Sagarin pre-season ratings

[WTFCollegefootballfan]

http://www.usatoday.com/sports/ncaaf/sagarin/

[Twentysix]

Sagarin predicts the FCS opener spread at 27.05 points?

[WTFCollegefootballfan]

Sagarin predicts the FCS opener spread at 27.05 points?

I guess you get an extra 2.61 points for home field advantage.

NDSU is ranked higher than 4 Big ten west teams, and 4 SEC teams. NDSU is at 44.

[RootinFerDukes]

I guess you get an extra 2.61 points for home field advantage.

NDSU is ranked higher than 4 Big ten west teams, and 4 SEC teams. NDSU is at 44.

I believe it too. You all could beat the lower 1/3 of any given P5 conference and have competitive games against the middle 1/3. That's why they won't schedule you, they would rather avoid a good team who will embarrass them.

[caribbeanhen]

I guess you get an extra 2.61 points for home field advantage.

NDSU is ranked higher than 4 Big ten west teams, and 4 SEC teams. NDSU is at 44.


hint hint.....

[Cocky]

How does a computer come up with a ranking before any games have been played? Does it have some human influence? If so then isnt it just another poll?

I ask this in another thread, does anyone have an answer?

[VikingPSU]

mix with hot water, same as nescafe coffee

[BisonTru]

I ask this in another thread, does anyone have an answer?

I would assume most of them are a combination of last years results and what they have returning turned into numerical values.

[bison137]

I would assume most of them are a combination of last years results and what they have returning turned into numerical values.



No. Sagarin makes no attempt to analyze returning players. It is simply a weighted average of the last few years. No meaning at all for 2016.

[Cocky]

So its snake oil? What I thought but just making sure.

[The Yo Show]

Yep snake oil, but we here at AGS sell only extra virgin olive oil. xlolx

[thebootfitter]

No. Sagarin makes no attempt to analyze returning players. It is simply a weighted average of the last few years. No meaning at all for 2016.
To say it has zero meaning on the current year might be stretching just a bit far, but certainly it has minimal value in predicting the relative strength of teams for the current season until the teams are "well connected" for the current year.

A lot can change from one year to the next, but a lot can also change from one game to the next and even one minute of a game to the next with a key injury or a good/bad bounce. However, when looking at statistical relevance, I'd be surprised if there's not at least some small amount of skill in the predictions that are derived from data from the most recent seasons. There's generally at least a few players and coaches who remain where they are one year to the next.

[Cocky]

To say it has zero meaning on the current year might be stretching just a bit far, but certainly it has minimal value in predicting the relative strength of teams for the current season until the teams are "well connected" for the current year.

A lot can change from one year to the next, but a lot can also change from one game to the next and even one minute of a game to the next with a key injury or a good/bad bounce. However, when looking at statistical relevance, I'd be surprised if there's not at least some small amount of skill in the predictions that are derived from data from the most recent seasons. There's generally at least a few players and coaches who remain where they are one year to the next.

Does he go back and adjust the original base values? If not the poll is tainted for the entire period of use.
The poll is fun but its just another poll but hiding under a computer chip.

[Hammersmith]

Does he go back and adjust the original base values? If not the poll is tainted for the entire period of use.
The poll is fun but its just another poll but hiding under a computer chip.

First, it's not a poll, it's a rating. Polls, ratings and rankings are three different things.

Sagarin starts the year based on past performance. Exactly how he does that is unknown. I would guess that most of it is based on the previous year, with less on the year before that, and even less as you go back. For example, the 2016 starting values could be based on 2012(12.5%), 2013(12.5%), 2014(25%), and 2015(50%). Or maybe 2012(10%), 2013(20%), 2014(30%), and 2015(40%).

As the season goes on, the model relies less and less on the previous seasons, and more and more on the current season. Eventually the model reaches "well-connected" status and is completely divorced from the prior seasons. At that point, only the information from the current season is used. Being well-connected depends on the schedule, but it normally happens after weeks 4, 5 or 6. Sagarin used to tell us when the changeover happened, but he stopped doing it a year or so ago for some reason.


