PED Differential: BYU vs Utah 2009

[Indy Coug]

Here's some background info on PED Differential

http://www.cougarboard.com/noframes/...tml?id=2279662


Offense Pass Efficiency

BYU - 167.69
Utah - 136.49

Defense Pass Efficiency

BYU - 127.22
Utah - 100.84


Pass Efficiency Differential

BYU - 40.47
Utah - 35.65


PED Differential: 4.82


Factor that in with home field advantage and BYU should be around a 4.5 point favorite.

[Indy Coug]

In the same vein, let's look at the top teams in Pass Efficiency Differential. This is a leading indicator of who wins championships:

http://www.cougarboard.com/noframes/...tml?id=4340007

This year, here are the top 6 teams in PED:

Florida: 75.69 (11-0)
TCU: 63.82 (11-0)
Boise State: 63.77 (11-0)
Cincinnati: 57.26 (10-0)
Virginia Tech: 54.68 (8-3)
Alabama: 54.42 (11-0)
Texas: 50.43 (11-0)

All six undefeated teams are in the top seven.

[Indy Coug]

I guess I should have posted something about fat and drunk Ute fans instead.

[cougjunkie]

Sorry Indy I missed this post.

I was looking at Pass Efficiency last night, not specifically PED though. Max Hall has a legit chance to finish 1st in the country in Pass Efficiency. Also BYUs secondary has not been as bad as years past. Can you post the PED for BYU the last 5 years or so?

[Indy Coug]

Sorry Indy I missed this post.

I was looking at Pass Efficiency last night, not specifically PED though. Max Hall has a legit chance to finish 1st in the country in Pass Efficiency. Also BYUs secondary has not been as bad as years past. Can you post the PED for BYU the last 5 years or so?

Season	Name	PED
2000	BYU	-26.54
2001	BYU	40.51
2002	BYU	4.27
2003	BYU	-10.03
2004	BYU	4.73
2005	BYU	-7.65
2006	BYU	60.85
2007	BYU	25.89
2008	BYU	27.12

[cougjunkie]

Season	Name	PED
2000	BYU	-26.54
2001	BYU	40.51
2002	BYU	4.27
2003	BYU	-10.03
2004	BYU	4.73
2005	BYU	-7.65
2006	BYU	60.85
2007	BYU	25.89
2008	BYU	27.12

So I was right this is our 2nd best team since 2000 regarding PED. Do you keep track of Utahs PED at all?

[Indy Coug]

Season	Name	PED
2000	Utah	3.83
2001	Utah	20.41
2002	Utah	-1.63
2003	Utah	30.71
2004	Utah	64.48
2005	Utah	29.40
2006	Utah	0.09
2007	Utah	28.68
2008	Utah	42.11

[jay santos]

Indy, I respect you which is why I don't get why you think this PED differential stat is meaningful. It's pretty much a proxy for MOV and has no SOS to it. It's weaker than weak sauce in terms of a predictive model.

[Indy Coug]

Indy, I respect you which is why I don't get why you think this PED differential stat is meaningful. It's pretty much a proxy for MOV and has no SOS to it. It's weaker than weak sauce in terms of a predictive model.

How is it a "proxy" for MOV? 90% of the pass efficiency value comes from the completion percentage and yards per attempt components.

As far as predictive value goes, take a look at the link I provided in the first post and notice the nearly linear PED Diff to MOV relationship.

[jay santos]

How is it a "proxy" for MOV? 90% of the pass efficiency value comes from the completion percentage and yards per attempt components.

As far as predictive value goes, take a look at the link I provided in the first post and notice the nearly linear PED Diff to MOV relationship.

Of course PED Diff correlates to MOV. That's my point. Am I using the word proxy wrong?

http://en.wikipedia.org/wiki/Proxy_(statistics)

You might as well use MOV. And if you use MOV, you might as well use any one of a thousand computer models that incorporate both MOV and SOS.

[Indy Coug]

Of course PED Diff correlates to MOV. That's my point. Am I using the word proxy wrong?

http://en.wikipedia.org/wiki/Proxy_(statistics)

You might as well use MOV. And if you use MOV, you might as well use any one of a thousand computer models that incorporate both MOV and SOS.

When you were talking about PED being a proxy for MOV I thought that you were saying that the scoring component (TD%) had an undue weighting in the pass efficiency formula.

So what you're saying is that whoever has the most points wins? That's pure genius. The whole point of PED is showing WHAT is the PRIMARY driver of MOV.

[pelagius]

Of course PED Diff correlates to MOV. That's my point. Am I using the word proxy wrong?

Yes and No. Indy isn't using it as a proxy for MOV. In his post he is using PED differential as a predictor of margin of victory. Which is different than using it as a proxy.

In order for PED differential to be just a proxy for MOV you need to show that PED differential is correlated with past MOV (MOV of the games used to compute the PED differential) and Past MOV (in some form, for example a MOV computer model) subsumes the predictive ability of PED differential. This doesn't just mean that past MOV as a predictor would just need to be better on average than PED differential but that after taking into account past MOV, PED differential doesn't have statistically significant predictive power anymore. Based on what's been posted one can't evaluate the hypothesis at all (although it is not obvious to me that PED differential would entirely be subsumed by MOV but it might be).

[jay santos]

Yes and No. Indy isn't using it as a proxy for MOV. In his post he is using PED differential as a predictor of margin of victory. Which is different than using it as a proxy.

In order for PED differential to be just a proxy for MOV you need to show that PED differential is correlated with past MOV (MOV of the games used to compute the PED differential) and Past MOV (in some form, for example a MOV computer model) subsumes the predictive ability of PED differential. This doesn't just mean that past MOV as a predictor would just need to be better on average than PED differential but that after taking into account past MOV, PED differential doesn't have statistically significant predictive power anymore. Based on what's been posted one can't evaluate the hypothesis at all (although it is not obvious to me that PED differential would entirely be subsumed by MOV but it might be).

Based on my understanding of college football models and the data, I can't see how PED differential could be any more meaningful than MOV. Any "differential" is getting close to MOV. Run differential, pass differential, total offense differential. It's all a way of measuring a team's combined offensive power and defensive power. I see no possible way that it can be nothing other than a proxy for MOV, and a bad one at that.

[pelagius]

Based on my understanding of college football models and the data, I can't see how PED differential could be any more meaningful than MOV. Any "differential" is getting close to MOV. Run differential, pass differential, total offense differential. It's all a weigh of measuring a team's combined offensive power and defensive power. I see no possible way that it can be nothing other than a proxy for MOV, and a bad one at that.

Yours is a perfectly plausible hypothesis (although I can see one arguing that PED differential would be less affected by luck and consequently would have some incremental predictive power ... it is also less discrete) ... Ultimately it is an empirical question. If one had the data it would not be a terrible hard hypothesis to test.

[Indy Coug]

Let me try this again. There are a limited number of "official" statistical measures you will find tracked by the NCAA (rushing/passing yards, turnovers, penalties, etc. etc. etc.). The point I'm trying to make is that pass efficiency (one of the official measures) has the greatest correlation with MOV of all the official measures you want to stack up against it. Of all the official statistics:

• Pass efficiency has the highest correlation with average points scored.
  
• Pass efficiency differential has the highest correlation with average MOV.
  
• Pass efficiency differential-differential has the highest correlation with predicted MOV.

This does not deny the existence of models that utilize entirely new, unofficial statistics (that go beyond simple differencing), such as a new pass efficiency formula that has different weightings or some model that is a massive composite of all of the official statistics and so forth.