CFP Discussion

[ibeattetris]

Of course the same computers put Georgia in over Alabama after they just lost to Alabama in the Conference Championship Game. Defend that logic!

What computer picked UGA over Alabama this year? (below chosen as they were the BCS computer polls)
Sagarin Bama > UGA http://sagarin.com/sports/cfsend.htm
Colley Bama > Uga https://www.colleyrankings.com/currank.html
Massey Bama > Uga https://masseyratings.com/cf/ncaa-d1/ratings
Wolfe Bama > Uga http://prwolfe.bol.ucla.edu/cfootball/ratings.htm

I'm fine with us getting mad about things (I sure as heck have been mad about the FSU snub since Sunday), but we should at least base them on reality. The biggest problem with all of the above is what I said before, all of the computer polls (except Colley) removed their BCS only ranking scheme because MOV based rankings are stronger predictors than Win/Loss only. It's possible the final real rankings would be different, but we can never know.

 

[Root4GT]

There are lots of ways to pick teams. There are lots of reasons to pick UGA over Alabama. We could easily set the criteria so that the loser of the championship game is out, though

But having an ad-hoc and inconsistent committee process for selecting teams with vague and subjective rules is about as bad as you can get.

If a computer model was secret and proprietary, that would also be terrible.

How would people here like it if our income decided at the end of the year by a committee with vague rules that weren’t clear ahead of time—especially if you met your goals but one of your teammates got sick at the end of the year?

How do you think the Public would react if the 4 CFP teams were picked by a computer ranking and they picked Georgia the Day after Georgia LOST to Alabama in at SEC Conference Championship Game and they both had only one loss. The sporting world would lose their minds and rightfully so!

The lack of transparency in the Committee inner workings is a problem. The lack of actual football expertise with the committee members is a problem. The influence of ESPN is a problem. The current system is chock full of problems.

Those problems pale in comparison to a computer system who picks the team that just lost a game the day before in a Conference Championship Game over the team that beat them.

GT grads often think everything can be solved with good algorithms. That is 100% false when there is not enough data for reasonable analysis. That is the case in College Football at the FBS level. Comparative data is not nearly good enough to make valid decisions.

Have you heard many complaints about the FCS playoffs? They have 24 teams with 10 automatic bids and the rest chosen at large by a committee. The committee also does the seeding. They actually get a real Champion!

Input from computer rankings makes sense as there are many to choose from. Humans need to be involved.

 

[jojatk]

There are lots of ways to pick teams. There are lots of reasons to pick UGA over Alabama. We could easily set the criteria so that the loser of the championship game is out, though

But having an ad-hoc and inconsistent committee process for selecting teams with vague and subjective rules is about as bad as you can get.

If a computer model was secret and proprietary, that would also be terrible.

How would people here like it if our income decided at the end of the year by a committee with vague rules that weren’t clear ahead of time—especially if you met your goals but one of your teammates got sick at the end of the year?

Yeah as it has turned out this is almost no different than the national champions picked by the polls years ago. The main difference is it's a smaller group of people who can have enormous pressure put on them by various constituencies to vote in certain ways and, as you said, have vague and subjective "rules" for how they are allowed to make their decisions.

 

[ibeattetris]

Those problems pale in comparison to a computer system who picks the team that just lost a game the day before in a Conference Championship Game over the team that beat them.

This literally did not happen.

Have you heard many complaints about the FCS playoffs? They have 24 teams with 10 automatic bids and the rest chosen at large by a committee. The committee also does the seeding. They actually get a real Champion!

The number of people who care about FCS is too small. I promise you, the first time a 11-2 GT team that loses its conference champions get skipped by a 9-3 SEC to be the number 12 team, you will hear complaints, and they will be certainly louder when it's a paid ESPN shadow council.

