brought to you by msnbc.com

http://nbcsports.msnbc.com/id/27149214/

PSU is just one of the teams mentioned in this weekend recap article, but we are spoken of very highly. The closing paragraph of our section is one that’s gotta make you chuckle a little:

Picture this: The Nittany Lions win the national championship. Coach Paterno, 82 years old and 59 seasons in at Penn State as an assistant or head coach, enters the office of the athletic director. "So about that contract extension …"

Looking at this logically, and without emotion at this point, Sagarin’s rankings are hilarious. How in the hell do we have one more win, but yet are behind to two teams with lower SOS than us?

We control our own fate by winning. I think we clearly are one of the top teams in the country (Alabama eking by Kentucky is not very impressive, and Texas falling behind 14-3 and giving up 35 points isn’t exactly domination). Let’s just be calm and rational at this point (and that’s bad when I’m the voice of reason and calm; ask my boys). We’re right where we need to be.

First, half the season really isn’t enough time for the computers to get their act together. 6 or 7 games really isn’t a big sample size. Once we hit 10 games or so, the computers will be much more accurate. Most teams still have only 2 or 3 conference games under their belt.

Second, PSU’s SOS isn’t all that great right now. If you follow SEC fan logic, you’d say PSU beat their opponents impressively thereby ruining their record and sending their season into a complete tailspin. (And any losses would be a result of every other team being a world beater in the SEC.) But in the non-SEC world, PSU’s best wins are over three teams that all have 3 losses, though of those 9 collective losses by Illinois, Oregon State and Wisconsin, 3 are to PSU, 1 is to Utah, 1 is to OSU, 1 is to Missouri and 1 is to Minnesota. I think PSU’s SOS will improve as some of their early opponents move into easier portions of their schedule and PSU works its way into its tougher games. But it will not improve enough to match up with the Big 12 or SEC champion.

I am normally knees deep into figuring out where we’ll be ranked, projected SOS, etc (at least I was in the 90s?!). But that said, this year its very simple:

If Alabama and Texas / Oklahoma State win out, then we will not play for the title. Period. The Big 10 is too weak this year. And though it would really, really hurt, I’d have no real complaint this time. The system in place will put us on the outside looking in.

But an undefeated PSU beats out all one loss teams in both of the human polls. Barring a colossal computer gaffe (PSU sub 12 collectively), it won’t matter and we’ll be in the BCS title game.

They’re not “computer rankings,” they’re “statistical rankings.” They don’t run a simulation of the remaining season or anything goofy like that. The ideas behind most of the rankings are hundreds of years old. Except for Billingsley. That guy’s a crackpot.

The reason for the spread in where Penn State’s placed is because they’re undefeated. Teams that are undefeated are basically impossible for those systems to rank well, because they typically work by maximizing the likelihood of each result (i.e. “how bad can I make USC to make the USC loss to Oregon State look more likely, without making the other games that USC’s won too unlikely”).

An undefeated team breaks that, because, obviously, the rating that makes Penn State’s results “most likely” is infinite. So the ratings systems use various different mechanisms to deal with undefeated teams – one way would be to inserting a likelihood distribution for team strengths, and adding that in. So Penn State wouldn’t be “infinitely good” because infinitely good teams don’t come along that often.

That method, however, would basically “floor” Penn State’s rating, because presumedly the prior distribution for team strengths is steep. (Guess which system uses that method).

One other method would be to rescale the ratings, such that the “best” rating isn’t infinite, but, say, “1.0”. Then, you could just average all of the possible results of an undefeated team, and get a result that way. That tends to produce a result that’s much higher, and it’s essentially how Colley’s rankings work.

You could also use a stepwise-convergence method to get to maximum likelihood, which would mean that a team that goes undefeated would “creep towards” an infinite rating, depending on the strength of the opponent they beat. That’s essentially how Sagarin’s ratings work.

Note the teams in Sagarin’s ELO-CHESS ratings. All undefeated at the top.
1: Alabama 6-0
2: Texas 6-0
3: Oklahoma State 6-0
4: Utah 7-0
5: Ball State 7-0
6: Penn State 7-0

Also, the reason that the “schedule strength” doesn’t seem to be right for the order of these teams is that the schedule strength is based on the “combined rating,” whereas the ELO rankings are based solely on ELO results.

Also to be noted is the fact that Sagarin’s PREDICTOR rating is noticeably biased towards high-scoring teams who blow out other teams, as he doesn’t have a decent 2D game-output-function (i.e. a 20-7 win looks the same as a 45-32 win). Hence USC first, Florida second, Penn State third.

Bottom line: Don’t pay attention to the statistical rankings this early in the year. They don’t have enough data yet, and they’re still severely hamstrung by undefeated teams.

Also, Billingsley’s a crackpot. He’s since pulled down the description of his rating system, which is too bad, because it’s hilarious.

I understood the words ‘crackpot’, ‘method’, and ‘hilarious’.

Somehow, though, I feel smarter for just having read that.

Thanks. I think.

It’s too bad Billingsley took down his description of his rankings. Everyone I’ve ever pointed to that description has come away laughing at the guy. It’s completely made up, with absolutely no justification to it.

It had gems like “I believe a strong defense is important, so I give a boost if a teams hold another team to less than a touchdown.” Good to know that that 6-4 loss Penn State had gave Iowa a boost, because that victory is so much more impressive than a 35-10 victory.

We are comfortably in third:
Texas – 0.9902
Alabama – 0.9592
PSU – 0.8849
USC – 0.7576
Texas Tech – 0.7263

I uploaded a spreadsheet with the top 27 teams to here. I included every team in the top 25 of each human poll and the top 15 of each computer poll. It is possible that some team not listed could actually deserve to make the tail end, but I didn’t care.