PDA

View Full Version : Sagarin's Top 8


coover
November 9th, 2008, 06:05 AM
1. JMU
2. Appy
3. Richmond
4. Montana
5. Villanova
6. Weber State
7. Wm & Mary
8. Cal Poly

However, I prefer his "Pure points" method where the points predict the difference in an upcoming game. Home field advantage is worth 2.52 points. Below are the point differences between #1 JMU and the teams behind it.

1. JMU ------
2. Montana -0.45
3. Richmond -2.62
4. Appy -3.36
5. Villanova -5.27
6. Cal Poly -5.77
7. Weber St -6.24
8. Wm & Mary -7.94

Of course, the Pure Points method is based on "margin of victory". Teams that do a lot of substitution when safely ahead of weaker teams may be rated lower than they should be by this method, but it is probably more accurate than a method that is based purely on Wins and Losses.
th0m
November 9th, 2008, 06:07 AM
Top 8?
Hoyadestroya85
November 9th, 2008, 06:11 AM
I like computer rankings, but at the same time, a good team can have a close game with a crappy team, and the computer rankings are influenced by margin of victory.. that's why the BCS computers love Texas Tech so much this year
but geez... look at how good the CAA is..
GolfingGriz
November 9th, 2008, 06:16 AM
No UNI or SIU...
coover
November 9th, 2008, 06:21 AM
No UNI or SIU...

UNI is rated 15th and SIU is 17th
GolfingGriz
November 9th, 2008, 06:23 AM
UNI is rated 15th and SIU is 17th

So not anywhere near the top 4...(this is a hint)
coover
November 9th, 2008, 06:26 AM
Top 8?

It is a big drop from 8 to 9, Wofford is number 9 and is considered 6.52 points worse than Wm & Mary and 12.29 points worse than JMU. Of course we know what happened to Wofford against Appy.
th0m
November 9th, 2008, 06:36 AM
Ok, hadn't noticed that.

However, #9 in predictor ratings, UNI, has a less significant drop though, rating at -10.2, followed by #10 UNH rating at -11.46
Drblankstare
November 9th, 2008, 06:42 AM
Oh Sagarin, will you ever be right? xrotatehx
JohnStOnge
November 9th, 2008, 07:57 AM
I like computer rankings, but at the same time, a good team can have a close game with a crappy team, and the computer rankings are influenced by margin of victory.. that's why the BCS computers love Texas Tech so much this year
but geez... look at how good the CAA is..

The BCS only uses models that don't use margin of victory. Models that do use margin of victory are, as I understand it, better at both describing past results and predicting future ones. If you don't use margin of victory, the model doesn't "see" any difference between losing to, say, USC by 1 and losing to USC by 60.

Take a look at this: http://www.thepredictiontracker.com/ncaaresults.php

If you read Sagarin's web site you'll see him say that the system he uses for the BCS (ELO chess) is not as good because it doesn't consider margin of victory. And you can see that the prediction tracker page I linked is consistent with that. Sagarin Predictive has gotten 74.4% of games correct while Sagarin ELO Chess has gotten 71% right. And if you go back in history by navigating the main page at http://tbeck.freeshell.org/ you can see that over time ELO chess has been the worst of Sagarin's models.

To me, you can intuitively see that a system that considers margin of victory should be better than one that doesn't. Basically, one that doesn't is ignoring available relevant information. One that looks at margin of victory and tries to get as close to it as possible automatically takes into account who won or loss because a - margin for a team is a loss and a + margin is a win.
Khan4Cats
November 9th, 2008, 09:44 AM
xcoffeex

Whatever. Ignore us now, we're irrelevant. In fact, don't even bother worrying if we sneak into the playoff field, it'll be a pushover. I'm sure there are plenty of coaches lobbying to get matched with UNI in the playoffs because we'll be so easy to deal with.xrolleyesx
uofmman1122
November 9th, 2008, 10:26 AM
xcoffeex

Whatever. Ignore us now, we're irrelevant. In fact, don't even bother worrying if we sneak into the playoff field, it'll be a pushover. I'm sure there are plenty of coaches lobbying to get matched with UNI in the playoffs because we'll be so easy to deal with.xrolleyesxI'd much rather play UNI over Cal Poly or Weber State again. xcoffeex
TheValleyRaider
November 9th, 2008, 10:35 AM
Of course, the Pure Points method is based on "margin of victory". Teams that do a lot of substitution when safely ahead of weaker teams may be rated lower than they should be by this method, but it is probably more accurate than a method that is based purely on Wins and Losses.


The BCS only uses models that don't use margin of victory. Models that do use margin of victory are, as I understand it, better at both describing past results and predicting future ones. If you don't use margin of victory, the model doesn't "see" any difference between losing to, say, USC by 1 and losing to USC by 60.

Take a look at this: http://www.thepredictiontracker.com/ncaaresults.php

If you read Sagarin's web site you'll see him say that the system he uses for the BCS (ELO chess) is not as good because it doesn't consider margin of victory. And you can see that the prediction tracker page I linked is consistent with that. Sagarin Predictive has gotten 74.4% of games correct while Sagarin ELO Chess has gotten 71% right. And if you go back in history by navigating the main page at http://tbeck.freeshell.org/ you can see that over time ELO chess has been the worst of Sagarin's models.

To me, you can intuitively see that a system that considers margin of victory should be better than one that doesn't. Basically, one that doesn't is ignoring available relevant information. One that looks at margin of victory and tries to get as close to it as possible automatically takes into account who won or loss because a - margin for a team is a loss and a + margin is a win.

Isn't it also true that the MOV factor has diminishing returns? That is, a 60 point win really isn't all that much better than a 50 point win, but a 20 point win is much better than a 10 point win?
Pantherpower
November 9th, 2008, 10:51 AM
I'd much rather play UNI over Cal Poly or Weber State again. xcoffeex

You very well may get your wish. However, it appears that UM and UNI will be fighting over who is the 3 and 4 seed provided that they win out. Under that scenario, our potential matchup would occur in Chattanooga. O.K. with you?:)
ASU88
November 9th, 2008, 11:12 AM
It is a big drop from 8 to 9, Wofford is number 9 and is considered 6.52 points worse than Wm & Mary and 12.29 points worse than JMU. Of course we know what happened to Wofford against Appy.
Top 8 also allows a Cal Poly fan to include Cal Poly in both methods. xlolx
uofmman1122
November 9th, 2008, 11:20 AM
You very well may get your wish. However, it appears that UM and UNI will be fighting over who is the 3 and 4 seed provided that they win out. Under that scenario, our potential matchup would occur in Chattanooga. O.K. with you?:)Well, I'm certainly not wishing to get UNI, WSU (impossible, but still), or Cal Poly in the first round. I'd much rather have the luxury most East Coast teams get of playing a cream puff to start the playoffs, but we haven't had such luck in the past 5 or so years.

Here's to hoping the West has a good showing in this year's playoffs. xthumbsupx
uofmman1122
November 9th, 2008, 11:21 AM
Top 8 also allows a Cal Poly fan to include Cal Poly in both methods. xlolxWould you rather he do top 7? xeyebrowx

8 seems like a logical number, being half of the playoff field total number. Granted, 10 or 16 would have been better.
coover
November 9th, 2008, 02:52 PM
Would you rather he do top 7? xeyebrowx

8 seems like a logical number, being half of the playoff field total number. Granted, 10 or 16 would have been better.

I was doing it at 3 o'clock in the morning. I'm only up because of insomnia, and when I saw the big drop off between 8 and 9, I figured 8 was enough. If I was doing it at noon (as I am as I write this), I think 16 or 20 would have been better.