Top Rankings for Computer and Human Poll Biases - Week after 10/24 Games

[eastbayaggie]

Top Rankings for Computer and Human Poll Biases - Week after 10/24 Games*


Top 10 Teams Where the Rankings are Biased Towards the Computer's Favor

1 Northern Arizona 0.387
Middle 5 Computer Rankings: #9, #10, #11, #12, #13
Human Rankings: #18, #21, #22

2 Delaware 0.375
Middle 5 Computer Rankings: #5, #8, #8, #9, #9
Human Rankings: #14, #16, #18

3 Massachusetts 0.366
Middle 5 Computer Rankings: #10, #11, #11, #12, #13
Human Rankings: #19, #19, #23

4 UC Davis 0.246
Middle 5 Computer Rankings: #18, #19, #19, #21, #21
Human Rankings: #30, #31, Unranked

5 Eastern Washington 0.207
Middle 5 Computer Rankings: #16, #16, #17, #18, #20
Human Rankings: #21, #24, #26

6 Jacksonville State 0.204
Middle 5 Computer Rankings: #13, #14, #16, #18, #19
Human Rankings: #20, #22, #22

7 Northern Iowa 0.183
Middle 5 Computer Rankings: #7, #7, #7, #11, #11
Human Rankings: #12, #12, #14

8 Southern Utah 0.142
Middle 5 Computer Rankings: #20, #21, #22, #23, #30
Human Rankings: #34-38(Tied), #44, Unranked

9 South Dakota State 0.118
Middle 5 Computer Rankings: #6, #6, #6, #7, #7
Human Rankings: #9, #9, #9

10 Montana State 0.107
Middle 5 Computer Rankings: #21, #23, #23, #24, #25
Human Rankings: #32, #34, #39-40(Tied)


Top 10 Teams Where the Rankings Are Biased Towards the Human Poll's Favor

1 McNeese State -0.528
Middle 5 Computer Rankings: #24, #27, #28, #28, #30
Human Rankings: #11, #13, #13

2 Appalachian State -0.519
Middle 5 Computer Rankings: #17, #18, #22, #23, #25
Human Rankings: #7, #8, #8

3 Central Arkansas -0.425
Middle 5 Computer Rankings: #23, #25, #26, #26, #34
Human Rankings: #12, #15, #15

4 Holy Cross -0.255
Middle 5 Computer Rankings: #27, #27, #32, #32, #33
Human Rankings: #17, #17, #20

5 South Carolina State -0.216
Middle 5 Computer Rankings: #15, #15, #16, #17, #18
Human Rankings: #10, #10, #10

6 Elon -0.187
Middle 5 Computer Rankings: #10, #11, #11, #12, #13
Human Rankings: #6, #6, #6

7 Liberty -0.139
Middle 5 Computer Rankings: #23, #26, #26, #26, #29
Human Rankings: #21, #23, #24

8 Colgate -0.103
Middle 5 Computer Rankings: #36, #38, #38, #40, #46
Human Rankings: #24, #25, #25

9 Southern Illinois -0.096
Middle 5 Computer Rankings: #5, #5, #5, #5, #7
Human Rankings: #3, #3, #3

10 Weber State -0.083
Middle 5 Computer Rankings: #15, #16, #17, #17, #17
Human Rankings: #13, #14, #16



*Computer Poll: Using BCS formula, % of points of the 7 computer rankings used in the GPI(-high and low rankings)
Human Poll: Using BCS formula, (% of points AGS + % of points SN + % of points coaches poll)/3
Bias = computer poll - human poll

Personal Terse Disclosure: Both computer and human polls have their own unique biases and misgivings.


Bonus Ranking

Top 3 Teams Where the Rankings are Biased Between Human Polls

1 Delaware 0.195** AGS poll - FCS poll (BCS formula)
2 Eastern Washington 0.131*** FCS poll - TSN poll (BCS formula)
3 Liberty 0.104**** AGS poll - FCS poll (BCS formula)

** Delaware received 992 points out of 88 voters in the Any Given Saturday poll and only 179 points out of 28 voters in the FCS Coaches poll.
*** Eastern Washington received 141 points out of 28 voters in the FCS Coaches poll and only 240 points out of 138 voters in The Sports Network poll.
**** Liberty received 440 points out of 88 voters in the Any Given Saturday poll and only 67 votes out of 28 voters in the FCS Coaches poll.

