Northern Illinois and Fresno State

[Spence]

I do not disagree. In the words of some Yankee, years ago, Wm Marcy, "to the victor goes the spoils." Have you ever noticed, since half time shows on television have been replaced with half time reports, much to the demise of our civilization, you listen to twenty minutes of facts, statistics, charts, graphs, compilations that support the theory of spontaneous generation and then the closing comments are, "yes, but on any given Saturday or Sunday..."

So here is the question, and it is not hypothetical. (We have this discussion yearly but it is always good to revisit.)

It is the begin of the season. No games have been played. The preseason polls come out and they have listed a Strength of Schedule. What "objective" criteria is used for the initial SOS?

That is so true. People listen to spin an except it as facts.

Duke, the problem with using SOS as an objective tool is you don't have enough information to compare teams to make the SOS objective. It would reqiure a form of national scheduling to make that happen. A tournament of champions is likely the easiest to make happen and probably the most fair. A great #2 team in a conference would get hosed, but they also had the ability to win their conference and didn't so that is on them.

[billybud]

The Massey Index gives you an idea of what the computers "see"....

The various algorithmns, like humans, see different aspect...but align pretty well in mass on a bell curve.

http://masseyratings.com/cf/compare.htm



And...the computers see NIU and Fresno State as #29 and #30....the computers are not as impressed with being undefeated since they factor in the quality of the wins.

[billybud]

If you are interested in the validity of various computer polls,,,and the categories differ from picking out straight up winners (and Sagarin is right up there at 76%) to accurate predictions of spread...

http://www.thepredictiontracker.com/awards2012.html

[donovan]

But isn't it true there is a difference between polls that are "predictor" polls, ones that try to determine a winner of a game not yet played and a "ranking" poll, one that ranks teams based solely on past performance for a given time frame?

If that is correct you can have rankings based on objective data. Wins, Losses, scores, spreads, attendance, number of hot dogs sold, not gallons of beer dispensed....some number are way too large to comprehend. (There are no rankings, I am aware, of that use data only for ranking, why...that is the definition of boring.)

I think most listings, are predictors. You post Mr. Billybud uses that terms and I semi remember, everything is now only semi remembered, a few years ago when we had the spirited conversation about this and Mr. Congrove chimed in. I think that all still holds true.

I think LSU will beat Alabama this weekend, and FSU and Ohio State will play in Pasadena because Oregon will get beat this year and if not before, by Oregon State. I've been wrong before.

[billybud]

You do have predictive ratings versus retrodictive ratings....

Retrodictive ratings do not attempt to predict how a future game between two teams might be...such as an Oregon vs FSU vs an Oregon vs Alabama

Predictive ratings are used to predict which teams are strongest while a retrodictive ranking determines, afterwards, which teams had the most impressive season.

All Vegas lines, for example, are predictive.


Both systems are judged differently..

Retrodictive: Judged by how well past results are fit. For example, what % of games were upsets based on the current ratings.

Predictive: Judged by how well the predictions work. For example, what % of games were upsets based on the ratings from the day before the game.

[billybud]

I have had my differences with CCR over the years....and I make no bones about it.

I sure don't have a clue about his proprietary algorithmns....Tried to back figure for years/

But I see that he is a major "outlier" in some areas. For instance...of the 115 computer systems tracked by the Massey Rankings Composite...most Have FSU as #1 and no ranking sytstem of the 115 has FSU lower than #3...

CCR has FSU at #5....a way out outlier.

[billybud]

In trying to compare CCR with Sagarin (and I admittedly have followed Sagarin), I pick the real differences between the ranking of a team between the two systems and try to see what is happening.

For instance:

NIU:

CCR Ranking....#13
Sagarin..........#46

Both systems have NIU with an SOS up in the 100's (Sagarin #134, CCR #124)...

Sagarin has Fresno State at #50....CCR at #11..

Obviously something very, very different is going on in the algorithmns.

The Massey Composite of 115 computer systems has Fresno and NIU at #29 and #30....I have come to the conclusion over the years that CCR weighs a straight win-loss record much heavier than the value of the wins (SOS). That was true when CCR had Ball State up so high some years ago.

