BCS, Schedules, and Conference Ranks

#2
#2
The problem is not only objectivity; it is a lack of information. If everybody played round robin, I agree that the computers could run a matrix and determine who performed the best. However, this doesn't happen in a 12 game football season.

It is impossible to quantify a night game in a hostile stadium, or a team that has improved throughout the season, or a game that turned on a bad call or a bad bounce, through purely statistical methods. This requires subjectivity, which is something the computers are wholly unable to produce (and are specifically designed not to produce.) Take a quick look at the "Billingsley National Champions" if you need some examples--they crowned Miami in '90, Penn St. in '86, Florida in '84, Oklahoma in '80, USC in '76, and they are far from the worst (that honor probably belongs to FACT or Sagarin.)

The human polls, of course, also have their flaws, as they do often lack objectivity. The BCS has tried to overcome these shortcomings by combining the two, but the fact is there is no perfect system.
 
#3
#3
I just where the computers help is in taking what happens on the field as what it is. Yes, they're flawed because it's all based on a very small sample. And even they would tell you that the game on the field that day is still up in the air.

The thing that bugs me the most as lazy thinking is that when the computers differ a lot from the human polls, it is put forth as evidence that the computers don't work. That's silly. That presupposes that the polls are correct -- that they are the benchmark. "Since we all agree that USC is the best team, the computers must be wrong." It could just as easily be that the pollsters are wrong.

It's all crap, but I sometimes like the objective look of the puters just for comparison sake.
 
#4
#4
cotton said:
The problem is not only objectivity; it is a lack of information. If everybody played round robin, I agree that the computers could run a matrix and determine who performed the best. However, this doesn't happen in a 12 game football season.

It is impossible to quantify a night game in a hostile stadium, or a team that has improved throughout the season, or a game that turned on a bad call or a bad bounce, through purely statistical methods. This requires subjectivity, which is something the computers are wholly unable to produce (and are specifically designed not to produce.) Take a quick look at the "Billingsley National Champions" if you need some examples--they crowned Miami in '90, Penn St. in '86, Florida in '84, Oklahoma in '80, USC in '76, and they are far from the worst (that honor probably belongs to FACT or Sagarin.)

The human polls, of course, also have their flaws, as they do often lack objectivity. The BCS has tried to overcome these shortcomings by combining the two, but the fact is there is no perfect system.
Click to expand...
Miami was the best team in '90. Penn State did win it in '86. Florida was three touchdowns better than BYU in '84. What's your point?
 
#5
#5
I have my own computer poll that I put together in spare time, so it's been interesting to see how exactly these get put together and how it bounces out. I think a lot of the general considerations for a human poll ("hostile crowd", other team's tradition, etc) are worthless, and actually playing good football is underrated. This is how a team like Rutgers, unbeaten as they are and having played a tougher schedule than a couple teams ahead of them, came in #15 with Louisville #3. A lot of this is still based on preseason rankings and doesn't take actual wins and losses into account against quality opponents.

By the way, tonight's game featured #5 defeating #1.
 
#6
#6
hatvol96 said:
Miami was the best team in '90. Penn State did win it in '86. Florida was three touchdowns better than BYU in '84. What's your point?
Click to expand...

I don't know why I included Penn St 86 in there--must have picked something up by mistake.

The point is that the computers yield a very different result than the system we have used to crown a champion. And there is definitely subjectivity in the computer polls, but the subjectivity lies in which data is used to rank the teams and how it is compiled. Which formula do you use?

As I stated before, Billingsley appears to produce, in my opinion, better results than some of the others, but take a look at Sagarin, also a part of the BCS:

2002: USC
1998: Ohio St.
1994: Penn St (I think that's where I goofed)
1992: Fla St
1989: Notre Dame
1987: Fla St
1986: Oklahoma
1985:Michigan

Another bit of evidence that there are problems relying on simply formulaic rankings to produce a champion is the number of times those formulas disagree.

The AP and Coaches Poll (BCS years included) have disagreed 10 times since 1956 (a nice 50 year sample,) yielding 2 National Champions: 1957, 1965, 1970, 1973, 1974, 1978, 1990, 1991, 1997,2003.

Computer polls have crowned multiple National Champions in 41 of those years, and at least 3 teams finished #1 in recognized polls 24 times (there were 5 computer champions in '60,'70, '73, & '80.)
 
#8
#8
cotton said:
Another bit of evidence that there are problems relying on simply formulaic rankings to produce a champion is the number of times those formulas disagree.

The AP and Coaches Poll (BCS years included) have disagreed 10 times since 1956 (a nice 50 year sample,) yielding 2 National Champions: 1957, 1965, 1970, 1973, 1974, 1978, 1990, 1991, 1997,2003.

Computer polls have crowned multiple National Champions in 41 of those years, and at least 3 teams finished #1 in recognized polls 24 times (there were 5 computer champions in '60,'70, '73, & '80.)
Click to expand...

