Some Levity On Rankings...

Fryowa

Administrator
First of all, everyone here knows I think the AP and Coaches polls are complete bullshit, and they are. They are the literal definition of a popularity contest with no standards whatsoever and no consistency. Total...Complete...Utter...Bullshit...

I do however believe in ranking systems that use standard defined criteria and which are the exact same for every team. I tend to follow Sagarin because his ranking system has shown to be a very good predictor of wins/losses, and his SOS rankings have a very small standard deviation throughout the year compared to final results.

What I like best about any ranking system is that there's zero subjectivity. However it's constructed, and however the rules are applied, they're applied to everyone the same. You can argue about the construct of the ranking system (and you should), but you can't argue that opinion has anything to do with the results.

So if one person likes Sagarin, and one person likes the AP poll, and one person likes KenPom, and another likes PRI, what do you do? You look at the Massey composite.

Massey's composite is an average ranking made up of over 100 different ranking systems. The vast majority are mathematical systems, but in the interest of being thorough there are a few popularity contests included like the AP and Coaches. The only thing I don't like is that each index is given equal weight, i.e., the AP poll (total bullshit) is given equal weight in the average to Sagarin (which is the most analyzed ranking out there for NCAA football). But, you can't always have your cake and eat it too.

Anyway, with such a huge sample size of rankings, it's definitely the best one we have. The noise gets smoothed out in a hurry. The website does a great job listing all the rankings for each team, gives the average, as well as the SDev which is huge because it indicates how much consensus there is among the rankings for any given team. For example, Iowa's SDev after this week is 3.87, which indicates that most polls/rankings have Iowa in about the same spot, which increases confidence in that average ranking being "real." Michigan, however, has an SDev of 8.72 which shows you that the ranking systems think UM is either fake news or the best in the country...not confidence inspiring. Either way, with the sample size being over 100, any bad apples are smoothed out easily.

Side note for all you dorks out there who think recruiting rankings mean anything more than diddly shit...If you look at Massey's composite (by far the largest data set), you'll see that the SDev skyrockets after about 10th place and levels out at that high level almost immediately. This alone tells you that any rankings are fairly reliable for most teams up to about 8-10 spots, and after that they become complete horseshit guesses. Just like recruiting rankings, but I digress.

Anyway, here's the link to the Massey composite for any of you who might not have heard about it before, make sure to check out the About page for explanation.

Oh and last thing...and because I really want to stir shit up bad...Iowa is ranked #4 in the nation when averaged over 100 ranking systems and Michigan is ranked #3.
 
First of all, everyone here knows I think the AP and Coaches polls are complete bullshit, and they are. They are the literal definition of a popularity contest with no standards whatsoever and no consistency. Total...Complete...Utter...Bullshit...

I do however believe in ranking systems that use standard defined criteria and which are the exact same for every team. I tend to follow Sagarin because his ranking system has shown to be a very good predictor of wins/losses, and his SOS rankings have a very small standard deviation throughout the year compared to final results.
What I like best about any ranking system is that there's zero subjectivity. However it's constructed, and however the rules are applied they're applied to everyone the same. You can argue about the construct of the ranking system (and you should), but you can't argue that opinion has anything to do with the results.

So if one person likes Sagarin, and one person likes the AP poll, and one person likes KenPom, and another likes PRI, what do you do? You look at Massey composite.

Massey's composite is an average ranking made up of over 100 different ranking systems. The vast majority are mathematical systems, but in the interest of being thorough, there are a few popularity contests included like the AP and Coaches. The only thing I don't like is that each index is given equal weight, i.e., the AP poll (total bullshit) is given equal weight in the average to Sagarin (which is the most analyzed ranking out there for NCAA football). But, you can't always have your cake and eat it too.