Link to the Gridiron Power Index (GPI)
http://www.collegesportingnews.com/stats/writer/GPI/20091026gpi.html

[CamelCityAppFan]

That's an interesting exercise. I think that early to mid-season, the human polls do a better job of ranking team's positions as the human polls take in to account factors that the computer polls cannot. Late in the season, the computer polls get much better as the data they are based on becomes more complete.

[ToTheLeft]

That's an interesting exercise. I think that early to mid-season, the human polls do a better job of ranking team's positions as the human polls take in to account factors that the computer polls cannot. Late in the season, the computer polls get much better as the data they are based on becomes more complete.

Such as Southern Utah being in the top 25. The computers don't have the whole picture yet. The GWFC is overrated to this point, and will balance out as they play each other.

Edit: The GWFC isn't bad. It's just not the second best conference.

[JMU2K_DukeDawg]

Well, that certainly support the East Coast Bias theory we find from so many West of the Mississippi.

[MacThor]

There are five computer rankings that have Northern Iowa #1.

I am becoming less and less concerned about computers becoming self-aware and overthrowing their human overlords.

[biggie]

Computers suck, that's about all I can ever figure out from computer polls. Believe that's why we like FCS over B(c)S.

[molly]

There are five computer rankings that have Northern Iowa #1.

I am becoming less and less concerned about computers becoming self-aware and overthrowing their human overlords.

Which five have UNI #1? All of the seven computer rankings used in the GPI have UR at #1; the humans and computers all agree on that.

The highest ranking for UNI of those rankings is #6.

[MacThor]

Dolphin Predictor, Sagarin Predictor, Reid, O'Malley, Markov.

You are correct, none of those are in the GPI.

I suppose those ratings are overweighting UNI's loss to BCS contender Iowa.

[UNIFanSince1983]

The computers rate us so high because they all have Iowa rated so high and a 1 point loss still holds water with the computer models. If you look at the BCS rankings Iowa is #1 in all but 1 of the computer rankings used, and in the other one they are #2.

If Iowa falls UNI will fall in the computers as well.

[GannonFan]

That's an interesting exercise. I think that early to mid-season, the human polls do a better job of ranking team's positions as the human polls take in to account factors that the computer polls cannot. Late in the season, the computer polls get much better as the data they are based on becomes more complete.


Computers, however, never get close to complete as even at the end of the regular season they only deal with a very small sample size (11 games) and very few intertwined games amongst the conferences. There's a reason why the GPI doesn't outperform people when it comes to predicting the final field of 16 - it's inherently flawed because of the computer components. xthumbsupx

[Native]

Such as Southern Utah being in the top 25. The computers don't have the whole picture yet. The GWFC is overrated to this point, and will balance out as they play each other.

Edit: The GWFC isn't bad. It's just not the second best conference.

I am guessing that the GWFC is rated that highly because of their out of concerence wins over the FCS competition.

Is there another conference besides the CAA with a superior OOC record?

[89Hen]

I am guessing that the GWFC is rated that highly because of their out of concerence wins over the FCS competition.

Is there another conference besides the CAA with a superior OOC record?
xeyebrowx Cal Poly > SDSU is really the only good win.

UCD beat Winston Salem and DII Western Oregon
SUU beat Texas State and DII Dixie State
UND beat NWSt and Stony, lost to NAIA Sioux Falls and got whalloped by SFA.
USD beat SELA, Drake and William Penn (a high school in Delaware xsmiley_wix)

Am I missing something?

[ToTheLeft]

I am guessing that the GWFC is rated that highly because of their out of concerence wins over the FCS competition.

Is there another conference besides the CAA with a superior OOC record?

And their SOS is held up by early season FBS games. Once everyone plays the Dakotas and SUU and the overall record of everyone's opponents goes down it will level out. But they did make a solid showing OOC, and I can't take that away from them. (Except for UND)

[89Hen]

The GWFC is overrated to this point, and will balance out as they play each other.
Don't count on it. The computers often self-implode and will continue to rank conferences high all season.

