NCAA 2009

[Berkut]

Sagarin ratings, which is, as far as I know, the only ranking that actually rank conferences in any kind of objective manner.

Computer rankings aren't allowed to consider the scores, which makes them highly suspect imo.

How so? Score is often a crappy measure of a teams quality, since it just gives even more credit to teams for playing shitty opponents.

[CountDeMoney]

For those(paging Dr. Money) who like the numbers behind the standings:

I'm glad you posted that, because I was going to yell at you.

And neither of those teams would get out of the Pac-10 without at least one loss, probably two.

You're talking about Florida and Alabama, right? 
Just making sure that you are, in fact, making the claim that Florida and Alabama, if they were in the PAC-10 would have 1 if not 2 losses.  Correct?  You're actually saying that, right?

Yep.

That's it.  You're going straight to Occupational Health first thing tomorrow AM.  You have a little cup to fill.  With urine this time.

[Eddie Teach]

How so? Score is often a crappy measure of a teams quality, since it just gives even more credit to teams for playing shitty opponents.

So a team that beats a mediocre opponent by a field goal gets a boost over a team that trounces a crap opponent by 50? Sounds like you're basically giving double weight to strength of schedule.

Humans consider both the score and the quality of the opponent when evaluating the outcome of a game. Computers aren't allowed that luxury.

[sbr]

Not at all irrelevant, since seeding will be based on the outcome of that game.

The loser would likely still be a top 6 seed, so maybe not completely irrelevant but mostly so.

[sbr]

For those(paging Dr. Money) who like the numbers behind the standings:

I'm glad you posted that, because I was going to yell at you.

And neither of those teams would get out of the Pac-10 without at least one loss, probably two.

You're talking about Florida and Alabama, right? 
Just making sure that you are, in fact, making the claim that Florida and Alabama, if they were in the PAC-10 would have 1 if not 2 losses.  Correct?  You're actually saying that, right?

Seems reasonable to me.

[Eddie Teach]

Yeah but you probably think ranking Oregon ahead of Boise is reasonable too(as does Sagarin and by extension, Berkut) 

[sbr]

Yeah but you probably think ranking Oregon ahead of Boise is reasonable too(as does Sagarin and by extension, Berkut) 

Well not now, obviously; but had Oregon won out I think it would have been reasonable, but I would understand anyone else who disagreed with me.

I am curious how USC is ahead of Oregon now though. 

[Berkut]

Yeah but you probably think ranking Oregon ahead of Boise is reasonable too(as does Sagarin and by extension, Berkut) 

If someone asked me to do a poll, I would not rank Oregon over Boise St.

If someone asked me to come up with a computerized and objective algorithm for ranking teams, it would not at all surprise if said ranking had Oregon over Boise St., at least up until last week.

[Berkut]

How so? Score is often a crappy measure of a teams quality, since it just gives even more credit to teams for playing shitty opponents.

So a team that beats a mediocre opponent by a field goal gets a boost over a team that trounces a crap opponent by 50? Sounds like you're basically giving double weight to strength of schedule.

Not at all - beating a average or mediocre team by a field goal SHOULD count for more than beating some patsy by 50. It isn't hard to beat a patsy by any number, and giving more than human weight to SOS creates extremely adverse incentives in scheduling.

Humans consider both the score and the quality of the opponent when evaluating the outcome of a game. Computers aren't allowed that luxury.

Indeed, because it would be impossible to come up with a algorithm that was objective that would not immediately be exploited if you threw in something like SOS.

So now when a human sees that USC struggled to get past ASU, they can say that is meaningful, while Florida beating up on Southwest Lousiiana State College of Law by 48 is not.

[Eddie Teach]

Not at all - beating a average or mediocre team by a field goal SHOULD count for more than beating some patsy by 50. It isn't hard to beat a patsy by any number, and giving more than human weight to SOS creates extremely adverse incentives in scheduling.

I don't think of high rankings as rewards so much. It should be, which team do I think has played better this year and which would I expect to win if they played each other now. Narrow victories against mediocre teams show vulnerability. Blowout wins against patsies don't, other than lack of confidence on the part of the AD.

Indeed, because it would be impossible to come up with a algorithm that was objective that would not immediately be exploited if you threw in something like SOS.

