Sagarin

hawkdrummer1

hawkdrummer1

Well-Known Member
Just for fun, here are the current Sagarin rankings. I've always thought these were pretty good, but some things make no sense.

Mizzou and Wyoming have played one game... against each other. Wyoming won. Mizzou is ranked 36, Wyoming 71. huh?

FYI, IOWA comes in @ 16, Rutgers @ 93, ismoo is 34

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


hawkdrummer1 said:
Just for fun, here are the current Sagarin rankings. I've always thought these were pretty good, but some things make no sense.

Mizzou and Wyoming have played one game... against each other. Wyoming won. Mizzou is ranked 36, Wyoming 71. huh?

FYI, IOWA comes in @ 16, Rutgers @ 93, ismoo is 34

https://www.usatoday.com/sports/ncaaf/sagarin/
Click to expand...

Small sample size will do that. Also Missouri was the road team and had the better stats except for turnovers and rushing. In other words odds are if they don't turn it over 3 more times than Wyoming, they win. Yikes that run defense by Missouri was putrid lol.

upload_2019-9-3_13-53-5.png

https://www.espn.com/college-football/matchup?gameId=401110728
 


JG10Hawk said:
Small sample size will do that. Also Missouri was the road team and had the better stats except for turnovers and rushing. In other words odds are if they don't turn it over 3 more times than Wyoming, they win. Yikes that run defense by Missouri was putrid lol.

View attachment 5543

https://www.espn.com/college-football/matchup?gameId=401110728
Click to expand...

They lost. Period. Shouldn't be 35 spots ahead of Wyoming. Sagarin's poll is supposedly based more on data, than speculation, projections or who voters "expect" to be good.

The small sample size should skew things the other way, with Wyoming ranked unusually high. With both teams moving toward more a reflective assessment as the difference in schedule strength is played out.
 


hawkdrummer1 said:
They lost. Period. Shouldn't be 35 spots ahead of Wyoming. Sagarin's poll is supposedly based more on data, than speculation, projections or who voters "expect" to be good.

The small sample size should skew things the other way, with Wyoming ranked unusually high. With both teams moving toward more a reflective assessment as the difference in schedule strength is played out.
Click to expand...

I suppose it is like in many sports where your stats are more predictive of future success then your record. For example, if 2 basketball teams face off, and one has a better record but a small margin of victory because they have won a lot of close games, and the other lost close games but blew everyone else out so they have a large margin of victory, the large MOV team is more likely to win.

I would guess the computer does not even consider head-to-head in its calculations. Anyone know if that is true?
 


I believe I have heard that Sagarin uses last years data as part of his early seasons rankings. I also have heard that the number of returning starters plays a part in early season rankings.

Can't remember exactly, but I believe by week #6, the old data slowly gets phased out and rankings are only based on this season's data after that.
 


guffus said:
I believe I have heard that Sagarin uses last years data as part of his early seasons rankings. I also have heard that the number of returning starters plays a part in early season rankings.

Can't remember exactly, but I believe by week #6, the old data slowly gets phased out and rankings are only based on this season's data after that.
Click to expand...

Could be...which would make it like most every other poll early in the season. Speculation.
 


guffus said:
I believe I have heard that Sagarin uses last years data as part of his early seasons rankings. I also have heard that the number of returning starters plays a part in early season rankings.

Can't remember exactly, but I believe by week #6, the old data slowly gets phased out and rankings are only based on this season's data after that.
Click to expand...
This is true, each week the previous year’s data has less and less weight.
 


That’s nothing. The coaches poll has Nebraska moving up in the rankings from NR to 25th. Now that is really dumb.

In regards to Sagarin, there are always flukes so it doesn't mean the lower rated team should be ranked higher. Michigan lost to Appalachian St. when they were 5th ranked I think. Doesn't mean Appalachian State would have won more than 1 out of 10.
 
Last edited:


Iowa has rated above their record on Sagarin for a number of years. Not sure what is showing up in their data that causes that.
 


AreWeThereYet said:
Iowa has rated above their record on Sagarin for a number of years. Not sure what is showing up in their data that causes that.
Click to expand...
Close scores against tough competition, largely. It’s my understanding that score differential is the biggest factor in his system, and Iowa typically plays close games against the highly rated teams on its schedule.
 


hawkdrummer1 said:
Could be...which would make it like most every other poll early in the season. Speculation.
Click to expand...
All polls are speculation with a small sample size.

The difference here, though, is that Sagarin’s ratings are completely objective and ignore any and all bias outside of how his formulas are set up. As the season goes on, the ratings system becomes a more accurate representation with each week.

One could argue the criteria used in the rating, but there is no bias whatsoever applied to rating teams because they all have the exact same formulas applied to them from top to bottom based on statistical information.

Polls are precisely that—a bunch of people’s mostly baseless opinions. The AP poll is no better than a figure skating score. Total bullshit.
 




