Sagarins

[Ronbo]

Not sure if this has been posted. Didn't see it.

http://www.usatoday.com/sports/sagarin/fbt09.htm



The FCS Top 25

68 Appalachian State
72 James Madison
75 Richmond
78 Montana
83 Northern Iowa
85 Massachusetts
90 New Hampshire
94 Villanova
96 Wofford
102 Southern Illinois
105 North Dakota State
108 Cal Poly
116 Delaware
117 William & Mary
120 Eastern Washington
121 Harvard
122 Furman
124 Georgia Southern
128 South Dakota State
129 Maine
136 Weber State
137 Yale
138 McNeese State
139 Elon
140 Hofstra

[BearsCountry]

Sargains ratings are pretty worthless overall. I dont care for his basketball ones either.

[UNHWildCats]

MASSACHUSETTS 6th? LMAO

[VT Wildcat Fan53]

Not sure if this has been posted. Didn't see it.

http://www.usatoday.com/sports/sagarin/fbt09.htm



The FCS Top 25

68 Appalachian State
72 James Madison
75 Richmond
78 Montana
83 Northern Iowa
85 Massachusetts
90 New Hampshire
94 Villanova
96 Wofford
102 Southern Illinois
105 North Dakota State
108 Cal Poly
116 Delaware
117 William & Mary
120 Eastern Washington
121 Harvard
122 Furman
124 Georgia Southern
128 South Dakota State
129 Maine
136 Weber State
137 Yale
138 McNeese State
139 Elon
140 Hofstra

ZooMass way too high. UNH probably too high by 3 spots to start the season. I'd move up both 'Nova and Wofford, but watch out if UNH finds a way to beat Ball State....

[BigHouseClosedEnd]

ZooMass way too high. UNH probably too high by 3 spots to start the season. I'd move up both 'Nova and Wofford, but watch out if UNH finds a way to beat Ball State....

How can you say teams are 'too high' or 'too low'?

Isn't this just a computer rating? At this point, the computer rating is based on littler more than strength of schedule, no?

[TexasTerror]

Hmmm...bottomfeeders.

238 Butler
239 NC A&T
240 Texas Southern
241 Valpo
242 St Francis
243 Savannah State
244 Old Dominion
245 Campbell

[mtgrizfan4life]

I think Weber State would be a little higher. They have a solid schedule compared to some above them.

[coover]

Pre-season polls are worth about as much as the paper ... er, the webpage they were displayed on, one second after it is edited .

[uofmman1122]

Pre-season polls are worth about as much as the paper ... er, the webpage they were displayed on, one second after it is edited .Hey, at least it gives us something to talk about. xnodx

[JohnStOnge]

How can you say teams are 'too high' or 'too low'?

Isn't this just a computer rating? At this point, the computer rating is based on littler more than strength of schedule, no?

At this point it is "bayesian," which means the subjective "belief" of Dr. Sagarin is a factor. The "belief" might be "objective" in a sense in that he may have some kind of algorithm based on the past performances of the programs that he uses. As the season goes on, the ratings will be based completely on what happened this year.

[JohnStOnge]

Sargains ratings are pretty worthless overall. I dont care for his basketball ones either.

To me, in order to judge something like this one way or another, you have to find some way of actually assessing its accuracy. I've always done that by looking at how accurate a power rating system is in predicting the outcomes of games both in terms of who wins and who loses and what the point spread is. In that regard, the Sagarin system is very close to the final Vegas line in terms of performance.

To me, the final Vegas line is a "consensus" opinion of a bunch of people who at least think they know about football. Vegas is trying to create a situation in which equal numbers of people bet on either side. These are people who are looking at all kinds of stuff: Injury reports, weather, matchups, etc. Yet Sagarin's system, even at the beginning of the season, consistently maintains performance comparable to that of the Vegas line without considering any of that.

You can go to http://www.thepredictiontracker.com/ncaaresults.php and see the performances for the 2008 season. The favorite by the Vegas line won 73.8 percent of the time, the favorite by both the Sagarin Predictive won 73.7 percent of the time, and the favorite by Sagarin's "regular" system also won 73.7 percent of the time. The line was off on the absolute value of the point spread by an average of 12.6 points, Sagarin Predictive was off by an average of 13.2 points, and Sagarin's "regular" system was off by an average of 13.4 points.

[JohnStOnge]

I'll mention that when we had this kind of discussion on this board a few years ago someone said they thought Sagarin's system isn't as accurate for I-AA/FBS schools. The site I linked above followed I-AA for a few years. The results show that there may be some truth to that but, if it's the case, the difference is not large.

