Famous NCAA Dynasties

[Spence]

I don't necessarily 'object' to how the BCS does it's 'ranking' but I would prefer a model that is more objective, than what they utilize.
Aren't there something like 8 computers applied? It's no wonder there isn't a 'concensus'. They should apply the most reliable computer model, and use it. Even their computer ranking lacks objectivity.

Having more computer polls makes it more objective, not less. Same idea as the human polls. The more voters the less likely a small group could have an effect. Limiting to one computer poll would take away from the effectiveness of that portion of the equation.

[ktffan]

85% ranked against unranked does sound impressive. My question would be when the ranked teams play each other how often does the higher ranked team beat the lower ranked team?

I'll give you a little more info on that when I get home. It'll probably be Monday.

I'm not against the AP poll or the coaches poll. I think they probably shouldn't have 2/3 weight, but I do think they do a good job adjusting. If you look at most relevant computer polls they look a lot like the wire service polls. Which means that most are using the same criteria.

You could be right. However, I think they ought to quit being hypocritical and eliminate the computers all together. Everytime their formula fails to name the same top two teams as the wire service polls have, they deem the formula a "failure" and change it so that they would have been in. If we are going to determine the top two teams by who finished #1 and #2 in the wire service polls, why not just use the wire service polls? Either they are going to have to change the collective opinions of the media and the fans, or they should just go with the wire service polls.

As to how biased computer polls are, I guess that would depend on whether the guy who wrote the formula went about it. If he did it without having a specific conference or team in mind and tried to make it as objective as possible it would be OK.

Computer polls are written to reflect the opinion of the programmer. Every programmer looks at his results and judged by what he thinks it should be and makes changes to reflect what he thinks it should be. That's not really unbiased, although, we hope, they don't change it in mid-season. You can elevate almost any of the major teams strictly by highlighting a statistical aspect that favors that particular conference, for instance, if you favor the PAC-10, you give more points to teams that play opponents that are deemed tougher, as the PAC-10 usually plays a tougher non-conference schedule. This is exactly what Hester and Anderson did when they made their poll, and not surprisingly, it favored PAC-10 teams for a while. Other conference you can elevate by counting wins more, than SOS, ie, the SEC or the Big 12. I view computer polls as just another opinion.

[Spence]

You could be right. However, I think they ought to quit being hypocritical and eliminate the computers all together. Everytime their formula fails to name the same top two teams as the wire service polls have, they deem the formula a "failure" and change it so that they would have been in. If we are going to determine the top two teams by who finished #1 and #2 in the wire service polls, why not just use the wire service polls? Either they are going to have to change the collective opinions of the media and the fans, or they should just go with the wire service polls.

I agree, if they keep telling the computer polls they have to change to be more like the wire service polls then it makes them a joke.

Computer polls are written to reflect the opinion of the programmer. Every programmer looks at his results and judged by what he thinks it should be and makes changes to reflect what he thinks it should be. That's not really unbiased, although, we hope, they don't change it in mid-season. You can elevate almost any of the major teams strictly by highlighting a statistical aspect that favors that particular conference, for instance, if you favor the PAC-10, you give more points to teams that play opponents that are deemed tougher, as the PAC-10 usually plays a tougher non-conference schedule. This is exactly what Hester and Anderson did when they made their poll, and not surprisingly, it favored PAC-10 teams for a while. Other conference you can elevate by counting wins more, than SOS, ie, the SEC or the Big 12. I view computer polls as just another opinion.

They are just another opinion, but if they are done right and they don't change them over time, then they could be good measures. If you change every year to reflect the changes in what you deem important, your poll loses validity. If you pick your criteria and stick with it, then over the course of time it becomes valad. Football styles in conferences change constantly, so if the criteria doesn't, then how relevant your poll is will show over time.

The computer polls that are ever changing are the ones that will never be valid.

[billybud]

The reasons why I have a hard time following a reward for conference SOS is that teams play schedules and not conferences and the variability within conferences is great...if you play a weak schedule because your conference opponents are weak, it should show up in your schedule strength (TEAM SOS)...but it really doesn't. If you schedule a decent OOC opponent while the other schedules a patsy, you make up a lot of ground. Much more than just the conference's average strength make up your schedule strength...OHio State playing a Texas is a heck of a lot stronger than Ohio State playing Buffalo.

Using the sagarin SOS...for end of year 2005

The Big 12 team's SOS ranged from the nation's toughest at #1 to #66.....Two teams played a #1 and #6 schedule while other teams in the Big 12 played SOS's ranked #66, #52, #47, #38, #37, #36, #29, #24..etc ...the average SOS for the conference was #30. But, that didn't really affect the two teams with SOS's ranked #1 and #6

The Pac 10 was much the same...schedules ranged from the #8 and #9 ranked SOS to the #67 ranked SOS...the average rank was #28.

It seems that a team can have a TOP 5 SOS while playing in a not so strong conference....so if you reward for tough scheduling...what does the conference have to do with it? The team being ranked plays a schedule which, incidentally, includes teams in their conference. It would play out in the individual ranking.

