.

[89Hen]

Remember when I got him to set that up? That was directly from AGS discussions (with JSO too IIRC)! :thumbsup:
:nod: JSO is our resident computer lover.

[*****]

Edited. I won't bother lowering myself to you.Raise yourself to Doc Mike, creator of the GPI! :nod:

[BigApp]

Gosh,

I guess I should just be quiet and accept the fact that some legitimate playoff caliber team could very well lose a well deserved playoff bid (anyone remember a GPI #4 ranked Villanova?) because a few 35-40 year olds who still live in mom's basement think a non-scholarship I-AA team is as good or better than Georgia Tech, TCU or Texas Tech.

If you can't see the problem in that, then maybe YOU should be the one looking in the mirror. Before you respond, check your blood pressure medication.

I want you (whomever is reading this) to ask yourself, right now, if you think San Diego (the example being used in this discussion) is as good as Georgia Tech/TCU/Texas Tech. Responses from all are welcome.

If the answer is yes, then you should seek therapy. If you answer no, then maybe you can understand, at least partly, what the real issue is.

[89Hen]

Raise yourself to Doc Mike, creator of the GPI! :nod:
And rather arrogant. :nono:

[BigApp]

So now polls are based on what a team did and aren't meant to be predictive?:confused: :confused: :confused:

Then why do we have pre-season polls?


:confused:

[*****]

Gosh....smack goes on the smack board smackman ----->

[*****]

And rather arrogant. :nono:well, his idea has done fairly well huh?

[89Hen]

well, his idea has done fairly well huh?
No better than the AGS poll.

[*****]

:confused:ZING! There is no preseason GPI.

[*****]

No better than the AGS poll.Really? Where was the AGS Poll in '98, '99, '00, '01, '02, '03? How has the best POLL of I-AA measured against the GPI since it started in 2004?

[BigApp]

smack goes on the smack board smackman ----->

You would know! And thanks for not answering, btw.: smh :

[usdtoreros]

Gosh,

I guess I should just be quiet and accept the fact that some legitimate playoff caliber team could very well lose a well deserved playoff bid (anyone remember a GPI #4 ranked Villanova?) because a few 35-40 year olds who still live in mom's basement think a non-scholarship I-AA team is as good or better than Georgia Tech, TCU or Texas Tech.

If you can't see the problem in that, then maybe YOU should be the one looking in the mirror. Before you respond, check your blood pressure medication.

I want you (whomever is reading this) to ask yourself, right now, if you think San Diego (the example being used in this discussion) is as good as Georgia Tech/TCU/Texas Tech. Responses from all are welcome.

If the answer is yes, then you should seek therapy. If you answer no, then maybe you can understand, at least partly, what the real issue is.

This is the exact reason the GPI takes an average of multiple polls (computer and human). In the same way USD is ranked among GT, TCU, and TT, they are ranked 64th in one poll. Some polls emphasize winning, some polls emphasize SOS and are put together to take in account as many variables as possible.

[*****]

You would know! And thanks for not answering, btw.: smh :ZING! smack board is over there ------>

[DocMike]

89Hen,
Am I arrogant because I "permit" computer models and polls to be what they are instead of making them what they "should" be? You have persistently posted your displeasure with the GPI itself, the components of the GPI, etc. You have championed the AGS poll (which in its brief history has done well) as being superior based on the past two years. The GPI history is much longer and has proven itself. The problem with a poll though goes back to some of the rationale behind the development of the GPI. In those days the selection basis could not be quantified. I began doing this to see if method of identifying teams that have earned an at-large spot in the playoffs. I agree with many here and have posted also that USD should not be considered a playoff team due to the poor schedule they have. I have confidence that the GPI will over the next few weeks with the upcoming games continue to identify those 8 at-large teams that have "earned" a chance to keep playing. But do remember that the GPI has never selected a single team for the playoffs. The committee has always done that and always will. If they want to use the GPI fine, if they want to ignore it fine. The GPI is what it is, and it has done what it was designed to do. You may disagree with that but you cannot disprove that. After a while you can compare the GPI to the AGS poll and draw conclusions to your hearts content. My "arrogance" is actually my attempt to imitate your technique of opinion over fact.

