The Letdown - Let's do some research!

[SFBadger96]

I would suggest a better metric is to look at point spreads. What causes a team that is a 7+point favorite to lose its bowl game?

Getting away from the beauty contests of the rankings and CFP selection, the odds makers have a lot of data to base their spreads on by the time we reach bowl season. They also take into account who is and who isn't playing, injury reports, and god knows what else. And their profits are related to how good they are at setting those spreads. 

Upsets happen in all sports, but is there any correlation between upsets in college football, the relative disparity of teams (K-State/Purdue being an easy example), and the way the favorite ended its season?

Games where the spread is close going either way shouldn't surprise anyone, but games where there is enough data to support more than a 1-score difference suggest something else happened on that field. Now, if there number of upsets isn't any different than for any other type of match-up, it suggests nothing unique about these games. But worth looking at. Also, how big the upset victory was, vs where the spread was.

Wisconsin has a couple of signature wins against the odds: 1999 Rose Bowl (vs. UCLA), 2006 Capital One Bowl (vs. Auburn) come to mind. Don't know how they figure against the spreads.

[utee94]

Wisconsin has a couple of signature wins against the odds: 1999 Rose Bowl (vs. UCLA), 2006 Capital One Bowl (vs. Auburn) come to mind. Don't know how they figure against the spreads.

Obviously, UCLA and Auburn were bitterly disappointed and didn't want to be there, which was the sole reason for their losses.


And yes, point spreads might be an interesting metric.  bwar already suggested score differentials as a potential way to assess.

I'm a little wary of using point spreads produced by betting houses, because they're not actually designed to be representative of an expected point differential, but rather they're created to entice equal betting on each side.  So they're SORT OF representative of what "the people" believe is an appropriate final point differential, but the added complexity of the betting house attempting to generate an even monetary distribution concerns me somewhat.  I guess it might only be worth a point or two of error in either direction?  But that point or two in either direction could change the outcome of the experiment.

[Riffraft]

If we include all of the "disappointed teams" and the result flips from 12-15 to 22-19 or something like that, does that flip your view on whether it's a "real thing" or not?

Should it?

The problem is you have yet to provide an objective criteria to pull out disappointed teams, where while you may object to the definition at least OEM has provided criteria that is objective. So you think his criteria is wrong, give some criteria that can be used to pull teams out. Otherwise quit complaining about his. 

[betarhoalphadelta]

And yes, point spreads might be an interesting metric.  bwar already suggested score differentials as a potential way to assess.

I'm a little wary of using point spreads produced by betting houses, because they're not actually designed to be representative of an expected point differential, but rather they're created to entice equal betting on each side.  So they're SORT OF representative of what "the people" believe is an appropriate final point differential, but the added complexity of the betting house attempting to generate an even monetary distribution concerns me somewhat.  I guess it might only be worth a point or two of error in either direction?  But that point or two in either direction could change the outcome of the experiment.

Yeah, I have the same concern about point spreads. Also have concern about the idea I floated, which is using historical ranking differential to generate expected point differential.
I think the ideal would be to use S&P. I believe that is actually designed to be a predictor. Only issue is that I'm not sure it goes back far enough to give us a reasonable sample size, and also not sure how easy it is to pull out the S&P ratings PRIOR to the bowl game as opposed to the final post-bowl season S&P, which would affect the validity. 

[utee94]

The problem is you have yet to provide an objective criteria to pull out disappointed teams, where while you may object to the definition at least OEM has provided criteria that is objective. So you think his criteria is wrong, give some criteria that can be used to pull teams out. Otherwise quit complaining about his.

In a scientific experiment you simply can't say, "Well the data set is flawed, statistically insignificant, and also not representative, but it's all we have so we're going to run with it anyway."  That's just bad science.
If it's your opinion that the criteria outlined by OAM are sufficient to be representative, you can state that, and we can debate that. I clearly wouldn't agree, but at least that debate is better than you telling me I should shut up and go away.

And MarqHusker offered a list of teams that he felt should be included on the other thread.  I've stated that's a good start.  I'm not sure what the objective criterion would look like, perhaps "any team excluded from the BCS/playoff that was withing 4 spots" or something?  I'm not saying it's easy.  But I am saying that I don't believe OAM's narrow criteria are sufficient, whether they're objective or not.