Anyway, with such a huge sample size of rankings, it's definitely the best one we have. The noise gets smoothed out in a hurry. The website does a great job listing all the rankings for each team, gives the average, as well as the SDev which is huge because it indicates how much consensus there is among the rankings for any given team. For example, Iowa's SDev after this week is 3.87, which indicates that most polls/rankings have Iowa in about the same spot, which increases confidence in that average ranking being "real." Michigan, however, has an SDev of 8.72 which shows you that the ranking systems either think UM is either fake news or the best in the country...not confidence inspiring. Either way, with the sample size being over 100, any bad apples are smoothed out easily.

Side note for all you dorks out there who think recruiting rankings mean anything more than diddly shit...If you look at Massey's composite (by far the largest data set), you'll see that the SDev skyrockets after about 10th place and levels out at that high level almost immediately. This alone tells you that any rankings are fairly reliable for most teams up to about 8-10 spots, and after that they become complete horseshit guesses. Just like recruiting rankings, but I digress.

Anyway, here's the link to the Massey composite for any of you who might not have heard about it before, make sure to check out the About page for explanation.

Oh and last thing...and because I really want to stir shit up bad...Iowa is ranked #4 in the nation when averaged over 100 ranking systems and Michigan is ranked #3.

Good info, thanks for putting in the effort.
 
First of all, everyone here knows I think the AP and Coaches polls are complete bullshit, and they are. They are the literal definition of a popularity contest with no standards whatsoever and no consistency. Total...Complete...Utter...Bullshit...

I do however believe in ranking systems that use standard defined criteria and which are the exact same for every team. I tend to follow Sagarin because his ranking system has shown to be a very good predictor of wins/losses, and his SOS rankings have a very small standard deviation throughout the year compared to final results.
What I like best about any ranking system is that there's zero subjectivity. However it's constructed, and however the rules are applied they're applied to everyone the same. You can argue about the construct of the ranking system (and you should), but you can't argue that opinion has anything to do with the results.

So if one person likes Sagarin, and one person likes the AP poll, and one person likes KenPom, and another likes PRI, what do you do? You look at Massey composite.

Massey's composite is an average ranking made up of over 100 different ranking systems. The vast majority are mathematical systems, but in the interest of being thorough, there are a few popularity contests included like the AP and Coaches. The only thing I don't like is that each index is given equal weight, i.e., the AP poll (total bullshit) is given equal weight in the average to Sagarin (which is the most analyzed ranking out there for NCAA football). But, you can't always have your cake and eat it too.

Anyway, with such a huge sample size of rankings, it's definitely the best one we have. The noise gets smoothed out in a hurry. The website does a great job listing all the rankings for each team, gives the average, as well as the SDev which is huge because it indicates how much consensus there is among the rankings for any given team. For example, Iowa's SDev after this week is 3.87, which indicates that most polls/rankings have Iowa in about the same spot, which increases confidence in that average ranking being "real." Michigan, however, has an SDev of 8.72 which shows you that the ranking systems either think UM is either fake news or the best in the country...not confidence inspiring. Either way, with the sample size being over 100, any bad apples are smoothed out easily.

Side note for all you dorks out there who think recruiting rankings mean anything more than diddly shit...If you look at Massey's composite (by far the largest data set), you'll see that the SDev skyrockets after about 10th place and levels out at that high level almost immediately. This alone tells you that any rankings are fairly reliable for most teams up to about 8-10 spots, and after that they become complete horseshit guesses. Just like recruiting rankings, but I digress.

Anyway, here's the link to the Massey composite for any of you who might not have heard about it before, make sure to check out the About page for explanation.

Oh and last thing...and because I really want to stir shit up bad...Iowa is ranked #4 in the nation when averaged over 100 ranking systems and Michigan is ranked #3.
where's the link?
 
Oh and last thing...and because I really want to stir shit up bad...Iowa is ranked #4 in the nation when averaged over 100 ranking systems and Michigan is ranked #3.

No link that I can see and too lazy to hunt it down. Is Penn State #2?? Figure they should still come in as the highest B1G team.
 
No link that I can see and too lazy to hunt it down. Is Penn State #2?? Figure they should still come in as the highest B1G team.