[jacksfan29]

And their SOS is held up by early season FBS games. Once everyone plays the Dakotas and SUU and the overall record of everyone's opponents goes down it will level out. But they did make a solid showing OOC, and I can't take that away from them. (Except for UND)

Who made a solid showing OOC? Poly beat SDSU at home, take a look at the other wins. Who exactly have SUU, Davis or USD beaten which would qualify as a quality OOC win? Drake, William Penn, W Wash, Dixie State? I think the computers putting the Great West in 2nd is proof that human polls should outweigh the computers.

[ToTheLeft]

Solid showing. Not "quality wins". They weren't just a doormat for the Big Sky, they played teams that they could beat and for the most part, they beat them. Wins over SELA and TXST aren't "quality" but they're solid.

I mean, I personally think the conference is overrated in this poll by quite a bit. Just trying to rationalize why the computers have them this way...

[eastbayaggie]

An alternative take is that there aren't many bad losses.
The teams won the games they should have won, and lost to good quality teams.

Cal Poly lost to: Montana, San Jose State, and Ohio
UC Davis lost to: Montana, Boise State, and Fresno State
Southern Utah lost to: Utah State, Northern Arizona, and San Diego State
South Dakota lost to: Montana State and Northern Iowa
North Dakota lost to: SF Austin, Texas Tech, and undefeated NAIA Sioux Falls

Interestingly enough, Sioux Falls is ranked between UC Davis and Southern Utah in the Massey rankings. Yeah. I know. They're still NAIA.xsmhx

[89Hen]

An alternative take is that there aren't many bad losses.
The teams won the games they should have won, and lost to good quality teams.

Cal Poly lost to: Montana, San Jose State, and Ohio
UC Davis lost to: Montana, Boise State, and Fresno State
Southern Utah lost to: Utah State, Northern Arizona, and San Diego State
South Dakota lost to: Montana State and Northern Iowa
North Dakota lost to: SF Austin, Texas Tech, and undefeated NAIA Sioux Falls
Then Towson should be in the Top 10. They lost to Northwestern, UNH and Delaware. :p

[eastbayaggie]

Then Towson should be in the Top 10. They lost to Northwestern, UNH and Delaware. :p
... as well as losing to 1-6 Northeastern and a #72 ranked Morgan State. :p

[jacksfan29]

An alternative take is that there aren't many bad losses.
The teams won the games they should have won, and lost to good quality teams.

Cal Poly lost to: Montana, San Jose State, and Ohio
UC Davis lost to: Montana, Boise State, and Fresno State
Southern Utah lost to: Utah State, Northern Arizona, and San Diego State
South Dakota lost to: Montana State and Northern Iowa
North Dakota lost to: SF Austin, Texas Tech, and undefeated NAIA Sioux Falls

Interestingly enough, Sioux Falls is ranked between UC Davis and Southern Utah in the Massey rankings. Yeah. I know. They're still NAIA.xsmhx

I respect Davis and Poly and SUU seems to be improving. But your alternative is to look at "good" losses? The FBS losses fine (though Utah State is a terrible fb program/team) and I'll give UND a break on the loss to Sioux Falls. That loss should be blamed on their AD and coach. That game was a set up from the day it was announced, there was no way the Sioux were going to get up to play an NAIA mid-season. However, South Dakota lost to UNI 66-7 and if I recall UND gave up 60+ to SFA. San Diego State is non-scholorship and UNC beat them. Those are not "good" losses.

[GoAgs72]

San Diego State is FBS and scholarship. University of San Diego is FCS and non-scholarship. UC San Diego doesn't have football. That should clear up the muddy water.

[jacksfan29]

San Diego State is FBS and scholarship. University of San Diego is FCS and non-scholarship. UC San Diego doesn't have football. That should clear up the muddy water.

oops, misread xoopsx.

[whoanellie]

that's why they play it off here!!!

[SDFS]

I respect Davis and Poly and SUU seems to be improving. But your alternative is to look at "good" losses? The FBS losses fine (though Utah State is a terrible fb program/team) and I'll give UND a break on the loss to Sioux Falls. That loss should be blamed on their AD and coach. That game was a set up from the day it was announced, there was no way the Sioux were going to get up to play an NAIA mid-season. However, South Dakota lost to UNI 66-7 and if I recall UND gave up 60+ to SFA. San Diego State is non-scholorship and UNC beat them. Those are not "good" losses.

