Keeper Ratings (a GPI Component)

[BigApp]

Jus' for you RussB:

http://home.att.net/~keepers.football/wsb/html/view.cgi-html2.html--SiteID-845888.html

I-AA October 22, 2006
1-North Dakota State
2-Massachusetts
3-Illinois State
4-Appalachian State
5-Montana
6-UC-Davis
7-Furman
8-Towson
9-Portland State
10-Youngstown State
11-Western Kentucky
12-Eastern Illinois
13-New Hampshire
14-James Madison
15-Princeton
16-Cal Poly
17-Maine
18-Montana State
19-Northern Iowa
20-Richmond
21-Coastal Carolina
22-Tennessee-Martin
23-Northeastern
24-Harvard
25-Wofford

[JMU Duke Dog]

This rating system still has JMU behind UNH... :bang:

[PantherRob82]

Jus' for you RussB:

http://home.att.net/~keepers.football/wsb/html/view.cgi-html2.html--SiteID-845888.html

I-AA October 22, 2006
6-UC-Davis
8-Towson
9-Portland State
19-Northern Iowa
20-Richmond


:confused: :confused: :bang: :bang: :bang: :bang:

[smallcollegefbfan]

This one and the DWiggins top 25s are a joke. Maybe we need to look at changing what is taken into the GPI. How about just the TSN, AGS, and CSN polls with Sagarin or something?

[*****]

I don't have the final Keeper from last year. As for using three polls and one system that doesn't even rank all I-AA's opponents isn't the GPI. And yes, the GPI is highly accurate.

[smallcollegefbfan]

I don't have the final Keeper from last year. As for using three polls and one system that doesn't even rank all I-AA's opponents isn't the GPI. And yes, the GPI is highly accurate.

I think all polls work themselves out. I was just giving a suggestion. I don't think those computer polls are worth much now but they always seem to work themselves out.

[*****]

I think all polls work themselves out...Some people here don't seem to get that point.

[MiloCat]

I don't have the final Keeper from last year. As for using three polls and one system that doesn't even rank all I-AA's opponents isn't the GPI. And yes, the GPI is highly accurate.
Highly accurate compared to what?

[*****]

Highly accurate compared to what?Compared to what the committee selects. It has never been off by more than one team, ever.

[MiloCat]

Compared to what the committee selects. It has never been off by more than one team, ever.
Isn't it one of the tools the committe uses? So if it is influencing the committee's decision, how can it be a bad "predictor".

[boonedocks]

Isn't it one of the tools the committe uses? So if it is influencing the committee's decision, how can it be a bad "predictor".


Just a disclaimer, I don't have a problem with the GPI. Luckily in I-AA we get to crown a champion on the field. However the issue that I gather from posts is that if these polls/rankings influence the playoff committee's decision, then it can affect who gets in or doesn't/who wins or doesn't. Also if these individual polls/rankings are somewhat questionable by a majority, I can understand a person being skeptical of the GPI. Obviously in years past it has been effective. I can see where some fans may be concerned based on these early results. :twocents:

Again ralph don't be upset i don't have a problem. I feel it will work out but i can also see the flip side and understand some concerns

[*****]

... I feel it will work out but i can also see the flip side and understand some concernsEvery year we preface the GPI discussion with a bit about how computer rankings work... ie. they get more precise as the year goes on. The GPI starts in October, giving the computers a chance to get filled with enough data to at least be presentable. I think the process towards precision is interesting because it has always resulted in a pretty good result for the GPI. I'll always be around to discuss the results and misconceptions/inaccuracies concerning the GPI so don't take that as being upset. Funny thing is though, the computer systems don't hear the critics... they just plow through the data and when mixed in the GPI, help finish with a top answer. No poll had Appalachian State #1 prior to the playoffs starting last year, the GPI did.

[texcap]

This rating system still has JMU behind UNH... :bang:

Yes, but you're catching up. Maybe if you can beat them a couple of more times you could move in front of them! :smiley_wi

[*****]

This rating system still has JMU behind UNH... :bang:The GPI doesn't. :D

[Tod]

The GPI doesn't. :D

The whole is greater than the sum of its parts. :thumbsup:

[Appstate29]

Yeah but the only reason the GPI had ASU #1 last year was because of us playing LSU, it boosted up our Sagarin rating, I think he had us #1 before the playoffs started too

[Tod]

Yeah but the only reason the GPI had ASU #1 last year was because of us playing LSU, it boosted up our Sagarin rating, I think he had us #1 before the playoffs started too

And you're arguing AGAINST the GPI? :confused:

[*****]

Yeah but the only reason the GPI had ASU #1 last year was because of us playing LSU, it boosted up our Sagarin rating, I think he had us #1 before the playoffs started tooFirst, get Sag out of your head because it is essentially a BCS I-A rating that also includes I-AA because I-AA is their only other opponent. It doesn't rate I-AA opponents who are not D-I. The GPI is a different concept than one rating. The reason ASU was number one in the GPI prior to the playoffs last year is because six of the computer ratings in the GPI had ASU #1. http://www.i-aa.org/article.asp?articleid=74448 The polls had ASU 2-5.