[Spence]

I think LSU will beat Alabama this weekend, and FSU and Ohio State will play in Pasadena because Oregon will get beat this year and if not before, by Oregon State. I've been wrong before.

Boy, I hope not. I don't think the Buckeye's are ready for the national championship. I'm betting on the Rose. result may not be much better, but it will give them a chance to get their feet wet for next year.

All the rankings whether predictive or retrodictive seem to be on the same page by the end. I do like predictive ranks better because it provides a check of how they are performing. I would say, though, that when I look at most of the computer rankings that I wonder how some teams fit where they fit. I guess if I am a Buckeye fan I should be happy with Ohio State ranked at the top or near the top, but if I don't agree with 10 of their top 25 I don't know if I should be alright with it.

Computer rankings are subjective to the point that the algorythms are based on the builders views on what makes a team great. Then you have to live with the results. I think once you decide to go public with your rankings you should not be tweaking the results. I think conforming their rankings to what the BCS committee wanted compromised the integrity of the group of computer rankings and also limited their abilty to have an equal impact on the process. The BCS is top heavy with the human polls. Sagarin and company have sold out to have their rankings used. They shouldn't have.

As far as Sagarin is concerned, I do think he rankings use past performance, which makes it harder for a team that is only good this year to do well.

Most people gravitate to polls and rankings that have their teams higher. It is human nature.

[Duke1632]

Duke, the problem with using SOS as an objective tool is you don't have enough information to compare teams to make the SOS objective.

I don't disagree, but insofar one thinks SoS ought to be used to measure team ranking, it's there to use. If there's little data, then it should not carry much weight. But, remember, it is not my argument that SoS should be an important factor. In fact, I do not think it should be as important as W/L, but today it appears to be the most important. Many say OSU and Baylor do not deserve consideration b/c their SoS is weak (you've even hinted at that). Well, it turns out, that's just not true (according to CCR, at least) and secondly, those folks are ignoring your argument that "you don't have enough information." Therefore, if you use SoS at all, it should be a function of the certainty/objectivity of those data.

As to Donovan's question if it's preseason, then the rankings can reflect no SoS factor being used at all. In fact, at that point there's nothing objective that is available, except for maybe last year performance measured against returning players and coaches, etc.

[billybud]

I think LSU will beat Alabama this weekend, and FSU and Ohio State will play in Pasadena because Oregon will get beat this year and if not before, by Oregon State. I've been wrong before.

Boy, I hope not. I don't think the Buckeye's are ready for the national championship. I'm betting on the Rose. result may not be much better, but it will give them a chance to get their feet wet for next year.

All the rankings whether predictive or retrodictive seem to be on the same page by the end. I do like predictive ranks better because it provides a check of how they are performing. I would say, though, that when I look at most of the computer rankings that I wonder how some teams fit where they fit. I guess if I am a Buckeye fan I should be happy with Ohio State ranked at the top or near the top, but if I don't agree with 10 of their top 25 I don't know if I should be alright with it.

Computer rankings are subjective to the point that the algorythms are based on the builders views on what makes a team great. Then you have to live with the results. I think once you decide to go public with your rankings you should not be tweaking the results. I think conforming their rankings to what the BCS committee wanted compromised the integrity of the group of computer rankings and also limited their abilty to have an equal impact on the process. The BCS is top heavy with the human polls. Sagarin and company have sold out to have their rankings used. They shouldn't have.

As far as Sagarin is concerned, I do think he rankings use past performance, which makes it harder for a team that is only good this year to do well.

Most people gravitate to polls and rankings that have their teams higher. It is human nature.

Wrong about Sagarin....Sagarin says that there are enough matches by week six for his rankings to be fully linked and there are no vestiges of last year. His first rankings do start out with a basis in the last year but all is adjusted as soon as enough current year games have been played...by week five or six.

[billybud]

Duke, the problem with using SOS as an objective tool is you don't have enough information to compare teams to make the SOS objective.