[ktffan]

The computer polls that are ever changing are the ones that will never be valid.

So, since Sagarin changed his formula a couple years ago, is it invalid?

[Spence]

The computer polls that are ever changing are the ones that will never be valid.

So, since Sagarin changed his formula a couple years ago, is it invalid?

I think Sagarin only changed that once for the BCS. I also think he still publishes his poll as it was meant to be. Although I do think it was wrong for him to bow down, he tries to get the best of both worlds.

[ktffan]

The computer polls that are ever changing are the ones that will never be valid.

So, since Sagarin changed his formula a couple years ago, is it invalid?

I think Sagarin only changed that once for the BCS. I also think he still publishes his poll as it was meant to be. Although I do think it was wrong for him to bow down, he tries to get the best of both worlds.

No, he created ELO-Chess for the BCS. He changed his regular ranking system after the 2003 season.

[mountainman]

If memory serves me correctly, the computer polls used by the BCS had Texas as the #1 team in the BCS Standing every week except for the first week the BCS Standing were released. I don't recall that being the case with the human polls.

I don't know what that means, but as it turned out they were correct.

Does anyone know what data the computers or the humans use for determining the rankings? Further, how is this data used?

[Spence]

Does anyone know what data the computers or the humans use for determining the rankings? Further, how is this data used?

All the computers compile the data into a formula that the deem will give them the best results. The human polls basically are opinion polls. I think most of them watch sportscenter final and decide who looks the strongest. As bad as that sound, though, they do a pretty good job. The polls evolve because they go by what happened last year to determine this years preseason poll. Ohio State this year will get more respect then they deserve because most of the offense will be back. The human polls seem to put more weight to offense. I believe that defense is more important, especially in the college game. Teams with good defenses and average offenses aren't as telegenic as teams with good offense. They don't get played in the highligts on sportscenter very often.

I don't believe the human polls are intentionally biased, I just think they don't actually see enough games.

[mountainman]

Yep, I figure that too, Spence. I would like to know what data the computer systems are being given, especially since already in this topic I read where the percentage of stadium capacity and graduation rate was being used. Neither of those two data elements will get a team a first down and its got my curiosity aroused.

[Spence]

I would imagine that all are different, but the legit ones, the ones without an agenda, wouldn't be using stadium capacity or grad rates as part of the criteria. I think his point was that computer polls aren't necessarily these great, unbiased entities that we think of when we think computer poll. That they can be very biased towards getting the result that they want based on the criteria they use. I would hope that people would check into things like that before giving a particular poll that much weight.

[Derek]

"I don't believe the human polls are intentionally biased, I just think they don't actually see enough games."

I can agree with most of that.....But I also think there is some bias given to certain teams that have a lot of notoriety. Maybe not every voter, or even enough to make a difference, but there are some that will give their Alma Matter or hometown a slide if they are close.

Nothing against Notre Dame, but they would get a number one ranking in the old system even if they had a loss and a lesser known school had none. This has proven itself out over time.

Public favorites like ND usually get the benefit of the doubt.

[ktffan]

Yep, I figure that too, Spence. I would like to know what data the computer systems are being given, especially since already in this topic I read where the percentage of stadium capacity and graduation rate was being used.

Those factors were used by the CFN "program analysis". They aren't really a computer poll, and since they consider is a program analysis, they aren't really ranking the best teams (or best conferences), but best "programs". In those respects, I suppose those factors are fair to include, but you have to wonder why stadium capacity rates more than tickets solds.

[Spence]

Those factors were used by the CFN "program analysis". They aren't really a computer poll, and since they consider is a program analysis, they aren't really ranking the best teams (or best conferences), but best "programs". In those respects, I suppose those factors are fair to include, but you have to wonder why stadium capacity rates more than tickets solds.

I guess it is fair to include grad rates when doing a program analysis. Especially now when grad rates are considered by the NCAA. As to why stadium capacity rates more then tickets sold, maybe the reasoning is that the school can control stadium capacity. It can't really control tickets sold. USC doesn't draw like Alabama traditionally, they never will.

[ktffan]

Those factors were used by the CFN "program analysis". They aren't really a computer poll, and since they consider is a program analysis, they aren't really ranking the best teams (or best conferences), but best "programs". In those respects, I suppose those factors are fair to include, but you have to wonder why stadium capacity rates more than tickets solds.

I guess it is fair to include grad rates when doing a program analysis. Especially now when grad rates are considered by the NCAA. As to why stadium capacity rates more then tickets sold, maybe the reasoning is that the school can control stadium capacity. It can't really control tickets sold. USC doesn't draw like Alabama traditionally, they never will.

I don't know. To me, it's a healthier sign of a program, if they are planning ahead and meeting their needs. I've always been perplexed that Nebraska fans are so proud about their sell out streak. Essentially, it's next to impossible to get a ticket, and they think this is a good thing. At the same time, they have a stadium that is a good deal smaller than the largest stadiums out there. It seems to be that if they had planned for such a thing in the 70s, they could easily have a stadium that would be meeting their needs now instead of a string of sell outs, which really mean nothing except people can't get in.