[*****]

89Hen,
Am I arrogant because I "permit" computer models and polls to be what they are instead of making them what they "should" be? You have persistently posted your displeasure with the GPI itself, the components of the GPI, etc. You have championed the AGS poll (which in its brief history has done well) as being superior based on the past two years. The GPI history is much longer and has proven itself. The problem with a poll though goes back to some of the rationale behind the development of the GPI. In those days the selection basis could not be quantified. I began doing this to see if method of identifying teams that have earned an at-large spot in the playoffs. I agree with many here and have posted also that USD should not be considered a playoff team due to the poor schedule they have. I have confidence that the GPI will over the next few weeks with the upcoming games continue to identify those 8 at-large teams that have "earned" a chance to keep playing. But do remember that the GPI has never selected a single team for the playoffs. The committee has always done that and always will. If they want to use the GPI fine, if they want to ignore it fine. The GPI is what it is, and it has done what it was designed to do. You may disagree with that but you cannot disprove that. After a while you can compare the GPI to the AGS poll and draw conclusions to your hearts content. My "arrogance" is actually my attempt to imitate your technique of opinion over fact.I am, like everyone else, reading this with great interest.

[*****]

89Hen and Bluehenbillk complaining about the GPI. And the sun rose. Some things that can be counted on...sorry, still chuckling after that... xlolx xlolx xlolx xlolx xlolx xlolx xlolx xlolx xlolx this goes waaaaaay back folks...

[Keeper]

If Mr 89hen is so distrusting of computers
why is he using one so much?

If the GPI is so abhorrent to his senses,
if it's all about the results on the field,
why so much gnashing of teeth about it?

These are media snapshots people.
They are not affecting your wallet or
your ability to reproduce. (At least not yet,
the computers haven't gotten quite there)

[HensRock]

Hopefully I can shed a little light here. I've been following the GPI for many years and I like it alot. I'm not a big fan of changing it, but that's a different story. Mathematics is sort of a hobby of mine and I can appreciate the efoort that goes into indexes such as the GPI. Many football fans do not have this appreciation for the theories behind the indexes. (Doc, I've read your papers long ago :thumbsup: )

I beleive there has been a misunderstanding between ranking and a predictive model which boils down to the definition of "predictive". While all predictive models must essentially rank teams, the converse is not true. Not all rankings are predictive. If a poll or index "ranks" team A higher than team B, then Yes, you should expect team A to beat team B on most Saturdays, however (89 take note here) this is NOT what "predictive" means in the mathematical sense. A predictive model attempts to quantify just how much stronger team A is than team B and so could be used to predict the expected point differential of A vs. B on a neutral field, not just who would win - but how how much would they win by. The GPI makes no attempt at this. Nor does any poll. Predictive models will usually say as much or at least tell you that the expected point differential can be derived from subtracting the index values of the two teams. Sagarin is predictive. Massey (even though it is a RATING) is not. The GPI is not predictive, nor is the AGS poll. (at least in the mathematical sense of the word)

[89Hen]

Am I arrogant because I "permit" computer models and polls to be what they are instead of making them what they "should" be? You have persistently posted your displeasure with the GPI itself, the components of the GPI, etc.
No, you're arrogant because of posts like...

"All you need to do is take the mirror test to see what the problem with the GPI really is. The mirror test requires you to go into your bathroom look into the mirror and there will be the problem....

Though your petulant attacks are quite amusing. I am sure that you won't let facts that you yourself have presented stop you from your erroneous beliefs."

And you are wrong about my displeasure about the GPI itself. My problem lies with the computer rankings. My apparent displeasure with the GPI is that it has been championed as the best, but yet has not outperformed the fans here that vote in the AGS Poll.

[89Hen]

If Mr 89hen is so distrusting of computers
why is he using one so much?

