2-2 Iowa is the number 6 team in the country by Sagarin ratings

I like Sagarin, but there are times when his formula fails badly. This is one of those times. Penn St 0-4 and ranked #21? Please.
 
I think the Sagarin "rankings" are fundamentally different than how we tend to rank teams as humans. Humans, I suspect, tend to very heavily weight W/L record. There are some very good and legitimate reasons for doing that, but there are also some reasons for it that are not so great. For example, most people - even professionals like journalists who cover CFB for a living - can't watch/analyze every single game. So, shortcuts have to be taken. Like if you saw Ohio State and Akron were both 8-0. Most of us would accept Ohio State being placed at or near the top of the rankings because we have good reason to believe they are probably pretty good. But if Akron was placed at or near the top of the rankings, we'd be understandably skeptical and be asking a lot of questions. In reality, we'd probably find that Ohio State's 8-0 and Akron's 8-0 were a lot different, so our suspicions would be confirmed...but it's possible that something surprising may be going on that runs counter to our suspicions. We'd have to look to be 100% sure.

For Sagarin (and other systems like it), it is simply a stats-based model to create a predictor model that produces output that allows you to compare the predictor score for two arbitrary teams, adjust for home field (if necessary), and get an idea of who "should" be favored. It becomes a "ranking" only in the sense that, when you dump the teams all into a big ass Excel spreadsheet and sort by predictor score, you get something that looks like what we'd call a "ranking". The model factors in W/L as one input, but it's weighted much lower than humans tend to weight it because (1) from a practical standpoint, a computerized model can very easily weigh many different factors across 258 teams without a hint of bias and (2) the model has been built by humans that understand that not all W's and L's are created equal.

I would argue Sagarin "ranking" Penn State at #21 isn't the model "failing"...it's simply the model saying that, based on what it's been programmed to know about how college football works, I think Penn State should be a (neutral field) underdog against teams 1-20 and a (neutral field) favorite against teams 22-258. Put like that, I personally don't think it's a completely outrageous assertion by any means.
 
Luftgekuehlt67 said:
I think the Sagarin "rankings" are fundamentally different than how we tend to rank teams as humans. Humans, I suspect, tend to very heavily weight W/L record. There are some very good and legitimate reasons for doing that, but there are also some reasons for it that are not so great. For example, most people - even professionals like journalists who cover CFB for a living - can't watch/analyze every single game. So, shortcuts have to be taken. Like if you saw Ohio State and Akron were both 8-0. Most of us would accept Ohio State being placed at or near the top of the rankings because we have good reason to believe they are probably pretty good. But if Akron was placed at or near the top of the rankings, we'd be understandably skeptical and be asking a lot of questions. In reality, we'd probably find that Ohio State's 8-0 and Akron's 8-0 were a lot different, so our suspicions would be confirmed...but it's possible that something surprising may be going on that runs counter to our suspicions. We'd have to look to be 100% sure.

For Sagarin (and other systems like it), it is simply a stats-based model to create a predictor model that produces output that allows you to compare the predictor score for two arbitrary teams, adjust for home field (if necessary), and get an idea of who "should" be favored. It becomes a "ranking" only in the sense that, when you dump the teams all into a big ass Excel spreadsheet and sort by predictor score, you get something that looks like what we'd call a "ranking". The model factors in W/L as one input, but it's weighted much lower than humans tend to weight it because (1) from a practical standpoint, a computerized model can very easily weigh many different factors across 258 teams without a hint of bias and (2) the model has been built by humans that understand that not all W's and L's are created equal.

I would argue Sagarin "ranking" Penn State at #21 isn't the model "failing"...it's simply the model saying that, based on what it's been programmed to know about how college football works, I think Penn State should be a (neutral field) underdog against teams 1-20 and a (neutral field) favorite against teams 22-258. Put like that, I personally don't think it's a completely outrageous assertion by any means.
Click to expand...
200.gif
 
I have heard that until a team plays 6 games, Sagarin mixes in data from the last half of the last season, which might explain why this year's rankings seem off.
 
You must log in or register to reply here.

Latest posts

Share this page

Top