So I'll continue to complain about it, thank you very much.

[ELA]

I also don't think anyone is saying it's never a factor, and that it doesn't play some role.  But my issue is in too narrowly defining disappointing teams and negating other factors.  As I pointed out, Michigan and Georgia both lost to the majority of top teams they faced this year (Michigan lost to all of them).  To fit your definition, a team has to lose its final two games, at the time of year they are likely facing their best opponents.  To me, it is just as likely as those teams getting exposed by the toughest part of their schedule.

Take a look at a couple of teams that should have met your definition, except for some fortunate bounces, 2001 Nebraska and 2003 Oklahoma.  Both should have been knocked out by season ending losses, but both managed to find their way into a national title shot.  Both fell flat a second straight time on the big stage.  It seems just as likely to me that teams who lose a big game in the finale, and then follow it with a bowl loss did so because it's as simple as they were overrated to begin with.  Is that every time?  No.  But I'm guessing it's just as much of a factor, and the "data" is probably just as supportive.

[utee94]

I also don't think anyone is saying it's never a factor, and that it doesn't play some role.  But my issue is in too narrowly defining disappointing teams and negating other factors.  As I pointed out, Michigan and Georgia both lost to the majority of top teams they faced this year (Michigan lost to all of them).  To fit your definition, a team has to lose its final two games, at the time of year they are likely facing their best opponents.  To me, it is just as likely as those teams getting exposed by the toughest part of their schedule.

Take a look at a couple of teams that should have met your definition, except for some fortunate bounces, 2001 Nebraska and 2003 Oklahoma.  Both should have been knocked out by season ending losses, but both managed to find their way into a national title shot.  Both fell flat a second straight time on the big stage.  It seems just as likely to me that teams who lose a big game in the finale, and then follow it with a bowl loss did so because it's as simple as they were overrated to begin with.  Is that every time?  No.  But I'm guessing it's just as much of a factor, and the "data" is probably just as supportive.

Extremely well said. 

[betarhoalphadelta]

If it's your opinion that the criteria outlined by OAM are sufficient to be representative, you can state that, and we can debate that. I clearly wouldn't agree, but at least that debate is better than you telling me I should shut up and go away.

And MarqHusker offered a list of teams that he felt should be included on the other thread.  I've stated that's a good start.  I'm not sure what the objective criterion would look like, perhaps "any team excluded from the BCS/playoff that was withing 4 spots" or something?  I'm not saying it's easy.  But I am saying that I don't believe OAM's narrow criteria are sufficient, whether they're objective or not.

The problem with a proposed criteria such as what you postulated is that it would include teams that COMPLETELY missed out on the entire idea of "letdown". Say, a 2-loss team wins its CCG to go from #9 (5 spots out of playoff) to #6 (2 spots out of playoff). What would they be let down or disappointed by?
The biggest critique in my opinion of OAM's data set is small sample size. And that small sample size came from restricting his data set to things that meet his definition of a letdown: a team that is basically in the CFP with a win, but loses their last game to be excluded. 
I think his criteria is actually quite good. I think the downside of that criteria is that not enough teams meet it to be truly statistically significant, at least upon the narrow grounds of mere W/L as the result. I think if we could have a more granular measurement such as score differential, it would give us a lot more info.

[betarhoalphadelta]

I also don't think anyone is saying it's never a factor, and that it doesn't play some role.  But my issue is in too narrowly defining disappointing teams and negating other factors.  As I pointed out, Michigan and Georgia both lost to the majority of top teams they faced this year (Michigan lost to all of them).  To fit your definition, a team has to lose its final two games, at the time of year they are likely facing their best opponents.  To me, it is just as likely as those teams getting exposed by the toughest part of their schedule.

Take a look at a couple of teams that should have met your definition, except for some fortunate bounces, 2001 Nebraska and 2003 Oklahoma.  Both should have been knocked out by season ending losses, but both managed to find their way into a national title shot.  Both fell flat a second straight time on the big stage.  It seems just as likely to me that teams who lose a big game in the finale, and then follow it with a bowl loss did so because it's as simple as they were overrated to begin with.  Is that every time?  No.  But I'm guessing it's just as much of a factor, and the "data" is probably just as supportive.

