PDA

View Full Version : Massey Margin of Victory Ratings



JohnStOnge
October 6th, 2008, 02:36 PM
Since ratings are being posted, here are the top 25 rated FCS teams by the Massey model that considers margin of victory. The link is http://www.masseyratings.com/rate.php?lg=cf . Click on "MOV" to sort by Margin of Victory Ratings. If you want to see how it turns out in the ratings that don't consider margin of victory, you can look at it as it comes up. Or you can click on the "Ratings" column and sort it that way if it's not already sorted.

1. James Madison
2. New Hampshire
3. McNeese State
4. Appalachian State
5. Villanova
6. Cal Poly SLO
7. Massachusetts
8. Richmond
9. Montana
10. Weber State
11. Northern Arizona
12. Wofford
13. Northern Iowa
14. Montana State
15. Jacksonville State
16. Elon
17. Western Illinois
18. Furman
19. Delaware
20. Central Arkansas
21. South Dakota State
22. Northeastern
23. Southern Illinois
24. Eastern Washington
25. North Dakota

89Hen
October 6th, 2008, 02:38 PM
Since ratings are being posted, here are the top 25 rated FCS teams by the Massey model that considers margin of victory. The link is http://www.masseyratings.com/rate.php?lg=cf . Click on "MOV" to sort by Margin of Victory Ratings. If you want to see how it turns out in the ratings that don't consider margin of victory, you can look at it as it comes up. Or you can click on the "Ratings" column and sort it that way if it's not already sorted.

1. James Madison
2. New Hampshire
3. McNeese State
4. Appalachian State
5. Villanova
6. Cal Poly SLO
7. Massachusetts
8. Richmond
9. Montana
10. Weber State
11. Northern Arizona
12. Wofford
13. Northern Iowa
14. Montana State
15. Jacksonville State
16. Elon
17. Western Illinois
18. Furman
19. Delaware
20. Central Arkansas
21. South Dakota State
22. Northeastern
23. Southern Illinois
24. Eastern Washington
25. North Dakota
No Bison? xconfusedx xconfusedx xeyebrowx

danefan
October 6th, 2008, 02:39 PM
No Bison? xconfusedx xconfusedx xeyebrowx
#27

89Hen
October 6th, 2008, 02:40 PM
#27
Not much else needs to be said about computer rankings then, does it. xcoffeex

Syntax Error
October 6th, 2008, 02:42 PM
Not much else needs to be said about computer rankings then, does it.other than they are objective..... xcoffeex

danefan
October 6th, 2008, 02:44 PM
other than they are objective..... xcoffeex

Oh boy....here it comes. Especially with the GPI coming out today.

Computer rankings are only as objective as the invidual who writes the code/formula.

If that person is biased, his/her code will be biased and as such the computer ranking will be biased.

JohnStOnge
October 6th, 2008, 02:58 PM
Oh boy....here it comes. Especially with the GPI coming out today.

Computer rankings are only as objective as the invidual who writes the code/formula.

If that person is biased, his/her code will be biased and as such the computer ranking will be biased.

I suppose you could say that a model like that is "biased" in that whoever does it has some idea as to what he thinks is a good way to try to describe, predict, and or explain results. But once that's established, there is no bias. And the mathematical bias can be measured. Usually, in that regard, these models do a good job.

JohnStOnge
October 6th, 2008, 03:03 PM
On North Dakota State:

There's really not a whole lot of difference between where Sagarin's and Massey's MOV system have the Bison. Sagarin has them just inside the top 25 and Massey MOV has them just outside of it.

North Dakota State got a good win Saturday but Southern Illinois is also considered a borderline top 25 team. And other than Saturday they've beaten Austin Peay and Central Connecticut State. They also lost to Youngstown State...which continues to look really bad after losing to Southern Utah this past weekend...and a bad FBS team (Wyoming).

To me it's not unreasonable at all to have them on the borderline between being in or out of the top 25 at this point.

danefan
October 6th, 2008, 03:04 PM
I suppose you could say that a model like that is "biased" in that whoever does it has some idea as to what he thinks is a good way to try to describe, predict, and or explain results. But once that's established, there is no bias. And the mathematical bias can be measured. Usually, in that regard, these models do a good job.

