Ranking / Polls

[Ronin]

OK, this is sure to ruffle some feathers or fur as the case may be. However, I found it interesting when comparing the rankings that some teams rank much higher on subjective polls, than they do in computerized rankings.

I've provided the raw data below with some simple analysis.


Team Subj. Avg Comp. Avg Rank Diff.
UNI 1.5 1 -0.5
Umass 4.00 3.3 -0.7
SIU 7.00 4.7 -2.3
McNeese 5.00 6 1
JMU 6.00 6.7 0.7
ASU 10.00 6.7 -3.3
UNH 8.50 8.3 -0.2
Montana 3.00 12.7 9.7
Delaware 8.50 11.7 3.2
Wofford 11.00 11.7 0.7
Richmond 14.00 11.7 -2.3
Elon 16.50 11 -5.5
EKU 17.50 19.7 2.2
Youngstown 20.00 19.3 -0.7
Norfolk 23.00 39.7 16.7
Fordham NR 43.7 NA

What's interesting is how close most of them are with two exceptions.

Haven't figured out yet how to do a nice table -- I'll work on it.

[WileECoyote06]

Voters often penalize teams for losing to strong teams by dropping them; where as computers will eventually reward a team for losing to a strong team, by increasing their schedule difficulty.

A team with a good record and name recognition looks good to voters.
A team with a strong record is weighed against its schedule for computers to figure out their take on it.

[grizbeer]

Montana is always ranked low by the computers, and is ranked high in the human polls until the playoff get going. Although the computer polls are supposed to be non-biased, their opening ranking has to come from somewhere, and this bias will last a long time, especially in a conference like the BSC where all the teams start off perceived as weak.

For example, the Dwiggins computer program dropped Montana 3 spots this week to #21 after beating UNC 52-7. I'm not sure what Montana could have done this week to improve their ranking if 52-7 dropped them 3 spots.

At the end of the season, and especially during the playoffs, the computers will start to make more sense, since everyone will have played their schedule, and conference SOS bias will start to drop out.

And yes I understand I just made SD case for them, but I do support letting an undefeated SD team into the playoffs. let's see what they can do.

[appfan2008]

Montana is one school that is ranked all over the place anywhere from 1 to about 10

[griz_fan_in_SanDiego]

Montana is always ranked low by the computers, and is ranked high in the human polls until the playoff get going. Although the computer polls are supposed to be non-biased, their opening ranking has to come from somewhere, and this bias will last a long time, especially in a conference like the BSC where all the teams start off perceived as weak.

For example, the Dwiggins computer program dropped Montana 3 spots this week to #21 after beating UNC 52-7. I'm not sure what Montana could have done this week to improve their ranking if 52-7 dropped them 3 spots.

At the end of the season, and especially during the playoffs, the computers will start to make more sense, since everyone will have played their schedule, and conference SOS bias will start to drop out.

And yes I understand I just made SD case for them, but I do support letting an undefeated SD team into the playoffs. let's see what they can do.

I agree...let them in. What's the harm in giving them a chance to prove themselves. If they get it handed to them in the playoffs after an undefeated season everyone will no whether they are TRULEY overhyped like most of you USD haters say in here. Most people say they are over-rated without ever even seeing them play. I have been to every home game this year, which doesn't make me a fan i just like to watch football and their games are inexpensive to attend, and I've seen a quality football team with some real talent.

[Khan4Cats]

Montana is always ranked low by the computers, and is ranked high in the human polls until the playoff get going. Although the computer polls are supposed to be non-biased, their opening ranking has to come from somewhere, and this bias will last a long time, especially in a conference like the BSC where all the teams start off perceived as weak.

For example, the Dwiggins computer program dropped Montana 3 spots this week to #21 after beating UNC 52-7. I'm not sure what Montana could have done this week to improve their ranking if 52-7 dropped them 3 spots.

At the end of the season, and especially during the playoffs, the computers will start to make more sense, since everyone will have played their schedule, and conference SOS bias will start to drop out.

And yes I understand I just made SD case for them, but I do support letting an undefeated SD team into the playoffs. let's see what they can do.

I'm not sure I would trust the Dwiggins at all, they still have SIU ahead of UNI despite the who 6 yards an 6 points thing. I would have given it a benefit of doubt about opponents' strength but SIU has played Indiana State and Southern Utah already while UNI has them coming up.xconfusedx

I almost agree on San Diego, they might just end up in if they finish 11-0.

[Ronin]

For example, the Dwiggins computer program dropped Montana 3 spots this week to #21 after beating UNC 52-7. I'm not sure what Montana could have done this week to improve their ranking if 52-7 dropped them 3 spots.

It could be that other teams had beat better opponents to move up in the rankings.

While the computer rankings early in the year mean little (no one has really proven themselves). The human polls are fallible to name tradition (name recognition) as a means of improving their rank.


PS I used GPI, Dunkel and Sagarin for the computer rankings, not Dwiggins.

[89Hen]

While the computer rankings early in the year mean little
They mean little later in the year too. xcoffeex

[grizbeer]

It could be that other teams had beat better opponents to move up in the rankings.

While the computer rankings early in the year mean little (no one has really proven themselves). The human polls are fallible to name tradition (name recognition) as a means of improving their rank.


PS I used GPI, Dunkel and Sagarin for the computer rankings, not Dwiggins.Dwiggins is part of GPI, FWIW. For Montana it would have been one of the extremes thrown out, but counted for others as part of ARC.

[mvemjsunpx]

The problem with most computer rankings I've seen is that they tend to greatly overvalue strength of schedule & undervalue actual wins & losses.

Take, for example, the RPI in basketball. It uses the formula of:

- 25% winning percentage
- 50% strength of schedule (opponents' winning percentage)
- 25% opponents' strength of schedule (opponents' opponents' winning percentage)

This means that about 70-75% of the rating is based on the strength of your opponents & only 25-30% is based on your own wins & losses (taking into account the fact that your team shows up repeatedly on your opponents schedules). I think approximately a 50/50 split is far more fair.

I like what the GPI does, but it weights the computer rankings more heavily than the polls (5/8 vs. 3/8). I think they should average the computer rankings and make that 50%, & then average the human polls and make that 50%. They should also extend the human poll rankings into the "others receiving votes" section, so you get about 40 ranked & there are fewer "26's"/fewer teams with the same number.