The link is located in the text, "here's the link to the Massey composite"

Thanks for the info, Fryowa. I personally LOVE seeing fOSU all the way down to the lowly #14. I hope they stay there (or worse) for months (or years). Their spoiled fan base must be losing their collective mind.
 
No link that I can see and too lazy to hunt it down. Is Penn State #2?? Figure they should still come in as the highest B1G team.
Anyway, here's the link to the Massey composite for any of you who might not have heard about it before, make sure to check out the About page for explanation.
Doesn’t appear to be highlighting like it used to for me either, but if you click on the bold it is in fact a hyperlink. Here it is as well: https://masseyratings.com/cf/compare.htm

OT: The formatting images hadn’t been loading for me for a while when replying (Bold, Italics, Underline, etc.) but those have reappeared, so perhaps that update screwed up hyperlink formatting.
 
Being on the bottom of the list with a narrow standard deviation is really bad news.

You probably noticed that SDevs at the bottom narrow.
 
In a nutshell there isn't much info to go on except UM is suspect. Not sure why the love for Cinc, ND, or UM anywhere.
 
Are we looking at Standard Deviations or a Standard Deviant?

That out of the way, Fry, where would you rank Iowa?
 
Are we looking at Standard Deviations or a Standard Deviant?

That out of the way, Fry, where would you rank Iowa?
Because I trust the objectivity of math, especially when the methods include a gigantic sample size, of course I would rank them fourth.
 
In a nutshell there isn't much info to go on except UM is suspect. Not sure why the love for Cinc, ND, or UM anywhere.
Suspect because you trust your own biased gut, or because you have a rigorously tested mathematical formula that puts them at a lower spot?
 
Doesn’t appear to be highlighting like it used to for me either, but if you click on the bold it is in fact a hyperlink. Here it is as well: https://masseyratings.com/cf/compare.htm

OT: The formatting images hadn’t been loading for me for a while when replying (Bold, Italics, Underline, etc.) but those have reappeared, so perhaps that update screwed up hyperlink formatting.
where's the link?
No link that I can see and too lazy to hunt it down. Is Penn State #2?? Figure they should still come in as the highest B1G team.
Hmm…not sure why the hyperlinks didn’t come through, fellas. Swear on my mother’s name I put ‘em in there. Weird.
 
Suspect because you trust your own biased gut, or because you have a rigorously tested mathematical formula that puts them at a lower spot?
I'm just impressed you understand statistics. Sort of like looking at corn yield trials.

The outcome I feel is fair. The eye test tells us differently. Never think you have KF figured out.
 
Good stuff, Fry.

How people put any stock in polls beyond pure entertainment value is beyond me. Yet, they are still a thing.

I'm a software developer and I also have always really like statistics and models, so I have been a long time enthusiast of these types of ratings and learning about how they are derived at.

I've found that, even today, many people consider them to be some kind of voodoo magic. "Oh, so some COMPUTER decided they were the 6th best team?!?" Yeah, the computer is the least important thing about these sometimes so-called "computer rankings". The computer just crunched the numbers. If you had the time and patience, you could work it all out on scratch paper, the computer just expedites the process.

The models of course are ultimately human-created and are simply trying to make evaluations on the "past performance as a predictor" basis...obviously, not perfect...but a hell of a lot better than what humans can do with just their eyeballs. Jesus H. Christ, there are 130 FBS teams. How many games do you think these pollsters are watching? There aren't enough waking hours.

Models that can be run on a computer though? Yeah, they factor every. Single. Game. And they "watch" the games with only the bias that it's baked into the model itself. They don't care of the uniforms say "Ohio State" or "San Jose State", it's all just numbers to the model.
 
I've noticed over the years that FPI tends to underestimate Iowa's performance and Sagerin overestimates. I had always assumed it was because Iowa style of football was a bit of an outlier. I've never delved into the formulas for this sort of thing and don't think I would want to.
 
Great stuff. I would add a perennial Notre Dame bias in the early major polls. This year though it looks like the domers have it.
 
Top