 

[Root4GT]

What computer picked UGA over Alabama this year? (below chosen as they were the BCS computer polls)
Sagarin Bama > UGA http://sagarin.com/sports/cfsend.htm
Colley Bama > Uga https://www.colleyrankings.com/currank.html
Massey Bama > Uga https://masseyratings.com/cf/ncaa-d1/ratings
Wolfe Bama > Uga http://prwolfe.bol.ucla.edu/cfootball/ratings.htm

I'm fine with us getting mad about things (I sure as heck have been mad about the FSU snub since Sunday), but we should at least base them on reality. The biggest problem with all of the above is what I said before, all of the computer polls (except Colley) removed their BCS only ranking scheme because MOV based rankings are stronger predictors than Win/Loss only. It's possible the final real rankings would be different, but we can never know.

Your links do not support rational seedings. One has Ohio State in the top 4. How would that sit with everyone.

CONGROVE COMPUTER RANKINGS (FBS

 

[ibeattetris]

Your links do not support rational seedings. One has Ohio State in the top 4. How would that sit with everyone.

CONGROVE COMPUTER RANKINGS (FBS

Please, for the love of god, read all of my post and apply rational thoughts to it. For starters, each of those polls is not used independently, a composite is made. There are plenty of reasons OSU could be a top 4 team when you include margin of victory as part of the equation. Their only loss is to the near consensus #1 team. Trying to determine a bottom for OSU when they have only one loss to the best team is hard. Secondly, all of the computer polls combined were equal to 33% of the final BCS poll. The humans have a bigger impact than any of the computers do. Thirdly as I mentioned all of my links are not using the true BCS computer polls since they do not exist anymore. Is OSU still ranked so high when losses and wins are treated equal? Does FSU improve when all of their end of season close wins are treated the same as other team's blow outs?

The biggest problem with all of the above is what I said before, all of the computer polls (except Colley) removed their BCS only ranking scheme because MOV based rankings are stronger predictors than Win/Loss only. It's possible the final real rankings would be different, but we can never know.

You are basically yelling at the clouds, since as was brought up earlier, even with MOV based computer polls would have had FSU in the final 4 with Bama above UGA. I think it gets the true final four wrong, but I also have said all along that 4 was a failure of a decision from the beginning. The top 12 overall looks pretty good to me.

CONGROVE COMPUTER RANKINGS (FBS

Why are you linking me some irrelevant poll? Any bozo can write a regression algorithm and post the results on a website.

 

[stinger78]

GT grads often think everything can be solved with good algorithms. That is 100% false when there is not enough data for reasonable analysis. That is the case in College Football at the FBS level. Comparative data is not nearly good enough to make valid decisions.

Input from computer rankings makes sense as there are many to choose from. Humans need to be involved.

Humans are fine as long as they have set metrics, those metrics are transparent, and the committee adheres to the metrics. Every year, it seems, we hear of a different criterion being applied that circumvents a rather straight-forward metric. The compounding issue here is the way sEcSPN openly comports itself. It clearly fluffs up the SECheat anytime it can, and it pumps more money into their coffers annually than any other conference. So then, when the back room deals and deviations from metrics occur, and the almost inevitable benefit goes to the SECheat, there is an understandable distrust of the process. Why did it change, and why does it seemingly always fall in the favor of the SECheat? It's so common that many were mockingly predicting that it was going to happen before it happened. That simply has to be fixed.

 

[stinger78]

Please, for the love of god, read all of my post and apply rational thoughts to it. For starters, each of those polls is not used independently, a composite is made. There are plenty of reasons OSU could be a top 4 team when you include margin of victory as part of the equation. Their only loss is to the near consensus #1 team. Trying to determine a bottom for OSU when they have only one loss to the best team is hard. Secondly, all of the computer polls combined were equal to 33% of the final BCS poll. The humans have a bigger impact than any of the computers do. Thirdly as I mentioned all of my links are not using the true BCS computer polls since they do not exist anymore. Is OSU still ranked so high when losses and wins are treated equal? Does FSU improve when all of their end of season close wins are treated the same as other team's blow outs?