If the GPI is so abhorrent to his senses,
if it's all about the results on the field,
why so much gnashing of teeth about it?
:confused: I wouldn't let a car wash give my kid a bath, but I'll certainly take my car there.

And where did I ever say the GPI is abhorrent to my senses? I've never even been close to saying that. Read the thread again and you'll see that I am arguing about what it is, what it does and why I think it IS predictive whether that's not the intent or not.

BTW, UC Davis at #6 and Towson at #8???

[89Hen]

boils down to the definition of "predictive".... A predictive model attempts to quantify just how much stronger team A is than team B and so could be used to predict the expected point differential of A vs. B on a neutral field, not just who would win - but how how much would they win by. The GPI makes no attempt at this... Sagarin is predictive. Massey (even though it is a RATING) is not. The GPI is not predictive, nor is the AGS poll. (at least in the mathematical sense of the word)
I agree 100% and was thinking about that very thing this morning. However, it has been (or should have been to anyone) very clear that my use of "predictive" is not the mathematical definition. I would never say that the GPI numbers correlate to any kind of prediction of point spread. But I would say the the GPI standing does correlate to a prediction of who should win a game head-to-head. Not that it will... no system is 100%. 75% seems to be a VERY high percentage rate for head-to-head, even-up predictions.

Even if the intent of the GPI is not to predict which team would win, we all know that's how it is viewed by 99% of the people that look at it. To say it can't possibly predict who would win head-to-head IMO is throwing it under the bus. IMO there is only one method of ranking that is truly 100% non-predictive and that is the conference standings. Those don't take into account who played who, only if they won or lost. It's also interesting to note that they are generally the only ranking system that doesn't assign values to the teams of 1, 2, 3, 4... when they are posted. They simply list the teams in order of W-L record.

I don't hate the GPI as others would have you believe. However, I don't think it's as great as others would have you believe. I am very suspect of many of the computer rankings it uses. I know that a big part of the GPI is to moderate (average, disseminate, take your pick...) the data that it uses to arrive at a final answer. However, I don't know why the saying of garbage in, garbage out seems to not apply to the GPI. The simple fact that the formula changes and evolves year to year tells me that it agrees with me that some of the components are in fact broken.

Anyway, don't let the rhetoric of the computer huggers make you think I don't appreciate the work of the GPI. I just prefer the polls.

[89Hen]

Really? Where was the AGS Poll in '98, '99, '00, '01, '02, '03? How has the best POLL of I-AA measured against the GPI since it started in 2004?
Hey, don't get me started again. You know it's your fault for not getting this website up and running in 1998 so we could have started the poll then. :p

As for the two years we do have to compare... the GPI and AGS are in 100% agreement. Both got 7 of 8 at large having both missed Lehigh and Lafayette for Cal Poly and YSU. Hmmm, maybe we need to add in a Patriot factor to both. ;)

2004 AGS indicates...
Automatics
1. Southern Illinois (29)
2. Furman
5. Montana
7. William & Mary
13. Jacksonville State
14. Hampton
24. Lafayette
15. Northwestern State

At-Large
3. Western Kentucky
4. Georgia Southern
6. New Hampshire
8. James Madison
9. Delaware
10. Sam Houston State
11. Eastern Washington
16. Cal Poly

17. Lehigh* Selected by committee over Cal Poly

BTW, once again the AGS had their missing team ranked higher and all 16 playoff teams were in the Top 24 of the AGS. :p :p :p

[bluehenbillk]

However, I don't know why the saying of garbage in, garbage out seems to not apply to the GPI. The simple fact that the formula changes and evolves year to year tells me that it agrees with me that some of the components are in fact broken.


I don't know how many times I have to say it, the GPI is a bad idea that was created from a worse idea. The only way it could get worse is to give the GPI credibility outsie of this website which luckily is what it has been contained to.

It'll be very interesting to see this year's results and any type of "last-week computer fixes" as what the computers are spitting out doesn't look to match the accuracy that very few people regularly tout.