I'm not sure I'm buying that rationale. The reason that could happen is because the human polls have a lot of "heard bias" -- that is, everyone kind of decides through the year who they think is best, a lot of which is dictated by a) group think and b) preseason rankings. Neither of those is relevant.

The other MAJOR reason is that there are so many teams and not much overlap in schedules. At the end of the year you have a bunch of teams that don't get to settle it on the field, so the bowls and voters effectively "decide" who gets consideration for the NC beforehand. Thus, you end up with one, two, or three options. Everyone else is left out regardless of merit.
 
#9
#9
Liper said:
I'm not sure I'm buying that rationale. The reason that could happen is because the human polls have a lot of "heard bias" -- that is, everyone kind of decides through the year who they think is best, a lot of which is dictated by a) group think and b) preseason rankings. Neither of those is relevant.
Click to expand...

First of all, I'm not sure that "group think" or preseason rankings are completely irrelevant. Consesus is sometimes flawed, but it is as good an indicator as any other as to who is best. Preseason rankings are consensus predictions that, until proven faulty, can be a tool used to sort the pack. If everybody thought Ohio St was better than Rutgers in August, and they both win out in similar fashions, what is the reasoning that Ohio St isn't better than Rutgers in January, the fact that there is a crystal trophy involved in the decision?

Liper said:
The other MAJOR reason is that there are so many teams and not much overlap in schedules. At the end of the year you have a bunch of teams that don't get to settle it on the field, so the bowls and voters effectively "decide" who gets consideration for the NC beforehand. Thus, you end up with one, two, or three options. Everyone else is left out regardless of merit.
Click to expand...

I believe this flaw is magnified in the computer rankings. How do you sort out teams with similar results and no overlap? That is why you end up with so many multiple NCs in the computers (see 1981, where Clemson won out in most rankings including both voting polls, Penn St. won Dunkel, and something called the National Championship Foundation awarded the title to Clemson, Pitt, Texas, and SMU.) It is also why you get results like Oklahoma as Berryman's NC in 2003 after getting dumped by LSU in the Sugar Bowl, leaving LSU, USC, and OU with 1 loss each.
 
#10
#10
From my experience in sports betting, you can make plenty of money by looking at the public consensus, then going the opposite way.

Therefore, I think in NCAAF, group think isn't a good barometer for which team is best.
 
#12
#12
cotton said:
If everybody thought Ohio St was better than Rutgers in August, and they both win out in similar fashions, what is the reasoning that Ohio St isn't better than Rutgers in January, the fact that there is a crystal trophy involved in the decision?

I believe this flaw is magnified in the computer rankings. How do you sort out teams with similar results and no overlap? That is why you end up with so many multiple NCs in the computers.
Click to expand...

1. If I'm reading this right, the OSU - Rutgers issue would be circular logic. We've decided they were number one; and because of that, they win the trophy; and if there is a consensus about it then the trophy proved the polls were right. It starts with a flawed assumption and then gets a result based on the flaw.

2. Exactly. How would you sort it out? I mean, some human poll is no better than a computer projection. The reason you could end up with a bunch of NCs is because there could potentially be that many if it were played.

The human polls would be better if no one voted until the season was over based on the sum total results on the field. At least then you wouldn't have all of these "they lost move them down, they won move them up" rules that are ridiculous.
 
#13
#13
the BCS is great for one thing and one thing, well, maybe two things...but first and foremost...none of us can stop talking about it...it's automatic fodder for all the fans and talking heads and is a huge point of interest for both. That's not a bad thing for college football. there's not another sport with a more meaningful regular season week to week.

2nd, it has provided matchups we would not have seen in the old conference alignment system.

there would have been no OSU/Miami, TX/USC, LSU/OK, USC/OK without the BCS. and those 1st two are highly regarded as two of the best bowl games ever played.

it's no where near perfect..and with L'ville losing last night...TX, USC, CAl, ND, UF, Auburn, ARK all have legitimate reasons to think they are the next best team. And don't forget about Rutgers. they are undefeated too. IMHO, if they are an undefeated conf. champion of a BCS conference, they in my mind are as deserving as anyone else. they may not be the next BEST team, but they're NO LESS DESERVING than any of the teams that have one loss.

and here's one for ya...say TN beats ARK this weekend, SC beats UF this weekend and Auburn wins out. ARK beats UF in the SEC CG and winds up a conf champ with 2 losses. UF has three losses.

who's the only one loss team in the SEC? 3rd place Auburn. Do they deserve a shot at a national title?

IMO, NO. it would be the same as when OK and NEB both got to play for it when they didn't win their conf. and that was horse crap too.

IMO when all is said and done, IF UF wins out and beats a top ten Auburn or ARk team for the SEC, IF TX wins out and beats Neb. again for the Big 12, IF USC or CAL win out and win the Pac 10, IF Rutgers was to win out and win the Big EAST, those are the teams with legit arguments as to who should be playing for the National championship.

With it ultimately coming down to three teams: UF, TX and CAL/USC (and only if one of those two should win out).

