By the way, I just realized that Nebraska hosts BYU Week 1 in 2015. That'll be a fun start to the season.

I'm also curious how Stanford didn't get punished by USC's loss - at least ranking wise.

There's just like so little movement for a week 3 slate. 2 teams in the top 21 moved more than 1 spot. Not sure if there's a leak somewhere in there, but think it definitely suggests we haven't had enough good games yet this season. Can't remember so few ranked vs ranked games this early.

i imagine power rankings don't move a ton based on results, especially close ones

uscs drop is probably significant

We don't know that. BYU has three decisive victories. (Yes, they only beat Houston by 8, but any stats will show that game wasn't nearly that close).

Until teams lose or play really close games, you don't really have any barometer for the upper end of their range. Like for example, we can be relatively certain that USC or Georgia or South Carolina is not the best team in the country - their range is capped, if you will.

Whereas a team like Baylor, it's not like the best team in the history of college football would have any different a resume than them at this point in the season, so we have no idea how good they are. Presumably, if the best team in the country and the 20th best team in the country played the same slate, their results would not be that strange if they were similar. Ditto BYU.

We'll know more soon. BYU plays 5 pretty decent teams in a row.

Sure, if we're only using the information generated directly by game action. However, we know quite a lot more than that about all of the college football teams. Coaching, talent levels, Vegas odds, various models that use more information than just yards, plays, and scores from games

Fair point. There are lots of different attempts to rate teams, and they account for a variety of different inputs, all in different ways. We could probably collect a bunch of different models together in one place, aggregate them, and get a pretty accurate idea of whether Michigan State or BYU is the better football team.

...

Unless our aggregate results had one team 17th and the other 16th, in which case it wouldn't help to settle the debate at all (except to confirm that it's valid to debate it, as there's not a clear and definitive answer).

We failed the eyeball test against Purdoo. But there's no accomplishment based reason to put teams like Baylor or UCLA higher.

Fair enough, fair enough.

just want to say to the thought that's being spent on this thread by BJJ, GB, gb, and others.

yes, week 3 rankings are going to be rough because it really doesn't matter until the end of the season and so much of the current result is based on the preseason preconceived notion of who is better than who and a limited sample size of the games played to date, but any effort to fine tune the early rankings to make them more accurate later in the year is much appreciated.

I mean, BYU is rated 11 points better than UVA

That at least puts them in the same ballpark as MSU, probably. So hard to tell with so few teams playing legit games right now

Oh, BJJ - in the other thread you asked about those histograms. As it is right now, yes all 17 point favorite graphs would look the same. Obviously that's the simplest way, but I'm not srue it's not also close to the right way.

I'm not sure what quality/team nature would impart a need for a significant difference that isn't already communicated by the line on the game. It is more than a shifting up and down a scale, for example shape of the curve in 21 point favorite is going to be flatter and wider (larger variance) than the shape of a the curve in a pick'em game. And the 21 point favorite game is going to be slightly more weighted toward multiples of 7.

A 17-point high-offense, weak defense favorite team would have higher shot at winning large

A 17-point weak offense, amazing defense team is less likely to win by 40

Perhaps ypp could be used to shape the graphs

Anyone calling Texas bad clearly didn't see them almost beat UCLA. They aren't good but they are at worst mediocre.

They aren't bad but UCLA isn't as good as some think, either

does anyone outside of the people who get to vote in polls think UCLA is good at all?

Goofball's model has them 6th behind 2-1 Stanford

I really like that he models this; I just wish it didn't have whatever allows Stanford to be top 5 and USC to be top 10

Does Duke play FSU this year? Where is it?

Will they give them a good game?

They lost at "home" to a shaky UCLA team led by a backup QB who threw flat routes and checkdowns. So read into it what you will. I'm not sure we learned much about either team Saturday night.

ACC Title game in Charlotte in December. Doubtful, but I'm not sure how good FSU is. Might keep it respectable for 2.5 quarters.

Does goof's model factor in the # of opening kickoffs per half?

Lol Texas.

Where does (if anywhere) good ypp data exist?

You must've missed this game

BYU 41 Texas 7


Feels like that model has too much preseason bias. Just from UCLA 6th, Washington 20th based on some shaky results vs supposed sub par comp. and the small gap between two schools who have a massively definitive and clearly not luck based result.

So the way team ratings update in my model is based on a formula I came up with. It's something I really like because it's simple but flexible. There's one constant in the formula that I calibrate to best fit past results, lines, etc.

That constant seems to always coalesce to a very specific value. Just for funsies tonight I dialed that value up to 11. Actually 1, which is the highest it can go. These are the rankings that it spit out.

Those probably look at lot more intuitively satisfying (with one notable exception, cough wookie). But they are worse at "predicting" the past, and also less consistent with this week's 5dimes lines (correlation of lines based on ratings from below = .877, correlation based on my actual rankins = .953).

I wonder if, from a predictive value standpoint, a single games means less than we think.