Well, I am sure that the AD would have preferred not to schedule that game, but I think UND lost three scheduled games during the summer this year (Idaho St in Grand Forks, Iona in Grand Forks and Weber St on the road.) and they had to scramble to fill the schedule the best they could. It is year two of the transition and it is what it is .. I do not mind UND scheduling the SFA and Texas Tech teams. If they are going to be successful at the FCS level the program needs to be prepared for the spread offense - it is a good bench mark to measure progress.

I think the primary reason why the Great West is ranked so high is they don't have rock bottom programs that tank the overall rankings and as UND and USD go through transition and fill the programs with FCS recruits. They will only get better. So, if you think you are upset now.. this is the low point of the conference.. two teams in 2nd year of transition plus SUU now at max scholarship levels only going up from here..

[jacksfan29]

Well, I am sure that the AD would have preferred not to schedule that game, but I think UND lost three scheduled games during the summer this year (Idaho St in Grand Forks, Iona in Grand Forks and Weber St on the road.) and they had to scramble to fill the schedule the best they could. It is year two of the transition and it is what it is .. I do not mind UND scheduling the SFA and Texas Tech teams. If they are going to be successful at the FCS level the program needs to be prepared for the spread offense - it is a good bench mark to measure progress.

I think the primary reason why the Great West is ranked so high is they don't have rock bottom programs that tank the overall rankings and as UND and USD go through transition and fill the programs with FCS recruits. They will only get better. So, if you think you are upset now.. this is the low point of the conference.. two teams in 2nd year of transition plus SUU now at max scholarship levels only going up from here..

You mean back to where it was when SDSU and NDSU were in it xwhistlex...

we shall see if USD and UND can fill the shoes of those who came before them.

[Keeper]

Computers, however, never get close to complete as even at the end of the regular season they only deal with a very small sample size (11 games) and very few intertwined games amongst the conferences. xthumbsupx

I agree with that aspect, thus you have named the challenge we attempt to master.
Would like to know how many games you actually watch each week to provide a
representative ranking 1 thru 125. To that end, the computers support unbiased
input into the argument for or against a particular team for playoff inclusion.
Human polls and computer rankings are each (hopefully) considered logically for what each furnishes the committee.

[89Hen]

I think the primary reason why the Great West is ranked so high is they don't have rock bottom programs that tank the overall rankings and as UND and USD go through transition and fill the programs with FCS recruits. They will only get better. So, if you think you are upset now.. this is the low point of the conference.. two teams in 2nd year of transition plus SUU now at max scholarship levels only going up from here..
Agreed on why the GW has an overall high ranking, but don't count your chickens on the UxD's and SUU just yet. Ask NDSU how year 6 is going. Several teams moving up have had quick success followed by mediocrity. Samford went to the semifinals a couple years after moving up (1991), went to the playoffs the next year but haven't been back since. Hofstra went to the playoffs 5 times pretty quickly upon moving up but haven't been back in 8 years. Not saying that will happen to USD and UND, but it's not a given that they will actually improve. xpeacex

[89Hen]

ranking 1 thru 125
That's really all the computers are good for... actually 30-125. xcoffeex

[I-AA Fan]

Computers suck, that's about all I can ever figure out from computer polls. Believe that's why we like FCS over B(c)S.

I think just the opposite, Use you team as an example. Sure, we knew that the team was better (a couple weeks ago) than their sub-.500 record. However, you have to show me that you are good. Any personal bias has to go. It allows you to be a fan of your favorite teams, and still remain fair.

http://petespoll.com/

Sorry for the shameless plug. We have new Conference power index, that shows a conference's improvement (or lack there of) each week.

[89Hen]

http://petespoll.com/

Sorry for the shameless plug. We have new Conference power index, that shows a conference's improvement (or lack there of) each week.

1 Richmond
2 Montana
3 New Hampshire
4 Southern Illinois
5 S. Carolina St.
6 South Dakota St.
7 Villanova
8 Elon
9 William & Mary
10 Northern Iowa
11 Appalachian St
12 Holy Cross
13 Stephen F. Austin
14 Central Arkansas
15 Massachusetts
16 Colgate
17 Liberty
18 McNeese St
19 Weber St.
20 E. Washington
21 Cal-Poly
22 Northern AZ
23 E. Illinois
24 Jacksonville St.
25 Delaware

If you're using computers... they suck.