You are basically yelling at the clouds, since as was brought up earlier, even with MOV based computer polls would have had FSU in the final 4 with Bama above UGA. I think it gets the true final four wrong, but I also have said all along that 4 was a failure of a decision from the beginning. The top 12 overall looks pretty good to me.
View attachment 15322

Why are you linking me some irrelevant poll? Any bozo can write a regression algorithm and post the results on a website.

I would not agree that MOV provides a more robust analysis. The reason is too many factors affect MOV. Suppose two otherwise equal teams (A & B) play the same third team (C). Team A wins by 4 TD while Team B wins by 2 TD. I would suggest that the MOV cannot be used to determine that Team A is the better team. The reason(s):
1. Suppose that the Team A-C game was played in early October on a sunny 70-degree day and the Team B-C game was played in late November in a snowstorm?
2. Suppose that the Team A-C game involved Team A's QB leaving with a concussion protocol (or other major injury) while the Team B-C game was injury free.
3. Suppose Team A had a bye week coming up and played his starters well into the 4th quarter, while Team B had a tough next game, pulled his starters, and ran out the clock in the 4th.
4. Rinse repeat with a myriad of possible confounding factors.

The only MOV metric that I would accept is the use of AVG MOV over all games. In that way, the confounding factors would tend to average out over the course of a season, but that is not assured.

 

[ibeattetris]

I would not agree that MOV provides a more robust analysis. The reason is too many factors affect MOV. Suppose two otherwise equal teams (A & B) play the same third team (C). Team A wins by 4 TD while Team B wins by 2 TD. I would suggest that the MOV cannot be used to determine that Team A is the better team. The reason(s):
1. Suppose that the Team A-C game was played in early October on a sunny 70-degree day and the Team B-C game was played in late November in a snowstorm?
2. Suppose that the Team A-C game involved Team A's QB leaving with a concussion protocol (or other major injury) while the Team B-C game was injury free.
3. Suppose Team A had a bye week coming up and played his starters well into the 4th quarter, while Team B had a tough next game, pulled his starters, and ran out the clock in the 4th.
4. Rinse repeat with a myriad of possible confounding factors.

The only MOV metric that I would accept is the use of AVG MOV over all games. In that way, the confounding factors would tend to average out over the course of a season, but that is not assured.

I am not saying MOV is all that is required to generate the best predictors, but a predictor using MOV is stronger than a predictor that only uses win/loss. I don't even think a "predictor" is what should be used for defining a poll (it's just all I have since the BCS polls stopped producing their BCS specific rankings), which is why I have made the caveat multiple times that the computers I linked were not true bcs. Even though I hate espn, I do like their "resume" rank which tries to determine a probability that a top random 25 team would have a better record than the given team's record. I think that type of approach is better in determining polling than something influenced by points. https://www.espn.com/college-football/fpi/_/view/resume
It has its own issues since it doesn't consider head to head. That's also why I like the approach of using composite polls to allow for different methodologies. My favorite poll overall is reddit's: https://poll.redditcfb.com/ It's a composite of around 300 polls, some human, some computer, and some hybrid. As with any poll that includes humans, it can be influenced by media at times, but I think it does a better job than most and includes analysis of user responses which can get you kicked out.

 

[stinger78]

I am not saying MOV is all that is required to generate the best predictors, but a predictor using MOV is stronger than a predictor that only uses win/loss. I don't even think a "predictor" is what should be used for defining a poll (it's just all I have since the BCS polls stopped producing their BCS specific rankings), which is why I have made the caveat multiple times that the computers I linked were not true bcs. Even though I hate espn, I do like their "resume" rank which tries to determine a probability that a top random 25 team would have a better record than the given team's record. I think that type of approach is better in determining polling than something influenced by points. https://www.espn.com/college-football/fpi/_/view/resume
It has its own issues since it doesn't consider head to head. That's also why I like the approach of using composite polls to allow for different methodologies. My favorite poll overall is reddit's: https://poll.redditcfb.com/ It's a composite of around 300 polls, some human, some computer, and some hybrid. As with any poll that includes humans, it can be influenced by media at times, but I think it does a better job than most and includes analysis of user responses which can get you kicked out.