[blur2005]

The whole is greater than the sum of its parts. :thumbsup:
Rousseau would be proud.

[*****]

Rousseau would be proud.As someone who double-majored in political science and philosophy I like Rousseau. :thumbsup:

[89Hen]

Some people here don't seem to get that point.
And others don't get that if they're wildly inaccurate at week seven, there's a pretty darn good chance they'll be inaccurate at week twelve. :nod: :nod: :nod:

[*****]

And others don't get that if they're wildly inaccurate at week seven, there's a pretty darn good chance they'll be inaccurate at week twelve. :nod: :nod: :nod:Chance maybe, reality:
The 2005 regular season-ending GPI was off by only one team in indicating at-large playoff selection, the committee chose No. 41 Lafayette over No. 12 Youngstown State. Here is the GPI indicated field:
1. Appalachian St*
2. New Hampshire*
3. Texas St
4. Montana
5. Cal Poly
6. Furman
7. N Iowa*
8. Ga Southern
9. S Illinois
10. E Washington*
11. Richmond
12. Youngstown St
14T. Hampton*
18. Nicholls St*
23T. E Illinois*
40. Colgate*
*= conference automatic qualifier
(Games through 11/19/05)

[89Hen]

they get more precise as the year goes on.

No poll had Appalachian State #1 prior to the playoffs starting last year, the GPI did.
My golf game gets more precise as the year goes on, but I'm not ready for the Tour by September.

BTW, if you want to play the hindsight game you may want to mention...

2004 - Southern Illinois #1 GPI, lost first round
2003 - Southern Illinois #1 GPI, lost first round
2002 - McNeese State #1 GPI, stayed #1 in final GPI :eek:

[DocMike]

Ralph,
With a double major of political science and philosophy you are eminently qualified in this type of work. It would only be better if you had gotten a minor in statistics. Maybe a degree in sports management? Law school? Hell, it wouldn't matter as no matter what background you have on this board, there is always an "expert" opinion to be found on any and all matters. :nod: :nod: :nod: :smiley_wi :smiley_wi :smiley_wi

Liberte, Egalite, Fraternite!! (That is as close as I am going to get to the French spelling (no hope for the accent marks!!):hurray: :hurray: :hurray:

AGS - The People's Board xsmileyclapx xsmileyclapx xsmileyclapx

[*****]

... if you want to play the hindsight game you may want to mention...
2004 - Southern Illinois #1 GPI, lost first round
2003 - Southern Illinois #1 GPI, lost first round
2002 - McNeese State #1 GPI, stayed #1 in final GPI :eek:Yep, every year you listed the GPI indicated 7 of the 8 at-large selections... even coining the term "woofed" for the one left out. :nod:

[*****]

... It would only be better if you had gotten a minor in statistics...I did take two stat courses because I love it but I'll leave it up to Massey to select the computers for the GPI. :D

[89Hen]

Yep, every year you listed the GPI indicated 7 of the 8 at-large selections... even coining the term "woofed" for the one left out. :nod:
You change your tune faster than my 10 year old with her hand on the radio buttons. If the GPI is just about picking the field, then STOP trying to use examples like picking AppSt #1 to show how good the GPI is. :mad: :nono:

[*****]

You change your tune faster than my 10 year old with her hand on the radio buttons. If the GPI is just about picking the field, then STOP trying to use examples like picking AppSt #1 to show how good the GPI is.xlolx xlolx xlolx xlolx That's just an example! One that rankles those afraid of the big bad GPI! xlolx xlolx xlolx

[89Hen]

xlolx xlolx xlolx xlolx That's just an example! One that rankles those afraid of the big bad GPI! xlolx xlolx xlolx
Doesn't rankle me when I know that two of the last three #1 GPI teams lost in the first round. Numbers are on my side in that one. :D

[*****]

Doesn't rankle me when I know that two of the last three #1 GPI teams lost in the first round. Numbers are on my side in that one. :DAGS! That's why they play the games and why the NCAA has tournaments. :nod:

[Longhorn]

I thought this thread got started because of questions about Keeper's rankings? Got to admit, his (Keeper's) numbers look WAY off...maybe a look should be given to eliminating his results from the GPI?