I agree. Its a different kind of bias then the "homerism" and sandbagging that occurs in human polls. But it is still a bias to me. A bias towards specific statistical indicators. And if its biased, it can't be objective.

89Hen
October 6th, 2008, 03:06 PM
other than they are objective..... xcoffeex
Throwing darts to select the Top 25 would be objective too. That doesn't make it a good way to do it. xpeacex

alexale23
October 6th, 2008, 03:07 PM
well they got 1 thing right JMU #1

CrazyCat
October 6th, 2008, 03:08 PM
I like this Massey guy :D

I Bleed Purple
October 6th, 2008, 03:11 PM
Don't like Margin of Victory Ratings unless they take into account blowout scores at the end of a game. 37-23 looks a lot different when two different games were 30-23 and 37-6 with five minutes to go.

gt_bison
October 6th, 2008, 03:14 PM
Part of the problem with early season computer rankings (just like polls) is that there's not enough data to use. Polls suffer from the pre-season poll bias that keeps teams flying higher than they probably should be. (I'm willing to admit the coaches poll has kept the Bison higher than they've proven themselves worthy of this season.) Computer rankings, while objective, suffer from just not enough data. There's one FBS ranking (http://football.kislanko.com/2008/ISOV.html) that has USC at 1, Penn State at 2, and Georgia Tech at 3. GT isn't ranked in any of the major human polls, although Sagarin has Tech in the low teens in his ranking. What's wonky with the ranking scheme I linked? It uses margin of victory, and Tech has three wins by large margins and will get a fourth this weekend against GW, in all likelihood. Once the computers have eight or so games and lots of interconnection to use, they tend to sort things out in a way that seems more reasonable to humans.

james_lawfirm
October 6th, 2008, 03:27 PM
[QUOTE=JohnStOnge;1138152]Since ratings are being posted, here are the top 25 rated FCS teams by the Massey model that considers margin of victory. The link is http://www.masseyratings.com/rate.php?lg=cf . Click on "MOV" to sort by Margin of Victory Ratings. If you want to see how it turns out in the ratings that don't consider margin of victory, you can look at it as it comes up. Or you can click on the "Ratings" column and sort it that way if it's not already sorted.


OK, I need some 'splainin' here. On the Massey webpage, it says margin of victory is not used. I don't know what Massey does use, but apparently not margin of victory. I would not think MOV is a good predictor anyway.

danefan
October 6th, 2008, 03:31 PM
YOu have to click the MOV column to get the MOV ratings to be used.

JohnStOnge
October 6th, 2008, 03:33 PM
Don't like Margin of Victory Ratings unless they take into account blowout scores at the end of a game. 37-23 looks a lot different when two different games were 30-23 and 37-6 with five minutes to go.

I've looked into it before and margin of victory ratings do a better job of predicting outcomes as well as in explaining past results. They're better. Like you can check out the site http://www.thepredictiontracker.com/ncaaresults.php and compare Sagarin's systems. The one he says is best on his site because it uses nothing but margin of victory is "Sagarin Predictive." The one he says is worse is the "Sagarin ElO" that doesn't consider margin of victory at all.

As you can see the "pure points" model has a slightly better percent correct (75% to 72%), a bias closer to 0 (0.64 to 1.14), and lower mean squared error (311 to 342). Another thing you can do is look at what you get with Massey's alternative that does not consider margin of victory. I don't know about you but I think the MOV ratings look more plausible overall than this does:

1. James Madison
2. McNeese State
3. Cornell
4. Montana
5. Wofford
6. Elon
7. New Hampshire
8. Villanova
9. Weber State
10. Jacksonville State
11. Cal Poly
12. Massachusetts
13. Richmond
14. Furman
15. Liberty
16. Appalachian State
17. Northern Arizona
18. Southern Illinois
19. Georgia Southern
20. Northern Iowa
21. Northwestern State
22. South Dakota State
23. Central Arkansas
24. Sam Houston State
25. Hofstra

JohnStOnge
October 6th, 2008, 03:37 PM
I personally like the average of ratings and polls at http://www.mratings.com/cf/compare1aa.htm . To me it looks very reasonable. For example: Though I'm a McNeese fan, I think having the Cowboys at #6 is more plausible than having them in the top 3 at this point. There's a lot of variation in the systems but I think the average comes out looking pretty good.

