Help I'm so confused

by Rockin on Nov 3, 2008 11:59 PM EST

The computer polls are confusing the hell out of me. I understand that they are based on Strength of Schedule and such but what I don't understand is the Colley Matrix and Peter Wolfe polls still having Texas ahead of Texas Tech. Don't these polls also take into account head to head matchups?

Team	AH	RB	CM	KM	JS	PW

1. Alabama	25	22	24	23	24	25

2. Texas Tech	24	25	21	25	25	22

3. Penn State	23	24	22	22	22	23

4. Texas	22	23	25	24	23	24

5. Florida	20	20	23	21	19	20

6. Oklahoma	19	19	20	20	20	21

7. Southern California	15	21	17	17	15	17

8. Utah	21	15	19	19	21	18

9. Oklahoma State	17	14	14	15	16	14

10. Boise State	18	16	18	16	18	19

What am I missing? Someone please help me understand what I see as a flaw in these polls. Is this just bad programming?

Thanks for the fanpost! Please do not post any content from a premium site that requires a subscription. Also, if you just want to share a link, quote, or video, please consider using fanshots instead. Thanks.

Tweet Comment 22 comments | 0 recs |

Do you like this story?

Comments

First...

First, stop calling them computer polls. They’re statistical polls. They have nothing to do with computers other than the fact that it’s quicker to have a computer do the math than to have you do it on a piece of paper. The statistics behind most of the polls (except Billingsley, who is insane, and I’m not sure about Anderson-Hester) are hundreds of years old.

Second: no, they don’t take into account head-to-head matchups in any special way. How could they? You’re not ranking Texas vs. Texas Tech. You’re ranking all 119 teams – in order to do that, you need to take into account all of the games for all of the teams.

If you think head-to-head matchups are so important, why isn’t Oregon State above USC? Why isn’t Ole Miss above Florida? The simple fact is that you can’t make a simple ordered list of teams and not have places where you violate head-to-head matchups. So if you’re going to violate them someplace, why does it matter where you do it?

As to why Texas is ranked above Texas Tech: when you get to unbeatens, it’s very difficult to separate them, because you have no information as to where to rank the team. They could be just slightly better than the best team they beat. They could be just slightly worse than the best team they beat, since the best team doesn’t always win (Otherwise, Florida would not have lost to Ole Miss).

The cases where Texas comes out ahead of Texas Tech is just because Texas’s wins so far are very strong (which is why they had such a huge lead), and their loss is not that bad (as it’s to a very good team). Texas Tech’s wins just don’t produce as strong a resume, even though they beat Texas. As an example – Texas already has a victory almost as good as Texas Tech, over Oklahoma. They also have a victory over Oklahoma State, which Texas Tech doesn’t have.

Most fans focus on head-to-head matchups exclusively, ignoring the cases where they can’t fit them into the polls as “flukes.” You have to accept that the better team doesn’t always win.

The easiest way to think about it is to ignore the actual teams that were beaten, and just think of them as “Texas beat #6/#9”, etc.

The main reason that Texas Tech is ranked low in Wolfe and Massey’s ratings is because Division IAA teams really suck in those two rankings, and Texas Tech played two of them. Take a look at Sagarin and Massey’s ratings, and you’ll note that Sagarin has a Division IAA team ranked 20. I can’t say I really agree with that.

by Bleed Blue 'n White on Nov 4, 2008 1:14 AM EST reply actions

It's the truth

Texas, IMHO, is still the #1 team in the country. To go 3-1 in 4 consecutive weeks against top 11 opponents, with your only loss coming with 1 second left on the road against a top 10 opponent (at the end of the 4 weeks, mind you, when the team is presumably beaten up) doesn’t give me any reason to believe they shouldn’t still be #1. It would be a travesty to leave UT out of the NC if they win out the rest of the way. That being said…

Can someone explain to me why Tech jumped PSU? Tech has 1 quality win, over Texas at home, and 1 decent win, a mauling of Kansas. Penn State has 1 quality win, Ohio State on the road, and what I would consider 2 quality wins, both at home, against Oregon St and Illinois. Even if you want to figure OrSt out of the equation (we have gone 5-1 since our loss to you), that still leaves the same (rank wise) number of quality wins for both teams. Now maybe it’s just me, but I’ve always had the traditional way of looking at this situation as such: you should be rewarded for a continuation of winning. Is beating Texas at home really that much better than beating Ohio State on the road? My basic instincts tell me no.

Now, I know this is a Penn St blog, so I’m sure you all have homer and biased views. But most of you seem like intelligent, level-headed fans. Could somebody PLEASE explain why Tech should be ranked above PSU at this current point?

Oregon State: where play action defense and healthy QBs thrive

by The VD Special on Nov 4, 2008 6:03 AM EST up reply actions

Probably not yet

but I don’t mind it simply because if they go undefeated, they would have jumped us at some point anyway. If they lose, they don’t have a chance of finishing above us as long as we take care of ours.

If anything I get mad about the disrespect it shows. But in the end, I think we’ll get the opportunity to prove them all wrong anyway.

by jimbo2psu on Nov 4, 2008 8:51 AM EST up reply actions

well it isn't really disrespectful

since the humans have us at #2, and computers and statistics can neither respect, nor disrespect, anyone. They just look purely at the numbers (except for Billingsley, as BBnW points out). I am a little surprised that TTU is tied for #1 in the computers, and I haven’t dug through the different methods/statistical categories used, but perhaps they just have the right “good” stats at the moment.

