Comments on Advanced NFL Stats: Team Efficiency Rankings - Week 6

Great question. It's due primarily to strength...

2009-10-17T11:13:59.586-04:00

Great question. It's due primarily to strength of opponent.

I regretfully don't understand why you rank Te...

2009-10-17T03:49:55.971-04:00

I regretfully don't understand why you rank Tennessee defense 4th, when they have allowed 7.1 yds per pass (among the worst) and 2.8 yds per carry (THE best, but small sample size). Their run defense is comparable to Arizona's run defense, in the fact that both have TERRIBLE pass defenses so running the ball is no more than a simple strategy to pull the safeties and cornerbacks in. Further explanation of the rankings would be much appreciated.

I looked everything over and ran some tests. I cou...

2009-10-15T18:16:28.403-04:00

I looked everything over and ran some tests. I could do a few more passes to reduce the correlation slightly, but it won't change the rankings very much. There will always be some correlation because the sample of opponents is not random. A team can never play itself, so a good team has one less good team as a possible opponent and vice versa.

Andy-Very good analysis. I'll have to check on...

2009-10-15T01:03:01.552-04:00

Andy-Very good analysis. I'll have to check on the routine that does the opponent adjustment and make sure it's working properly. I've built a new tool this year that does all this stuff automatically so there may be some glitches.

I'm a little surprised that the correlation be...

2009-10-14T15:58:04.798-04:00

I'm a little surprised that the correlation between rank and previous opponent quality is so high. The correlation is p=.36 and is significant. The correlation between own GWP and opponent GWP is similarly negative and significant. Higher ranked teams are playing worse teams than lower ranked teams, at least according to your metric.

Perhaps I'm misunderstanding the methodology but isn't your construction attempting to control for opponent quality in measuring efficiency? Shouldn't the expected correlation between your efficiency measure and opponent quality thus be 0, at least in expectation? If so, what should we make of this correlation - a problem with the methodology or just noise?

"Two factors -- close games (the Titans could...

2009-10-14T15:01:52.641-04:00

"Two factors -- close games (the Titans could easily be 3-2)"

Pretty misleading thing to say. The Titans have a -55 point differential. I don't think anybody disagrees that they are running bad, but they'd need an insane amount of good luck to be 3-2, given their last two losses have been utter blowouts. Every Efficiency Ranking formula like this will have certain data points that just don't make sense. This is one of them.

Two factors -- close games (the Titans could easil...

2009-10-14T12:55:43.555-04:00

Two factors -- close games (the Titans could easily be 3-2) and strength of schedule.

Interesting to see the Titans (0-5) ahead of the V...

2009-10-14T11:33:23.116-04:00

Interesting to see the Titans (0-5) ahead of the Vikings (5-0).

this is, but do u think you could do some more wit...

2009-10-14T10:17:33.848-04:00

this is, but do u think you could do some more with fantasy football?

Hi Brian. Opp GWP is that future schedule or past?...

2009-10-14T05:29:54.596-04:00

Hi Brian. Opp GWP is that future schedule or past?

Juri