[*****]

I thought this thread got started because of questions about Keeper's rankings? Got to admit, his (Keeper's) numbers look WAY off...maybe a look should be given to eliminating his results from the GPI?No, this thread got started with no question...
Jus' for you RussB...Look at Keeper's ratings last year... pretty awesome.

Folks, the GPI is superfine every year. Don't try to fix what isn't broken.

[Longhorn]

Come on Ralph...the question was implied...why else start the thread? I'm not questioning the GPI...just suggesting that a quick glance at Keeper's '06 rankings indicate something VERY different is going on...as compared to the other polls. Perhaps by year's end events will prove Keeper's system is accurate, and he's some kind of mad genius. Or, perhaps his system has jumped the tracks...at least for '06 and needs to be retooled. Admittedly, last year's results for Keeper were good, or better than, other systems contributing to the GPI. But Keeper's results are (to date in '06) so different they deserve watching.

[Keeper]

Ralph mentioned last year's ranking, and yes,
my system had AppSt #1 all of November like most of the
other computers, but had Northern Iowa about #12,
so not too much to read there. This year, I admit I am
not happy with the results and am not sure if the off-season
tweaks are responsible. Guess I should have left it alone.
But it would be unethical to change the system back in the
middle of the season in my view. A complete review is
definitely in the works.

Just wanted to mention that last week, mine was the
only system to predict wins by NE over NH, NIowa over
Youngstown, and SCState over Hampton. Maybe some
luck involved, but because my system depends on net
yardage instead of scores, it hopefully produces more
upset picks. Also, the rankings are heavily slanted on
won-lost vs overall strength of-schedule, but is
still a subjective measure. One person e-mailed and
said that I should blow it up and stop posting embarrasing
rankings. His only ammunition was that Towson is ranked
eighth, nothing else mentioned. Seems a little harsh for
disagreement about one team. Also, a Jackrabbit complained
about SDState's ranking, noting who they beat and two of
the losses. He conveniently did not mention their loss to
Wisc-LaCrosse. Am I supposed to just throw that out and
ignore it as a freak result? No can do. No mention of San Diego's
rank from any of their fans, surprise surprise.


I personally don't agree with my ranking order either,
but always looking at improvements in the formula to
appear more "logical". Bottom line to me is that the system
which predicts the most wins by the lowest average
differential should be acclaimed the most accurate.
The best retrodictive systems produce the best mathematical
result regarding who-beat-whom, but any ranking, whether
human or computer, is still totally subjective. I have also
said often that any composite poll should have as many
qualified voters and competent computers as possible to
expect a reasonably accurate result. Basic statistics.
Then any "out-of-range" or abnormal inputs would be
sufficiently averaged out and you fans would have less of
a reason to freak out over one person's ranking of one team.
I have confidence in the caretakers of the GPI to govern
which and how many entries into whatever percentage they
deem fair. I happen to agree with the results so far, just my
opinion.

Besides, if all rankings were in agreement to all, what would
you all have to chat about, and why post them at all?
Challenge the differences but also embrace them.
In the end all that matters is who wins their last game.

[boonedocks]

I think that post says alot, thanks keeper :hurray: that validates IMO what some have been saying. If the GPI is a determinant in the playoff field, and the creator of one of the components says the results thus far are perhaps suspect, then why shouldn't there be some trepidation from blindly trusting it. I hope and think that it will work out in the end, and that's fine. Ralph, i just wonder why you seem unwilling to allow a debate about it? I understand you think its great, but why is it unreasonable to question computer rankings? Particularly in our subdivision which we all take such pride in knowing a champ is crowned on the field. That is the issue i take from this thread. that since this can potentially have an effect on that beloved process, some can be concerned about the process and results. :twocents: :confused:

[Keeper]

Everything is subject to review,
computer ratings,
opinion polls,
the formulas and composition,
that is the fair thing to do.
Many of you want to only pick on the computers.
Many of you want to use opinions only.
Again I say, the most accurate sampling that
can be achieved is by accumulating the most
relevant data possible, ie the most informed opinions
plus the best predictive computer systems.
If you don't trust the computers, how can you possibly
trust the one you're using right now? Computer science
gave you this forum to be able to slam their legitmacy
in matters of leisure time topics discussion.


No problem with scrutinizing any aspect of the GPI
as long as all are equally viewed.
I would like to see all your opinion votes posted so that
I have the opportunity to question your rankings and
make you justify your vote.
Fair enough?