I do agree, though, that as it stands now, it doesn’t matter, since the next 2 weeks will settle the argument one way or another.

by The JuggerNitt on Nov 4, 2008 10:24 AM EST up reply actions

Texas Tech

Texas Tech is #1 in Billingsley (insane), and Sagarin and Massey, both of whom for some crazy reason have several Division IAA teams waaay high (as high as #20!), including the two Texas Tech played.

I’m going to have to stare at those a bit more. The disagreement there isn’t small, it’s gigantic: Sagarin has JMU at #20, Wolfe has JMU at #90. I’ve always had a nagging belief that including Division IAA is “bad” due to the poor connection between the two – I’ve just never had much evidence to support it.

To explain more: when two sections become more disconnected from each other, the disconnected sections tend to spread out along the ranking. Think of it this way: if you ranked Division IAA by itself, and Division IA by itself, the best team in IAA and the best team in IA would have similar rankings. When you add connections in, you start realigning the two leagues.

Colley has a (rather arbitrary) trick around it, and I’m not sure why Wolfe’s rankings seem a bit more reasonable.

by Bleed Blue 'n White on Nov 4, 2008 1:13 PM EST up reply actions

computers and statistics can neither respect, nor disrespect, anyone

But the statistical models used to make the calculations were created by humans. So, there should be someone behind this we can blame?

by NJ lion on Nov 4, 2008 2:58 PM EST up reply actions

Yeah, but most of them are dead.

by Bleed Blue 'n White on Nov 4, 2008 3:05 PM EST up reply actions

Can we at least blame ELO? I always hated that band.
Rats! Just from writing this I now have “Evil Woman” stuck in my head.

by confirmy on Nov 4, 2008 3:28 PM EST up reply actions

I like ELO...

Blame Kasparov.

There is a tractor in the parking lot, West Virginia license EIEIO. Your lights are on.

by leeharvey418 on Nov 4, 2008 3:56 PM EST up reply actions

In the spirit of today

can’t we blame the terrorists somehow?

by Bleed Blue 'n White on Nov 4, 2008 4:01 PM EST up reply actions

If PSU doesn't play for the national championship

then the terrorists have won.

I bleed Blue and White.

by Horse N Buggy on Nov 5, 2008 8:50 AM EST up reply actions

Not the first time I've posted this, but THIS is the source of Sagarin's system:

by Cairo on Nov 5, 2008 2:42 PM EST up reply actions

shit, good point

I forgot the BCS standings were the only ones with us at #3. Good enough then. It’ll all come out in the wash.

by jimbo2psu on Nov 5, 2008 10:45 AM EST up reply actions

I agree

Although is rare for a team to jump from #7 to #2, while jumping another undefeated team this late in the year, I don’t really have a problem with it. They win out and they deserve it; they lose and it won’t be a problem.

by Screen Name 20 on Nov 4, 2008 10:48 AM EST up reply actions

man, those statisticians were really prepared 100s of years ago

to start working that early on a way to figure out what 2 college teams should play for the MNC…man, that’s looking ahead!

by The JuggerNitt on Nov 4, 2008 10:20 AM EST up reply actions

The burst really came in the early 1900s, but some of the ideas are pretty old. Colley’s method is almost directly a copy of a statistical inference method from Laplace.

Obviously they didn’t care about football, but networks of pairwise comparison have applications in lots of areas: it’s crucial in understanding preference studies in social sciences, but it also has application in the stock market and disaster response.

by Bleed Blue 'n White on Nov 4, 2008 12:58 PM EST up reply actions

oh I know

well…maybe I didn’t know the specific details, but yeah, I’ve had enough statistics to understand. I just liked the thought of guys sitting around a room a couple hundred years ago thinking, “some day we’ll need to figure out a slightly less arbitrary way of picking a national champion in some sport that may or may not already exist. Instead of having them play it out on a field, let’s just calculate who we think is the best!”

by The JuggerNitt on Nov 4, 2008 1:36 PM EST up reply actions

Sounds like

total revenge of the nerds.

by jimbo2psu on Nov 5, 2008 10:47 AM EST up reply actions

Minority Report

Someday we’ll get to be like that Tom Cruise movie “Minority Report” where they predict murders and arrest people before they commit them. We’ll have a computer tell us who is going to win and we won’t have any need to play the game. Maybe it will even tell us when the running back is going to tear his ACL so he can just go right to the hospital and have the surgery done without dealing with the pain.

Mike
Black Shoe Diaries

by BSD on Nov 5, 2008 5:21 PM EST up reply actions

computer rankings of who is good

So when you say that “texas had a good win beating #6/#9”, where do those numbers come from, do the polls feed into the statistical rankings?

by JoePaPa on Nov 4, 2008 10:27 AM EST up reply actions

Nope

The numbers come from the ranking system themselves.

You’re basically arranging the entire college football system to have the fewest number of unlikely results as possible. An “unlikely result” would be something like Texas ahead of Texas Tech: but there’s a bit of fuzz there, because it has to be possible for a lower-ranked team to beat a higher-ranked team. So really, you’re trying to make Texas “lower, or not that much higher” than Texas Tech.

Texas ends up being higher because pushing them lower would make their wins less likely. So it’s a balance between having Texas too high or too low.

For Texas Tech it’s a bit different: I’ve got that described in the first In Defense of Statistics here.

by Bleed Blue 'n White on Nov 4, 2008 1:05 PM EST up reply actions