So then you're saying that determining MNC, BCSCG participants, or CFP participants, is a horribly flawed beauty contest where the voters / committee routinely get these things so wrong that they're overrating teams on a regular enough basis to throw off all of OAM's data set?
Sounds like a great way to crown a "champion" 

[utee94]

The problem with a proposed criteria such as what you postulated is that it would include teams that COMPLETELY missed out on the entire idea of "letdown". Say, a 2-loss team wins its CCG to go from #9 (5 spots out of playoff) to #6 (2 spots out of playoff). What would they be let down or disappointed by?
The biggest critique in my opinion of OAM's data set is small sample size. And that small sample size came from restricting his data set to things that meet his definition of a letdown: a team that is basically in the CFP with a win, but loses their last game to be excluded.
I think his criteria is actually quite good. I think the downside of that criteria is that not enough teams meet it to be truly statistically significant, at least upon the narrow grounds of mere W/L as the result. I think if we could have a more granular measurement such as score differential, it would give us a lot more info.

I understand that is YOUR biggest concern.
It is only one of mine. 
We've pretty much beaten this one to death.

[OrangeAfroMan]

I'm just sort of dumbfounded that the suggestion being offered to me is to take my admittedly small sample size and stuff it full of much more diverse, subjective data to make it more valid.


WHAT???

[OrangeAfroMan]

For a team to be "left out" of the playoff or BCSNCG because of a tie-breaker or rankings or even an early-season loss, you have someone else to blame.  You can blame the voters, the system, the baby Jesus,..whoever you want.  You feel cheated.  You learn that life is unfair.  You're probably pissed.




For a team to lose their last game and go from IN to OUT, it's on them.  They have no one else to blame.  THAT, is perhaps, the specific difference being glossed over here.  To fail when you had the power to achieve your season-long goals is an especially damning thing, imo.  When you fail in that way, you're broken.  Emasculated.




If you don't see that colossal difference, I have nothing left to offer.

[TyphonInc]

I observed teams tending to lose an inordinate percentage of the time, in a specific situation.  The fact you're labeling that so negatively is odd.  What I observed set the parameters, not my bias.

I've observed almost exactly half the bowl teams losing their last game. Can we analyze the significance of that disappointment?

[utee94]

For a team to be "left out" of the playoff or BCSNCG because of a tie-breaker or rankings or even an early-season loss, you have someone else to blame.  You can blame the voters, the system, the baby Jesus,..whoever you want.  You feel cheated.  You learn that life is unfair.  You're probably pissed.




For a team to lose their last game and go from IN to OUT, it's on them.  They have no one else to blame.  THAT, is perhaps, the specific difference being glossed over here.  To fail when you had the power to achieve your season-long goals is an especially damning thing, imo.  When you fail in that way, you're broken.  Emasculated.




If you don't see that colossal difference, I have nothing left to offer.

You asked for reasons to dispute what, in your opinion, is a trend.  This was your exact question:
" If someone else has a better explanation for these teams having a losing record, when most of them were favored/higher ranked in their bowl, I'm all ears."
Several individuals have offered multiple possibilities for disputing your opinion.  Those challenges consist of:
1) A non-representative sample that does not include the appropriate teams that represent the class you're attempting to differentiate and identify
2) A sample that is too small to be statistically significant
3) The idea that the "let-down losers" are often simply overrated, as they've already demonstrated the capability of losing to other ranked teams, so their losses in the bowl games are neither surprising nor notable.  This doesn't necessarily apply in all cases, such as KSU 1998, but it certainly applies to this year's batch of let-down losers, as Michigan failed to beat ANY of the three teams it faced that will complete the season ranked in the top 15, and Georgia lost to 3 of the 5 teams it faced that will complete the season ranked in the Top 15.
4) The idea that there are potentially other numerous contributing factors, such as "let-down losers" simply encountering their toughest stretch of opponents at the end of the season in rivalry games, CCGs, etc. Again, making it neither notable nor surprising that they would also lose to another ranked team in a bowl to close out their season.

You apparently have already made up your mind and are completely closed off to debate.  So yes, I'd say you have nothing left to offer.