[boonedocks]

No problem with scrutinizing any aspect of the GPI
as long as all are equally viewed.
I would like to see all your opinion votes posted so that
I have the opportunity to question your rankings and
make you justify your vote.
Fair enough?


I think that is very fair. Again I don't have a problem with the rankings and polls. They give us something to discuss all week, between games. I too, have also thought it would be fun to see everyone's individual rankings. I don't want anyone to confuse the fact, I am not lambasting the GPI, or the Keeper rankings. I am just trying to make a point that some people seem not to welcome any discussion/scrutinization/debate on the GPI. My only wish is that this doesn't get read as a few App fans teaming up. I don't care if some don't have ASU as #1. Even I can without reservation understand the arguments for more than a couple other teams. I enjoy seeing all the polls, knowing the criteria etc. It makes the games more enjoyable, and provides for some good discussion.:thumbsup:

[Keeper]

I understand the natural concern for the
playoff selection process, as anyone who
is not close to that body.

The selection or non-selection of some
teams over the years will always be debated,
but I'm sure the selection committee uses
every available source to winnow it down.

As I understand it, were the GPI not in place,
the committee would still view all the polls and
some of the computers to help make their
decision. The GPI is simply a composite of
those rankings, and as such has been shown
historically to be a competent measurement,
so why not consult it?

The bigger concern to me is the lack of exposure
of many teams/conferences to the selection panel
to help jusify their picks. Unfortunately, strength
of schedule has to be the primary consideration.
If there is a mistrust of computer resolution of
the strength-of-schedule (a valid concern overall IMO)
I think the panel is aware of that.

Besides, what would we have the selectors do differently,
other than dissuade them from certain polls?
Until a chisled-into-stone qualification process is adopted
like the NFL, all available information should be weighed.
The GPI has as much informed info/data as you'll find
anywhere relative to CS football. Which would you rather
use: the current GPI, or an Internet Fan Vote to pick
the playoff teams?

anyway, it seems to me that, thanks to these message boards,
everything is pretty much scrutinized to pieces.
nothing escapes the die-hard fans of I-AA!

[Keeper]

BTW, no response to NDakotaState number 1?
Even tho they "lost" to Minnesota?

Nothing from San Diego fans for the low rank?

Aren't you comforted that my extreme ranks will
be thrown out anyway?

Also, the fact that some teams are ranked behind
others they beat is not basis for an argument.
That's why they call them upsets. Or close shaves.
Or AGS. The final score should be reward enough for you.
Rankings are predicated on the entire schedule, not just
the games you wish to be considered. If you don't like
your team's rank, they must have lost a game they shouldn't
have, or the schedule is too soft. Everything is relative,
nothing is moot.

In conclusion, I do have one big issue regarding
the other computers. Many of them calculate
on the basis that all teams begin equally from zero,
competing mathematically on the same standard
in the theory that the formula will sort it out.
This is done in the interest of fairness, or as some
professors call it, transparency.

We all know that realistically there are different
levels of competition. My strength-of-schedule
component is heavily reliant on actual out-of-conference
competition which presets the conference for rating.
This may be construed as unfair or even biased.
But it is based strictly on actual competitive results.
Not a theoretical algorithm like some, or a blend of
statistical ranking comparisons like others. I use head-to-head
yardage results vs the strength of the opponent for every game
with recent games having more of a factor. This season the
results are mixed, indeed not as good as last year I am fully
aware. Still I will stand behind my methodology.
No matter how off-the-track from average it looks.
Vive le difference!

[bluehenbillk]

Keeper-

First of all, thanks for coming on the boards & "manning up" regarding your computer system. Listening to Ralph run PR for the 3-letter system puts an even more negative spin on a questionable system.

I was happy to see that you aren't happy with the results of your model, and you shouldn't be. Let's look at the results, the 3-letter system throws out the high & low ranking of all the computer models so that means that when looking at how the system ranks it's top 25 teams, 13 out of those 25 have your results thrown out.

Lets look at some of the more mystifying of your rankings:

NDSU #1 (hey they did lose last week)
Furman #7 (CCU?)
Towson #8 (Must be that killer OOC slate: Morgan St, Elon & Liberty)

San Diego #64 (see Towson schedule remark, so much for SOS)
S. Illinois #30 (Indiana?)
JMU #14 (whitewash of UNH?)

As for your system being retrodictive: Garbage in is Garbage out

[th0m]

Nothing from San Diego fans for the low rank?


Low? :D #64 looks about right ;) Now #14 JMU on the other hand...:p