What's the difference between a poll, rating and ranking? A poll is based on the opinions of a group of people. One person does not make a poll. A rating is a statistical model looking at a group of teams with minimal bias. There's some inherent bias in what the model looks at and how it weighs that info, but once the model is set up, it's not touched. A ranking is one person's opinion. It may be objective or subjective. A ranking of 2015 FCS attendance is objective; there's no bias involved with that. A ranking of the best 2015 FCS quarterbacks is subjective; there's going to be a lot of opinion in that ranking. You could say that a poll is a collection of individual rankings.

[Cocky]

First, it's not a poll, it's a rating. Polls, ratings and rankings are three different things.

Sagarin starts the year based on past performance. Exactly how he does that is unknown. I would guess that most of it is based on the previous year, with less on the year before that, and even less as you go back. For example, the 2016 starting values could be based on 2012(12.5%), 2013(12.5%), 2014(25%), and 2015(50%). Or maybe 2012(10%), 2013(20%), 2014(30%), and 2015(40%).

As the season goes on, the model relies less and less on the previous seasons, and more and more on the current season. Eventually the model reaches "well-connected" status and is completely divorced from the prior seasons. At that point, only the information from the current season is used. Being well-connected depends on the schedule, but it normally happens after weeks 4, 5 or 6. Sagarin used to tell us when the changeover happened, but he stopped doing it a year or so ago for some reason.


What's the difference between a poll, rating and ranking? A poll is based on the opinions of a group of people. One person does not make a poll. A rating is a statistical model looking at a group of teams with minimal bias. There's some inherent bias in what the model looks at and how it weighs that info, but once the model is set up, it's not touched. A ranking is one person's opinion. It may be objective or subjective. A ranking of 2015 FCS attendance is objective; there's no bias involved with that. A ranking of the best 2015 FCS quarterbacks is subjective; there's going to be a lot of opinion in that ranking. You could say that a poll is a collection of individual rankings.

Ok then its a rating ranking. This years teams havent played and were not last years teams. Like I said the rating ranking is fun but it is what it is, one personal opinion hiding under a computer chip. Doesnt mean it is as acurate as others but it not a computer only system.

[thebootfitter]

Ok then its a rating ranking. This years teams havent played and were not last years teams. Like I said the rating ranking is fun but it is what it is, one personal opinion hiding under a computer chip. Doesnt mean it is as acurate as others but it not a computer only system.
Do some research. Nothing even close to a "personal opinion." The only way personal opinion has a play is in determining which metrics are used in the calculations and how much weight they are given. Once the formulas are set up, they are what they are. Completely objective and not swayed by personal opinion in any way.

But you are correct that it is fun, and especially in the early season shouldn't carry much weight. But it can still be used to correctly predict the outcome of roughly 75% of games -- even in the early season using data from previous years. So it is clearly not worthless.

[Daytripper]

I like Sammy coming in at # 91.

[UNIFanSince1983]

Do some research. Nothing even close to a "personal opinion." The only way personal opinion has a play is in determining which metrics are used in the calculations and how much weight they are given. Once the formulas are set up, they are what they are. Completely objective and not swayed by personal opinion in any way.

But you are correct that it is fun, and especially in the early season shouldn't carry much weight. But it can still be used to correctly predict the outcome of roughly 75% of games -- even in the early season using data from previous years. So it is clearly not worthless.

This. Also it is a little misleading that these are called computer ratings. They are mathematical ratings. The only thing the computer does is compute the formulas that the mathematician entered into the computer. So nothing is hiding under a computer chip.

[Grizalltheway]

So its snake oil? What I thought but just making sure.

Yes but it leads NDSU fans to believe they could compete in the SEC so they like it.

[Hammersmith]

Yes but it leads NDSU fans to believe they could compete in the SEC so they like it.

???

Bit of hyperbole there since Sagarin says we'd fare worst in the SEC.

Sagarin puts us near the bottom of the SEC. If you were to add NDSU to the SEC, we would be 11th of 15. We'd be around the bottom third mark of the SEC and Pac-12, and mid-pack in the ACC, Big 12 and Big 10. It predicts we'd be very competitive in the G5. Second in the AAC and MWC, and top in the rest.

SEC - 11th of 15 (33rd percentile)
Pac 12 - 9th of 13 (38th percentile)
ACC - 10th of 16 (44th percentile)
Big 12 - 7th of 11 (45th percentile)
Big 10 - 8th of 15 (53rd percentile)