If all three of those teams do what's expected and win out, any one of those three has the right to expect to play for the NT.
 
#16
#16
VolinArizona said:
From my experience in sports betting, you can make plenty of money by looking at the public consensus, then going the opposite way.

Therefore, I think in NCAAF, group think isn't a good barometer for which team is best.
Click to expand...

Not a good analogy, in my opinion. First, the group think in betting is actually very good--favorites win more times than they lose, and point spreads are usually indicative of whether or not a game will be competitive. There are errors, which good gamblers exploit, but as a whole, it is not that bad of an idicator (it would probably be better if you used money line betting rather than point spread.) Second, AP sportswriters or Coaches are a different group than money players, so the "group think" is going to be different, and hopefully, better educated.

jakez4ut said:
the BCS is great for one thing and one thing only...
Click to expand...

...and that thing is pitting BCS1 vs BCS2 fot the BCS national title. That doesn't mean that it is perfect, or even very good, at picking its 1 and 2. It is good when there are clearcut 1&2s, such as FSU and Va Tech in '99. It is very questionable any time the issue is not so clear cut, such as 2003.

Liper said:
1. If I'm reading this right, the OSU - Rutgers issue would be circular logic....

2. Exactly. How would you sort it out? I mean, some human poll is no better than a computer projection....

The human polls would be better if no one voted until the season was over based on the sum total results on the field.
Click to expand...

1. The logic is semi-circular, but that doesn't mean it is wrong.

2. That is debatable. At least in my opinion, the voting polls get it right fairly often. The computers get it right some, wrong some, and disagree with each other almost all of the time. It just occurred to me that the BCS is actually trying to make an AP type poll for the computers by averaging them, with each computer getting a "vote." Maybe the average will end up being as good as the voting polls--they keep changing the comps, so I don't think there is enough data to tell--but even then, I don't think they will be any better.

3. Voting at the end of the year would not be better than voting throughout, you just wouldn't see the illogic until the end of the season.
 
#17
#17
cotton said:
1. Well, if it's circular, then it doesn't mean anything one way or the other. Thus, not a factor.

2. This is what I'm trying to point out. The only reason we say the computers get it wrong sometimes and right sometimes is whether or not they agree with our predefined criteria, i.e. human polls. That's not a defensible position because it is based on our opinion.

3. I think starting out with rankings poisons the process. Teams start out ahead or behind based on nothing, and then get reshuffled accordingly. Moreover, it has a trickle down effect because the opponent is viewed under this same light, i.e. they beat/lost to a good team or a bad team so it means X.

The season as it unfolds, and the results throughout, are not retroactively taken into account in the next poll. It's proven that people tend to extrapolate and over emphasize whatever they have experienced most recently.
Click to expand...
 
#18
#18
1. Just because everybody thought Ohio St was better than Rutgers in the beginning of the season doesn't mean they are not. I don't understand your logic there. You actually have a season's worth of data backing up what would prove to be a very good job of preseason ranking, should they both win out.

2. You are correct. The only reason we think the computers are wrong is because they give us results that people and other computers believe don't make any sense.

3. The process would be "poisoned" anyway. Do you not think people are going to be opining on who has the best team from the preseason if the poll doesn't come out until Groundhog Day? And actually, winning or losing against a particularly good or bad team should mean something. Who is playing particularly well at the end should mean something. Even if you could remove these factors by delaying the poll, I don't know why you would want to.

EDIT: 500 posts. I need a better hobby.
 
#19
#19
cotton said:
1. Just because everybody thought Ohio St was better than Rutgers in the beginning of the season doesn't mean they are not. I don't understand your logic there.

3. The process would be "poisoned" anyway...And actually, winning or losing against a particularly good or bad team should mean something. Who is playing particularly well at the end should mean something...
Click to expand...

1. Correct, it does not mean they are not. It also does not mean they are.

3. Correct, it should mean something. But that "meaning" should not be derived from how good or bad we thought they would be but how good they are; two different things. An even bigger issue is that the unfolding season, with more reality being brought to light, is not then used to recalibrate the human poll correctly. The computers, good or bad, take this into account on an objective, consistent basis. All available information is baked into the cake.
 
#20
#20
:dance2: I am so excited. This was a great read with some good points and counterpoints. I swear, the solution to this is about as hard of a debate as religion or politics (which I guess it is the politics of a certain religion, really). Anyway...good discussion. I realized long ago that I don't have any answers. I think a playoff SOUNDS great...but I don't see how to pull it off in a good manner without adding too many games to an already hard season. Anyway, Kudos y'all.
 
#21
#21
I don't think a playoff is viable, but I had that thought process in a different thread that I don't feel like searching for.

Here's one to try on for size...in Ohio, the high school playoff teams and seeds are determined based on a precise mathematical formula rather than a simple W-L listing. Thus, every game counts for something rather than just a certain few.
 
New posts
You must log in or register to reply here.

VN Store



Back
Top