Writers of Pro Football Prospectus 2008

13 Dec 2005

In Numbers Game, Aikman's Add Up

This sounds like a step in the right direction. (free registration/bugmenot required)

Posted by: P. Ryan Wilson on 13 Dec 2005

21 comments, Last at 14 Dec 2005, 8:46pm by doktarr

Comments

1
by Chris (not verified) :: Tue, 12/13/2005 - 2:00pm

The Yardage has and always will be terrible. It makes average games against bad defenses look great, it makes offenses against prevent at the end of blowouts look great, and it makes defenses facing good offenses look bad. It doesn't take ANYTHING useful into account.

2
by SJM (not verified) :: Tue, 12/13/2005 - 2:15pm

The article mentions that Aikman's ratings are not used for the official Fox power rankings, but fails to mention that those ratings are based on an even more complex, comprehensive and carefully calibrated system than AER.

FO can't get no love?

3
by Richard (not verified) :: Tue, 12/13/2005 - 2:18pm

It's just nice to see somebody inside the football "establishment" thinking "outside the box".

Cliché quota reached.

4
by Jerry F. (not verified) :: Tue, 12/13/2005 - 2:20pm

Actually, it doesn't mention power rankings. It complains that it doesn't use Aikman's ratings during the games on TV.

5
by Pat (not verified) :: Tue, 12/13/2005 - 2:26pm

Yah, in my mind, the lack of opponent adjustments (and fumble adjustments) really separates DVOA from the Aikmans. I think that would be even more relevant if there was a DVOA/Aikmans for college football - there, strength of schedule is usually the difference between 11-0 and 9-2.

Fumble adjustments are huge in my mind, too. When you've got something so random that can change a game completely, if you want to separate actual team skill from luck, you've got to compensate for fumbles.

6
by SJM (not verified) :: Tue, 12/13/2005 - 3:09pm

Re: #4

You are correct. I retract my previous comment.

7
by Kevin (not verified) :: Tue, 12/13/2005 - 3:35pm

Re: #5

What separates FO statistics and aikman's statistics is not just fumble adjustments and defensive adjustments. The real difference is the concept of taking statistics in the context of the game instead of as totals. yds/att is better than yds, but it's still a total of unlike and incomparable situations. I bet there's a greater differance between VOA and AER than DVOA and VOA.

8
by the fumble (not verified) :: Tue, 12/13/2005 - 4:33pm

Does anyone know if AER is additive amongst the different categories? Example: Is more weight given to 3rd dn conv than yds/rush? Are offense and defense simply added? I've run a similar metric over the last couple years and I've noticed that the offensive and defensive numbers are correlating differently to winning percentage based on the year. For instance, this year, the defensive numbers are correlating to winning % at a noticeably higher rate than last year (.87 vs .79), while offense has mostly stayed the same. I think that this could potentially be reflective of a 'more defensive environment' this year. I wonder if AER takes yearly fluctuations into account, or does it simply add up all categories as equals. Still 82% of games correct is very impressive, especially considering home-field fluctuations, anyone in a pick 'em poll?

9
by josh (not verified) :: Tue, 12/13/2005 - 5:17pm

Frankly, DVOA isn't so great. It doesn't seem to know whether it wants to be predictive or descriptive and the result is a muddle of the two.

10
by IzzionSona (not verified) :: Tue, 12/13/2005 - 6:32pm

thefumble: I read that to be "the team that had the better AER within the game wins 82% of the time, as opposed to the team that wins net yardage within the game only getting the 60-some-odd%"

If that's the case, they're not making any claims as to the a priori predictive power of the AER

11
by doktarr (not verified) :: Tue, 12/13/2005 - 7:23pm

Josh #9,

VOA is descriptive. Weighted DVOA is predictive. DVOA is the intermediary between them.

12
by tunesmith (not verified) :: Wed, 12/14/2005 - 12:28am

Where are the AER's anyway? Are they updated somewhere for each week? I tried googling and couldn't find them. That 87% stat is pretty impressive.

13
by Gatts (not verified) :: Wed, 12/14/2005 - 7:18am

tunesmith,

click my name

14
by admin :: Wed, 12/14/2005 - 9:50am

"Still 82% of games correct is very impressive, especially considering home-field fluctuations"

Just to make clear, the claim about the team with the better Aikman rating winning 82% of the time is NOT a claim that the team with the better Aikman rating BEFORE the game wins THAT GAME 82% of the time. It is a claim that the team with the better Aikman rating FOR THAT GAME wins THAT PARTICULAR GAME 82% of the time. This is not a claim that Aikman Ratings pick winners in future games.

(This also means that "home-field fluctuations" are not really an issue here.)

15
by the fumble (not verified) :: Wed, 12/14/2005 - 10:49am

OK, thanks guys, I saw 82% and got a little ahead of myself. But now I'm really not all that impressed with it. Sure it's a more useful metric than yardage, but what isn't? I am impressed with Aikman's effort to make a difference though.

16
by Pat (not verified) :: Wed, 12/14/2005 - 12:45pm

Just to make clear, the claim about the team with the better Aikman rating winning 82% of the time is NOT a claim that the team with the better Aikman rating BEFORE the game wins THAT GAME 82% of the time. It is a claim that the team with the better Aikman rating FOR THAT GAME wins THAT PARTICULAR GAME 82% of the time.

Man, Aikman's ratings suck, then! I'm gonna come up with my own rating - let's call it Pat Efficiency Rating (PER), where PER is equal to how many points the team scores.

Let's see how it does... see, I'm awesome! The team with the higher PER in a given game wins 100% of the time! Man! I rule!

:)

17
by Stuart (not verified) :: Wed, 12/14/2005 - 1:16pm

Well, sure, Pat, it's easy to data mine and come up with a system that is right based solely on past experience. Are you willing to put your PERs to the real test and use them to predict next week's games? Do you honestly think that the team that scores the most points will win every week?

I didn't think so.

18
by Pat (not verified) :: Wed, 12/14/2005 - 2:10pm

Well, sure, Pat, it’s easy to data mine

Um, it wasn't data mining... it's kindof the definition of the game that the team that scores more points wins.

:)

Are you willing to put your PERs to the real test and use them to predict next week’s games?

Actually, points scored is a pretty good predictor of future success. It's not perfect, but of all of the "simple" stats, it's pretty much the best.

19
by elhondo (not verified) :: Wed, 12/14/2005 - 3:25pm

Re # 16

Pat, your rating system is obviously biased and does not take into account the following

1) Vick's intangibles
2) Seattle's swagger
3) Injuries to Patriots

In fact, it's clear from your stats that you're an obvious Colt's homer. Why don't you go post your ridiculous formula on a Colt's fanboard and leave us alone.

20
by Kibbles (not verified) :: Wed, 12/14/2005 - 8:00pm

Any statistic that thumps its chest by saying it predicted who would win after the fact is just stupid. If you told me who would have the higher AER, I could predict who would win 82% of the time. Of course, good luck guessing who is going to have the higher AER.

Re #17: Well, sure, Pat, it’s easy to data mine and come up with a system that is right based solely on past experience. Are you willing to put your PERs to the real test and use them to predict next week’s games?

Why? The article isn't using AERs to predict future performance, so why should Pat use his PERs?

21
by doktarr (not verified) :: Wed, 12/14/2005 - 8:46pm

Well, to make a meaningful comparison, then: how often does the team with the higher VOA for that game win the game? More than 82% of the time?