I don't disagree, but insofar one thinks SoS ought to be used to measure team ranking, it's there to use. If there's little data, then it should not carry much weight. But, remember, it is not my argument that SoS should be an important factor. In fact, I do not think it should be as important as W/L, but today it appears to be the most important. Many say OSU and Baylor do not deserve consideration b/c their SoS is weak (you've even hinted at that). Well, it turns out, that's just not true (according to CCR, at least) and secondly, those folks are ignoring your argument that "you don't have enough information." Therefore, if you use SoS at all, it should be a function of the certainty/objectivity of those data.

As to Donovan's question if it's preseason, then the rankings can reflect no SoS factor being used at all. In fact, at that point there's nothing objective that is available, except for maybe last year performance measured against returning players and coaches, etc.

Duke:

We have had this discussion over the years...but just wins-losses by themselves carry little meaning without a value attached. Its like having a pocket full of one dollar bills vs a pocket full of 100 dollar bills. You may have the same number of bills...but the total value of those pieces of paper will differ greatly.

Is going 12-0 in the MAC the same as going 12-0 in the SEC?

Does beating Alabama and LSU and Auburn mean the same as beating NIU, Ball State, and Toledo? The top three in the SEC West vs the top 3 in the Mac West.

If wins do not have a value, there is nothing that separates a 6-2 Michigan from a 6-2 Louisianna-Lafayette.

[billybud]

The problem is in attaching value to wins and to losses.....Humans do that in their voting based on their bias, an "eye test", etc.

[Duke1632]

Duke:

We have had this discussion over the years...but just wins-losses by themselves carry little meaning without a value attached. Its like having a pocket full of one dollar bills vs a pocket full of 100 dollar bills. You may have the same number of bills...but the total value of those pieces of paper will differ greatly.

Is going 12-0 in the MAC the same as going 12-0 in the SEC?

Does beating Alabama and LSU and Auburn mean the same as beating NIU, Ball State, and Toledo? The top three in the SEC West vs the top 3 in the Mac West.

If wins do not have a value, there is nothing that separates a 6-2 Michigan from a 6-2 Louisianna-Lafayette.

I certainly understand this and actually agree with you. But there's the other side too that I also agree with: that it is in fact possible that an undefeated team in the MAC can be competitive with the very best of the rest of the league. Such would be rare and absolutely astounding, and something that ought to be cultivated, not forbidden.

So why not advocate something of a compromise? Why draw that hard line and say in advance, if you are from the MAC, there's no possibility that you are good enough so your accomplishments (and actual quality) are now and in the future meaningless. And there is, after all, a valid compromise that is different than the two extremes of (1) only W/L matters or (2) only SoS matters.

Seems to me CCR does that as well as anyone else. It also seems to me that the predictive rankings models are the way to go, at least to satisfy your concern (e.g., don't want a bowl game blowout). Suppose a model that get's 80% right one year says NIU will not get blown out by 1-loss Missouri or X. Does that satisfy your concerns? If not, why not?

And this brings up a question others have noted. As I understand it, Sagarin runs at least two models, one for the BCS, and one for prediction. Since the BCS one is less accurate, why are any of those values (like SoS, for example) favored in this conversation? I apologize that this conversation is newer to me than many of you, but I am getting good info from it.

[billybud]

If you are in the MAC...schedule and beat good teams in OOC...that's all you have to do. Beat some good teams, not fodder.

Beat Ohio State (a MAC team has never done so in about a score of attempts), beat a ranked team or two.

We all saw Ball State go 12-0 on a weak schedule, get ranked high, and get blown away by Tulsa in the bowl game.

[Spence]

BB. Since Sagrin does start the year using past results, wouldn't that make me right? As you know, I don't have a problem with using past results. It generally gives you a good impression of what is happening this year. I also agree that all wins aren't equal. I don't think anyone would disagree with that. But even if a team plays the weakest team in the country, they can still be a great team. If all the teams in the ACC this year are at the bottom of all cfb, that doesn't change the fact that Florida State isn't a great team. It just makes them harder to identify.

I don't think that letting an undefeated team in the natioinal championship that plays in a traditionally weak conference is the answer. I think if you are going ahead with playoffs, you have a tournament of champions. It won't necessarily give you the best line up of teams, in fact in guarantees that you don't have the best teams, but at least in this format you have given every one a fair chance to earn a right to play for the national championship. It will also eliminate the need for subjective rankings and public opinion polls.