ToTheLeft
October 6th, 2008, 03:41 PM
I do not approve of a poll that uses MOV....

(this may or may not be because we beat Western Carolina and Youngstown by 6 combined points)

dbackjon
October 6th, 2008, 03:43 PM
I do not approve of a poll that uses MOV....

(this may or may not be because we beat Western Carolina and Youngstown by 6 combined points)


Or all three FCS wins by a total of 11 points :p

JohnStOnge
October 6th, 2008, 03:44 PM
Here's another example using FBS teams. Take a look at what Massey got when he used and didn't use Margin of Victory to rank teams (top 10):

Using Margin of Victory

1) Oklahoma
2) Penn State
3) Texas
4) USC
5) Missouri
6) Alabama
7) Texas Tech
8) LSU
9) Florida
10) Georgia

Not Using Margin of Victory

1) Vanderbilt
2) Alabama
3) Missouri
4) Utah
5) Oklahoma
6) Northwestern
7) Penn State
8) Texas
9) Virginia Tech
10) Georgia Tech

Now, which looks more plausible?

ToTheLeft
October 6th, 2008, 03:47 PM
Clearly the second one. Everyone knows that the 'Dores are the best thing in FBS right now!

ToTheLeft
October 6th, 2008, 03:48 PM
Or all three FCS wins by a total of 11 points :p

What, I couldn't hear you, there's a really loud East Coast Bias in between us. :)

james_lawfirm
October 6th, 2008, 03:48 PM
YOu have to click the MOV column to get the MOV ratings to be used.


Thanks, Danefan. Predicting a game based of MOV seems akin to driving your car looking solely through the rear view mirror.

And, what if your Coach puts in 3rd & 4th stringers anytime the differential is +28?? Doesn't that skew everything?

BDKJMU
October 6th, 2008, 03:51 PM
Throwing darts to select the Top 25 would be objective too. That doesn't make it a good way to do it. xpeacex

Neither does the heavily biased human polls, which I believe are no better in general than computer rankings.

I think the best way to do it is have a blend of the human and computer polls, therefore hopefully all the biases cancel each other out.

danefan
October 6th, 2008, 03:58 PM
Neither does the heavily biased human polls, which I believe are no better in general than computer rankings.

I think the best way to do it is have a blend of the human and computer polls, therefore hopefully all the biases cancel each other out.

Or amplify each other......xcoffeex xcoffeex xcoffeex

We'll see that blended ranking very shortly when the GPI comes out.

JohnStOnge
October 6th, 2008, 04:36 PM
Ok guys. I'm going to post another one. Top 10 in all Division I by Sagarin according to his standard ratings using margin of victory and his ELO ratings that don't use margin of victory:

Using Margin of Victory

1. USC
2. Oklahoma
3. Alabama
4. Penn State
5. Texas
6. Missouri
7. Florida
8. Georgia
9. Boise State
10. Texas Tech

Not Using Margin of Victory

1. Alabama
2. Northwestern
3. Missouri
4. Vanderbilt
5. Utah
6. Oklahoma
7. Ball State
8. Penn State
9. Georgia
10. Texas

C'mon fellas. Surely you're starting to get the picture! Using margin of victory may not be perfect. Like I don't think there's any way Boise State is a top 10 team. But NOT using Margin of victory is worse.

89Hen
October 7th, 2008, 09:28 AM
Part of the problem with early season computer rankings (just like polls) is that there's not enough data to use.
Therein lies the problem. There is NEVER enough data for them to be accurate. 11 or 12 games per year with only 3 or 4 games OOC. Any computer person (or scientist for that matter) will tell you that there simply isn't enough data to make these accurate.

89Hen
October 7th, 2008, 09:30 AM
Neither does the heavily biased human polls, which I believe are no better in general than computer rankings.

I think the best way to do it is have a blend of the human and computer polls, therefore hopefully all the biases cancel each other out.
Garbage in, garbage out. I've challenged JohnStOnge for four years to a comptetion of computer vs human, but he continues to dodge me. Most people in the gohens Top 25 competition are running around 80% correct on games this year. Computers would be hard pressed to match them. xpeacex