Over the three year period during which entire seasons were considered the site author looked at 1743 I-AA games and 2032 I-A games. Sagarin's "regular" model picked the winner correctly for 71.3 percent of the I-AA games and 73.7 percent of the I-A games. If you compare the two, the difference is "significant" at the 91 percent confidence level (p = 0.09). The most widely used convention is to conclude that there is a difference when the confidence level reaches 95 percent (p = 0.05). So if someone were writing a paper or something they'd likely say there is not sufficient evidence to conclude that Sagarin's system is any better at picking winners for I-A games than it is at picking winners for I-AA games. As a practical matter, though, if you were betting you'd do well to bet that it is slightly better at picking winners of I-A (now FBS game). But the emphasis should be on slightly. Very slightly.

[bigblue78]

Hmmm...bottomfeeders.

238 Butler
239 NC A&T
240 Texas Southern
241 Valpo
242 St Francis
243 Savannah State
244 Old Dominion
245 Campbell

That's got to be embarassing for Campbell to be ranked behind a team that hasn't won a football game in almost 70 years...I can also see what our AD used as a guide for putting our schedule together this year!

[dgreco]

does anyone know the site that consolidates all the computer polls? Its no longer a bookmark and I cannot remember it.

[Ronbo]

Sticky: 2nd thread on the FCS discussion page.

[dgreco]

Sticky: 2nd thread on the FCS discussion page.

I mean the computer rankings, it is something like 20 or so computer rankings.

[UNHWildCats]

Hey, at least it gives us something to talk about. xnodx
if you'd post pictures we could talk about the guys of Tokyo xsmiley_wix

[UNHWildCats]

does anyone know the site that consolidates all the computer polls? Its no longer a bookmark and I cannot remember it.
http://www.masseyratings.com/cf/compare1aa.htm

[19Duke97]

Ok not smack at all, but how can ODU be ranked ahead (much less at all) with a team that has NEVER played a down of college, much less Division 1 football?

[RabidRabbit]

Ok not smack at all, but how can ODU be ranked ahead (much less at all) with a team that has NEVER played a down of college, much less Division 1 football?

Isn't ODU playing schollie ball and Campbell is one year ahead, but non-schollie? That could explain the bottom rung for the Camels.

[dgreco]

http://www.masseyratings.com/cf/compare1aa.htm

ty

[89Hen]

As the season goes on, the ratings will be based completely on what happened this year.
Not true John and you know it. xnonox

[89Hen]

the favorite by both the Sagarin Predictive won 73.7 percent of the time, and the favorite by Sagarin's "regular" system also won 73.7 percent of the time.
So Sagarin wouldn't be in the top 10 in the gohens.net pool. :p

[WileECoyote06]

According to his ratings, App. State should upset ECU

68 Appalachian State
73 East Carolina

[UNHFan99]

I actually enjoy the sagarin rating because it is normally holds a good argument against FCS haters.

[Minuteman87]

MASSACHUSETTS 6th? LMAO

Geez, the season hasn't even started yet and you're already obsessing about UMass being one spot above you in a meaningless ranking? Well, I guess inferiority complexes are a way of life at the University of No Hardware.

[JohnStOnge]

So Sagarin wouldn't be in the top 10 in the gohens.net pool. :p

I don't know anything about the gohens.net pool. Does it involve picking all games that occur or subsets of them?

[JohnStOnge]

Not true John and you know it. xnonox

If it's not true I certainly don't know it. I interpret what's written as his site to mean that once enough information is available he transitions to a model based entirely on what's happened during a given season. I'll admit that the way he describes the situation is confusing. However, I think you can tell what's going on by comparing what he says at the beginning of a given season to what he says at the end.

For instance, if you look at the top of the current 2009 ratings description you'll see the statement, "RATING and PREDICTOR are now bayesian and the ELO-CHESS is bayesian now."

If you look at the top of any of the season ending ratings in his archives, you'll see this statement in that spot:

"RATING and PREDICTOR are now unbiased and the ELO-CHESS is UNBIASED now."

I normally don't use Wikipedia but I'll go ahead and use the following quote from the Wikipedia article at http://en.wikipedia.org/wiki/Bayesian_inference because I do believe it puts what "bayesian" means in a nutshell:

"Bayesian inference uses a numerical estimate of the degree of belief in a hypothesis before evidence has been observed and calculates a numerical estimate of the degree of belief in the hypothesis after evidence has been observed."

I interpret what's written at his site to mean that he starts off with ratings that are not based on any "evidence" from the current season. As the season progresses, the initial ratings play less and less a role until a point is reached where they play no role at all and the ratings are based entirely on what's happened during the current season.

[JohnStOnge]

I don't know anything about the gohens.net pool. Does it involve picking all games that occur or subsets of them?

Never mind. I went to gohens.net and looked at it. It only involves picking games that are going to be played by teams in the AGS top 25. My belief is that if you do that you're likely to get a higher percentage correct than if you try to pick every single game. If you think about it I think you can see reasons for that.