Sagarin

[OL FU]

Since we are talking rankings a lot let's put up Sagarin.

1- 67 New Hampshire AA = 68.85 3 0
2- 75 North Dakota State AA = 66.67 3 1
3- 81 Montana State AA = 65.27 2 2
4 - 87 Appalachian State AA = 63.98 3 1
5- 93 Illinois State AA = 62.97 3 1
6- 94 Harvard AA = 62.83 2 0
7- 97 Delaware AA = 62.36 3 0
8- 99 Portland State AA = 61.85 3 1
9- 102 Western Kentucky AA = 61.59 2 1
10- 106 McNeese State AA = 60.89 1 1
11- 107 UC Davis AA = 60.89 2 2
12- 108 Cal Poly-SLO AA = 60.58 3 1
13- 109 Wofford AA = 60.56 2 1
14- 110 Montana AA = 60.54 2 1
15 -111 Villanova AA = 60.13 2 1
16 113 Western Carolina AA = 59.74 2 1
17 114 Massachusetts AA = 59.25 3 1
18- 117 Nicholls State AA = 58.74 1 1
19- 118 Northern Iowa AA = 58.39 2 1
20 -123 Stephen F. Austin AA = 56.44 3 1
21 -125 Georgia Southern AA = 56.41 2 2
22 -126 Princeton AA = 56.34 2 0
23 127 Rhode Island AA = 56.26 3 1
24 128 Southern Illinois AA = 56.22 2 1
25 129 Texas State AA = 56.10 2 1

I understand from most that Sagarin is a pretty good predictor of individual games. But I don't think it is a very good predictor of rankings in I-AA until the season is almost over. This year is even more difficult with games being cancelled. See McNeese at number 10.

A few thoughts :

1. One criticism of polls is where you rank early on does not depend on record. While power polls correct themselves fairly quickly, I think power polls work similarly just reverse it: where you rank depends on where your opponents were ranked early on.
Take Georgia Southern - Lost to Wofford and McNeese both ranked in the top 25.

2. Can someone tell me how much margin of victory enters into this calculation. That would explain Appalachian State. Close win a EKU, but stomping CCU and Citadel by more than his power points would have predicted.

3. Ralph has mentioned D-II games are not considered which would work to Wofford's advantage beating ranked GSU but barely beating D-(something less than I) Georgetown.

Missing

Lehigh
W&M
James Madison

and oh yeah, Furman - Maybe one of the reason I am usually not to fond of Sagarin.

[Dallas Demon]

The main thing I don't like about Sagarin is the initial ratings seem to be skewed in favor of certain conferences/teams in the beginning. This helps immensely for those conferences that schedule a minimum number of OOC games. If all teams are ranked high in a particular conference, even with lots of losses a team can still have a high Sagarin rating as the season goes on.

Conversely, if you schedule an OOC game against a weak opponent and blow them out by 50 points you can still hurt your rating just by playing that opponent. Likewise, if you play an opponent with a high rating and get blown out your rating will likely improve, especially as the season progresses.

[OL FU]

Is there a location where I can find out what MOV is and how it works?

[OL FU]

MOV= Margin of Victory. Each rating system should explain how they figure it. Here's Massey's explanation:
http://www.masseyratings.com/theory/massey.htm#gof

Thanks Ralph

[bluehenbillk]

Sagarin is the classic computer poll that is way off in September, but normally by the end of the season it's fairly accurate.

[HensRock]

Sagarin uses 2 components to arive at the overall Rating.
The "Predictor" component does use MOV.
The "ELO-Chess" component does not. Only this component is used for the BCS because it is more "politically correct" to not use MOV.

The main knock on Sagarin for I-AA is that he does not use any results from games other than Div-I vs. Div-I. So especially in this early part of the season where lots of I-AA teams have played Div-II or even NAIA teams, those games do not enter his calculations. Only results from I-AA vs. other I-AA or vs. I-A are factored in.

Personally, I don't think this is a big deal later in the season, but it is now.

The other thing is that he will change his mathematical model at about week 6. This is when the games (to use his terrminology) are "connected". Meaning that we can get a good feel for the relative strengths of teams because of common opponents and common opponents of common opponents, etc. Early in the season, his model relies on a pre-season ranking. This pre-season bias is dropped about midway through the season.