I'm not against MOV, I just think it has to be used in a way that minimizes the error that it introduces. As you no doubt know, any composite factor used as an indicator forming or reflecting a variable in a linear regression introduces error. That is, that factor doesn't represent the variable exactly. I think error can be mitigated by using average MOV rather than specific game by game MOV. MOV has usefulness and can make an analysis more robust, but only if the error is mitigated. Just my feeling on the matter.

 

[Root4GT]

Please, for the love of god, read all of my post and apply rational thoughts to it. For starters, each of those polls is not used independently, a composite is made. There are plenty of reasons OSU could be a top 4 team when you include margin of victory as part of the equation. Their only loss is to the near consensus #1 team. Trying to determine a bottom for OSU when they have only one loss to the best team is hard. Secondly, all of the computer polls combined were equal to 33% of the final BCS poll. The humans have a bigger impact than any of the computers do. Thirdly as I mentioned all of my links are not using the true BCS computer polls since they do not exist anymore. Is OSU still ranked so high when losses and wins are treated equal? Does FSU improve when all of their end of season close wins are treated the same as other team's blow outs?


You are basically yelling at the clouds, since as was brought up earlier, even with MOV based computer polls would have had FSU in the final 4 with Bama above UGA. I think it gets the true final four wrong, but I also have said all along that 4 was a failure of a decision from the beginning. The top 12 overall looks pretty good to me.
View attachment 15322

Why are you linking me some irrelevant poll? Any bozo can write a regression algorithm and post the results on a website.

Your Sarigan link was pre championship game. My point is computers have built in bias just as humans. We had some terrible BCS selections during the BCS Era.

The BCS was dumped as it sucked. The CFP made a bad decision this year based on the on field performance. They clearly were influenced by ESPN and TV ratings concerns.

There is no good way to pick 4 teams this year. Cases can be made for 7 teams.

This all goes away next year.

Thank Gif the BCS is done. Thank Gif the 4 team CFP is done.

Use computers as aids. Never let them make the final decision.

 

[slugboy]

Your links do not support rational seedings. One has Ohio State in the top 4. How would that sit with everyone.

CONGROVE COMPUTER RANKINGS (FBS

You keep bringing up red herrings.

1. Make a set of clear rules
2. Try to make them fair rules
3. Publish them publicly in advance
4. Follow the rules
5. The rules decide—not opinions
6. Don’t use arbitrary or hidden processes

Before the first game of the season, everyone should know their path to the CFP. If the decision process is a computer margin of victory normalized vs strength of schedule, then everyone knows what they have to do, and their fate is in their own hands.

If it’s a different set of rules, like the NFL tiebreaker rules, that works too.

Last weekend, the CFP committee was capricious. That robs every team of their free choice and ability to chart their own destiny.

The players and the teams should be able to earn their way in. Period

 

[stinger78]

Your Sarigan link was pre championship game. My point is computers have built in bias just as humans. We had some terrible BCS selections during the BCS Era.

The BCS was dumped as it sucked. The CFP made a bad decision this year based on the on field performance. They clearly were influenced by ESPN and TV ratings concerns.

There is no good way to pick 4 teams this year. Cases can be made for 7 teams.

This all goes away next year.

Thank Gif the BCS is done. Thank Gif the 4 team CFP is done.

Use computers as aids. Never let them make the final decision.

There is a great way to pick 4 teams this year. Start with the 3 undefeated P5 conference champs then pick one of the 2 once-defeated P5 conference champs - the winner of the head-to-head matchup.