So now when a human sees that USC struggled to get past ASU, they can say that is meaningful, while Florida beating up on Southwest Lousiiana State College of Law by 48 is not.

Agreed, and that's why I prefer the human polls to the computer ones.

[stjaba]

But even human polls have their flaws. For instance, in the Harris and Coach's Polls, USC is now rated ahead of Oregon, even though Oregon beat the Trojans 2 weeks ago.

[Berkut]

Not at all - beating a average or mediocre team by a field goal SHOULD count for more than beating some patsy by 50. It isn't hard to beat a patsy by any number, and giving more than human weight to SOS creates extremely adverse incentives in scheduling.

I don't think of high rankings as rewards so much. It should be, which team do I think has played better this year and which would I expect to win if they played each other now. Narrow victories against mediocre teams show vulnerability. Blowout wins against patsies don't, other than lack of confidence on the part of the AD.

I think beating a patsy by any number tells you nothing about a team.

Beating a mediocre team however, tells you that you beat a mediocre team. Which means that you are, at least in theory, better than mediocre.

And it isn't about rewards, it is about what the game tells you. Florida can beat WSU by 50, and so can Arizona, and so can OSU. So what?

The real issue is that if you give MoV weight in the computer polls, you are going to reward good teams even more for playing shit teams, and then running up the score on them, since the computer algorithms cannot possibly be sophisticated enough to not be exploited.

Indeed, because it would be impossible to come up with a algorithm that was objective that would not immediately be exploited if you threw in something like SOS.

So now when a human sees that USC struggled to get past ASU, they can say that is meaningful, while Florida beating up on Southwest Lousiiana State College of Law by 48 is not.

Agreed, and that's why I prefer the human polls to the computer ones.

But human polls have an entirely different set of problems, in that they are not objective at all, and do in fact rewards not losing way too much, which then creates crap like the SEC never playing anyone OOC, and even having most conferences play fewer conference games so they can load up on patsies.

I don't think computer polls are better per se, but they are more objective, and more useful for comparing aggregate numbers, like SoS and such, or overall conference strength from top to bottom. They will however, have their inevitable outliers, of course.

[Eddie Teach]

I think beating a patsy by any number tells you nothing about a team.

Beating a mediocre team however, tells you that you beat a mediocre team. Which means that you are, at least in theory, better than mediocre.

I agree on the first point. There shouldn't be any real upside to beating a team you're expected to blow out.

However, if you struggle against a team you're expected to beat easily, that's reason to downgrade the expectations. The computers won't catch that, to them a win is a win, period.


p.s. Central Michigan and Northern Arizona?  Methinks you're overstating the SOS differences just a tad. Heck, every team in the SEC except Ole Miss has at least one non-conference game against a BCS foe.

[Berkut]

p.s. Central Michigan and Northern Arizona?  Methinks you're overstating the SOS differences just a tad. Heck, every team in the SEC except Ole Miss has at least one non-conference game against a BCS foe.

NAU is a patsy of course.

Central Michigan is actually rather good - I think they have two losses this year. Certainly not a patsy at all.

And of course you are kind of forgetting Iowa. Funny that.

And the real kicker is not comparing 3 OOC games to the SECs three - it is comparing that 4th OOC games that they play against Arizona playing USC/Oregon/OSU/Cal/Stanford. That is why the Pac-10 has such a strong SOS, compared to the SEC, that 4th game is being played against a top-50, if not top-25 team, rather than another blowout against some patsy.

[Berkut]

I think beating a patsy by any number tells you nothing about a team.

Beating a mediocre team however, tells you that you beat a mediocre team. Which means that you are, at least in theory, better than mediocre.

I agree on the first point. There shouldn't be any real upside to beating a team you're expected to blow out.

And that is the problem with the human polls - there is upside to it - you get to continue to be talked about as a national title contender for blowing away a bunch of them every year. The way the system works now, the Pac-10 is foolish to play such a strong OOC schedule, and not load up with yet another patsy in place of a conference game.

At least from the perspective of getting into BCS bowl games. Why risk having a bad day against a decent team when you can just pummel some joke of a school like the SEC does? Much less playing another conference opponent.