Fryowa said:
It’s a perception problem by media and coaches who make up the poll.
Click to expand...

I'm not sure human polls have a lot of weight in Sagarin's ratings. They use a bunch of weighted algorithms base on performance metrics. I'm not sure how it works or whether it works.
 




Fryowa said:
All polls are speculation with a small sample size.

The difference here, though, is that Sagarin’s ratings are completely objective and ignore any and all bias outside of how his formulas are set up. As the season goes on, the ratings system becomes a more accurate representation with each week.

One could argue the criteria used in the rating, but there is no bias whatsoever applied to rating teams because they all have the exact same formulas applied to them from top to bottom based on statistical information.

Polls are precisely that—a bunch of people’s mostly baseless opinions. The AP poll is no better than a figure skating score. Total bullshit.
Click to expand...

Except as previously pointed out, his formula for the first 6 weeks of the season is fundamentally flawed. It uses assumptions based on last season, not performance from this season.
 


hawkdrummer1 said:
Except as previously pointed out, his formula for the first 6 weeks of the season is fundamentally flawed. It uses assumptions based on last season, not performance from this season.
Click to expand...
It’s not flawed if it uses the exact same formula for everyone. The data becomes better as sample size increases, but that doesn’t make it flawed.

Again, not trying to be argumentative but what’s telling you it’s flawed are your human biases (that we all have to some degree) that subjectively say that Team A should be ahead of Team B.

For example, in your Missouri/Wyoming scenario above, you think the Sagarin model is way out of wack, but it’s using the exact same formulary model on Iowa and none of us take issue with the Hawkeye’s placement. That’s right there is a perfect example of bias taking over.

I know you aren’t necessarily arguing that Sagarin’s system is total garbage, but wanting a better statistical model that works with the smallest possible sample size (1) is wanting to have your cake and eat it too.
 


Fryowa said:
It’s not flawed if it uses the exact same formula for everyone. The data becomes better as sample size increases, but that doesn’t make it flawed.

Again, not trying to be argumentative but what’s telling you it’s flawed are your human biases (that we all have to some degree) that subjectively say that Team A should be ahead of Team B.

For example, in your Missouri/Wyoming scenario above, you think the Sagarin model is way out of wack, but it’s using the exact same formulary model on Iowa and none of us take issue with the Hawkeye’s placement. That’s right there is a perfect example of bias taking over.

I know you aren’t necessarily arguing that Sagarin’s system is total garbage, but wanting a better statistical model that works with the smallest possible sample size (1) is wanting to have your cake and eat it too.
Click to expand...

I'm with you Fry. I like Sagarin's system better than the beauty contest method.

I'm just pointing out that early in the season he deviates from his performance data-based formula. He uses last year's record and returning players to "guesstimate"...similar to the classic polls. The reality is, NO poll is going to be real accurate for a few weeks due to "sample size" (though you can't sample things that haven't happened yet) ...be they subjective or more objective like Sagarin. So he might as well stick with his formula.

Using all of today's available data (not just a sample) Wyoming should be ranked above Mizzou were Sagarin's system pure. One cannot simply assume that Mizzou deserves to be higher, they have to earn it by generating the necessary data.
 


It all depends on how much weight they give to the factors they are considering. Im sure theyve run statistical analysis to figure out which are most likely to produce wins.
At his point so early in the season, what else is there really to go on other than results from last year? You cant just base rankings entirely from one game or the whole thing would be out of whack. There are no "like" opponents to compare to, every game so far has taken place in a vacuum.
 


Ree4 said:
It all depends on how much weight they give to the factors they are considering. Im sure theyve run statistical analysis to figure out which are most likely to produce wins.
At his point so early in the season, what else is there really to go on other than results from last year? You cant just base rankings entirely from one game or the whole thing would be out of whack. There are no "like" opponents to compare to, every game so far has taken place in a vacuum.
Click to expand...

I agree and disagree. You can produce rankings based on what has already happened, and it is 100% real data. (There are "standings" from game 1 in the pros and in conference play, knowing that they will change a lot over time) Which of course points out the absurdity of polls before Oct 1. One game does not a season make (unless you're ISU).

Even as the season goes "like opponents" are mostly in conference, and we all know the flaws of that A beat B and B beat C so A should beat C. The best "assumptions" I think you can make are tougher schedules and conferences overall. Sagarin does a pretty good job of incorporating that as the season progresses.

ANYWAY. I like Sagarin's data-grounded method. Which is why I wish he'd stick to it, even early in the season. If Mizzou loses to Wyoming and Wyoming is #71 by Sagarin's own method (even after beating Mizzou)...then Mizzou sure as hell isn't worthy of #36.
 
Last edited:


It would be fun if Sagarin did release week 1 "fun" rankings based on just 1 weeks worth of data.

Teams like MD would be in the top 5 based on their 79-0 win against Howard.
 


You must log in or register to reply here.


Latest posts

Share this page






Top