[GannonFan]

I agree with that aspect, thus you have named the challenge we attempt to master.
Would like to know how many games you actually watch each week to provide a
representative ranking 1 thru 125. To that end, the computers support unbiased
input into the argument for or against a particular team for playoff inclusion.
Human polls and computer rankings are each (hopefully) considered logically for what each furnishes the committee.

See, that's the part where humans have the advantage - there's really no need to look much past #25 - there's a whole host of data there that just means very little in the grand scheme of things. Looking at the top teams (and for arguments sake you can at least just start with teams with the required 7 DI wins) then you can narrow down the analysis a lot more. Humans can then look at stats from the game, even play by play, and how teams match up in terms of offensive/defensive philosophies, and also look at who played and didn't play, what games came before and after, and so on. Hard to get all of that into a computer. You watch enough football, you get a better and better feel for things. A computer won't ever get that in a tiny sample size and limited interconference games. xthumbsupx

[Keeper]

Good points, Gannon.

Also, humans can reflect on who's hot & not at the end of the season
and more deserving of inclusion. I am all for a selection committee to
seed the playoffs and participants. However, I suspect that computer
results are consulted as a part of a tie-breaker for seeding purposes
or to pick the last at-large qualifiers. Nothing wrong with that.

[SDFS]

We shall see if USD and UND can fill the shoes of those who came before them.
You do realize that UNC, NDSU, SDSU, USD and UND are all following UNI. I think UNO will be the next to jump in the near future.

[SDFS]

Agreed on why the GW has an overall high ranking, but don't count your chickens on the UxD's and SUU just yet. Ask NDSU how year 6 is going. Several teams moving up have had quick success followed by mediocrity. Samford went to the semifinals a couple years after moving up (1991), went to the playoffs the next year but haven't been back since. Hofstra went to the playoffs 5 times pretty quickly upon moving up but haven't been back in 8 years. Not saying that will happen to USD and UND, but it's not a given that they will actually improve. xpeacex

Valid point, but I do not see NDSU down for long.

[NC Aggie]

I respect Davis and Poly and SUU seems to be improving. But your alternative is to look at "good" losses? The FBS losses fine (though Utah State is a terrible fb program/team) and I'll give UND a break on the loss to Sioux Falls. That loss should be blamed on their AD and coach. That game was a set up from the day it was announced, there was no way the Sioux were going to get up to play an NAIA mid-season. However, South Dakota lost to UNI 66-7 and if I recall UND gave up 60+ to SFA. San Diego State is non-scholorship and UNC beat them. Those are not "good" losses.

San Diego State is a scholarship D1A school (although not that good). I think you are confusing them with the USD Toreros who are "non-scholarship". Now you could have confused them with UCSD......but they're only good with Pascal and Cobol

[89Hen]

Valid point, but I do not see NDSU down for long.
I wouldn't expect them to be either. xpeacex

[JohnStOnge]

Next year, I really do need to do an experiment in which I get some of you "computer" skeptics to agree to pick games randomly selected from all FCS vs. FCS games and also predict the margin of victory each of your favorites will enjoy. I'll pick one of the more well known models, like Sagarin, and compare how it does in those two respects to how each of you skeptics do. And I'll tell you right now what will happen: If you beat the model in terms of correctly picking winners, you won't beat it by much. And you probably won't beat it. When it comes to how close you come on average to margin of victory, you will have no shot against the model. It will be much, closer on average than you will be.

As I've written before I've done that before with a local radio talk show host who fancies himself to be good at picking games. For quite some time he kept ridiculing power ratings when I brought them up so I did that to him; only strictly with BCS conference I-A teams so that he'd be familiar with them. Before the season I randomly selected some number of games. Then, each week on the Thursday I'd call him up and get his picks. I'd record the Sagarin system's picks. When it was all over the Sagarin system beat him both in terms of percent winners picked correctly and "bias." Bias is the average of the quantity (observed - expected); which in this case is (actual margin of victory by favorite -predicted margin of victory by favorite). The target is 0.

Anyway, I think if I go through the trouble of doing that some of you skeptics are going to be in for a rude awakening. By random chance some of you may beat it by a small margin in terms of percent correct. But none of you are going to beat it when it comes to the bias measure.