Very simple. Very rational. Very defensible.

Would there be gripes? Always.

The answer is: There were 3 undefeated P5 conference champions who were invited, and the two once-defeated P5 conference champions played each other this season. The winner got the invitation.

Case closed.

 

[Techster]

I don't think they understood the ferocity of a rabid fanbase.

The CFP website has a contact page, where anyone can find the emails and phone numbers of committee members, and an address listed. The page had to be removed after enough FSU fans bombarded the committee with letters, phone calls, and emails voicing their outrage.

 

[Root4GT]

There is a great way to pick 4 teams this year. Start with the 3 undefeated P5 conference champs then pick one of the 2 once-defeated P5 conference champs - the winner of the head-to-head matchup.

Very simple. Very rational. Very defensible.

Would there be gripes? Always.

The answer is: There were 3 undefeated P5 conference champions who were invited, and the two once-defeated P5 conference champions played each other this season. The winner got the invitation.

Case closed.

Yup, however, money always rules. ESPN knows Alabama vs Michigan is a huge TV rating. Better than FSU. Committee is flawed and no doubt influenced by the check writers of College Football.

 

[stinger78]

Yup, however, money always rules. ESPN knows Alabama vs Michigan is a huge TV rating. Better than FSU. Committee is flawed and no doubt influenced by the check writers of College Football.

Sure, but you said, "There is no good way to pick 4 teams this year."

Sure there is. See above.

In fact, I would say it was one of the easier years, TBH. The money aspect screwed up everything.

 

[Root4GT]

Sure, but you said, "There is no good way to pick 4 teams this year."

Sure there is. See above.

In fact, I would say it was one of the easier years, TBH. The money aspect screwed up everything.

 

[Root4GT]

This year would have been good for an 8 team playoff.

Locking into a computer generated decision has failed in the BCS Era several times. The committee failed this time in my opinion.

None of it matters as thank God we are dumping this bad system. 4 teams based on a BCS system would still be bad.

It would be nice to have the FCS middle but that would never happen. Bowls still have TV money as they fill air time. 82 teams vs 24 teams.

FBS really doesn’t have a true champion. Just the winner of a 4 team tournament.

 

[Yaller Jacket]

So we hear that the selection committee was told to try to pick the four best teams to ensure competitive games and thus good tv ratings. Who exactly did the telling? It gripes me to no end that the goal of the playoffs is now to have the highest ratings instead of finding the champion.

 

[cpf2001]

You keep bringing up red herrings.

1. Make a set of clear rules
2. Try to make them fair rules
3. Publish them publicly in advance
4. Follow the rules
5. The rules decide—not opinions
6. Don’t use arbitrary or hidden processes

Before the first game of the season, everyone should know their path to the CFP. If the decision process is a computer margin of victory normalized vs strength of schedule, then everyone knows what they have to do, and their fate is in their own hands.

If it’s a different set of rules, like the NFL tiebreaker rules, that works too.

Last weekend, the CFP committee was capricious. That robs every team of their free choice and ability to chart their own destiny.

The players and the teams should be able to earn their way in. Period

Eh, SOS is way too out of anyone's hands. Matchups are often set years in advance for OOC games, and for conference games you have even less control over how good your opponents are in any given year. The drumbeat against the computers not being able to account for factors like that in the BCS era was constant - nobody really thought stuff like that was fair or that teams fully had their fate in their own hands. It's basically the USC/LSU scenario, where the voters got mad at the computers for favoring the SEC and Big XII strength of schedules and ignoring that OU lost it's conference championship game badly. An inverted version of this year - "three undefeated teams, computer picks two of them, one of those picked teams lost their QB in the last weeks" - would've caused people to absolutely lose their minds too if the BCS was still around.

I'm not expecting everyone to be happy with FSU getting left out (though: screw them, personally ) but arguing against the idea that the BCS was good somehow is a good way to pass the time waiting for some stuff to run for work.