Writers of Pro Football Prospectus 2008

11 Nov 2008

Smarter Stats: Divisional DVOA

This week's "Smarter Stats" column for the Washington Post looks at the best-to-worst divisions based on combined DVOA through Week 9. The top division isn't a surprise, but the movement in power from the AFC to the NFC since the last time we did this cumulative DVOA thing is pretty interesting.

Posted by: Doug Farrar on 11 Nov 2008

14 comments, Last at 19 Nov 2008, 5:22am by bengt


by Bowl Game Anomaly :: Tue, 11/11/2008 - 4:44pm

Some notes:

The deepest division is the NFC South (last place team with highest DVOA).

The NFC North is the most heavily swayed by an outlier (by eyeballing, I didn't do the math. The outlier is Detroit. The rest of the division is above average).

The AFC West is so bad, it would rank lower than the next-worse NFC West even if current division leader and #4 overall Arizona (21.2%) was dropped to below average. On that note, the West sucks (except Arizona).

While the NFC North technically averages closest to 0.0, the true "most average division" is the AFC East, since is has the lowest variance from 0.0 (again eyeballed, not calculated).

(Formerly "The McNabb Bowl Game Anomaly")

by BlueStarDude :: Tue, 11/11/2008 - 9:44pm

Sorry, the deepest division is the NFC East (unless Romo doesn't come back healthy). The Cowboys' offensive DVOA has been killed by BJ and the defense hasn't been helped by the offense's inability to sustain more than a few drives over the last three games. This is a glaring example of how a statement based on DVOA without bringing some context to it isn't really fruitful. The team that pasted the Packers and outlasted the Eagles hasn't turned to dust.

by Bowl Game Anomaly :: Tue, 11/11/2008 - 10:08pm

Sorry, let me clarify:

The deepest division based on past performance is the NFC South. If we ignore Dallas' performances with Brad Johnson and assume that Romo will return to form, the NFC East will probably (but not definitely) turn out to be the deepest.

(Formerly "The McNabb Bowl Game Anomaly")

by Doug Farrar :: Tue, 11/11/2008 - 11:38pm

That's the thing, though -- injuries happen, and we can't go back and say that Team X would be on a higher stat plane if Quarterback Y wasn't out for a month or a whole season. Teams are penalized in DVOALand for failing to bring base-quality depth to their roster just as they're penalized on the field and in the standings. The context is the wild swings in game-by-game passing DVOA between Romo and Johnson, but the Cowboys have to own their DVOA with Johnson because they're the ones who put him out there.

by BlueStarDude :: Thu, 11/13/2008 - 9:10am

Hi Doug. I'm not arguing that the Cowboys should be rated higher by DVOA. I'm not saying junk DVOA because it doesn't account for injuries. Injuries are a part of the game and so is signing a washed up QB to be your backup and keeping him on the roster for two seasons.

In recent years, though, I think this site (to generalize, not intending to bring up the old FO hive mind thing) has moved more and more away from Aaron's initial caveat about applying DVOA too literally.

Romo's not out for the season. He's coming back this week and it's reasonable to think he'll play at a level to make this team competitive again -- if he returns to form, they'll be more than merely competitive. Bowl Game Anomaly's not wrong in his statement. And I probably read more into it than what he intended it to say, a bit unfairly. But I disagree with the idea (implied at least) that there was anything relevant in strictly looking at the DVOA in this instance.

by mm (not verified) :: Tue, 11/11/2008 - 11:12pm

I've been meaning to ask if you guys could list divisional dvoa's on a weekly basis with the dvoa report as well as adding a section to the 'playoff odds' that gives each divisions chances of sending 0, 1, or 2 wild card teams to the playoffs (obviously, with only 2 of those numbers, we could infer the third).

You could also add things like variance, and other statistical tidbits, as well as comparing the 2 conferences.

There seems to be more interest in divisional rankings the last few years, and this sort of stuff would be interesting. It might be a little bit too much work to do before the offseason. However, once it's coded, the computer could spit it out every week without any more effort.

by Doug Farrar :: Tue, 11/11/2008 - 11:41pm

Not sure about coding right now, but if there's interest, I could run the numbers every week and put up a quick Extra Point with the updates.

by Red Hedgehog (not verified) :: Wed, 11/12/2008 - 2:54pm

Ooh, please do!

by mm (not verified) :: Wed, 11/12/2008 - 10:17pm

Well, I don't know if we need to necessarily track the Division DVOA week by week if you're calculating it by hand, though it would be nice to review it once 3-4 weeks from now, and then again after the season ends.

A wildcard report by division (0,1, or 2) would definitely be interesting to track week by week.

by bengt (not verified) :: Wed, 11/19/2008 - 5:22am

I'd say everybody can already calculate the probabilities for the number of teams from each division to reach the playoffs by themselves, using the numbers from the playoff odds. P_1team = P_divwint1*(1-P_wct2)*(1-P_wct3)*(1-P_wct4)+P_divwint2*(1-P_wct1)*(1-P_wct3)...

Of course, the wild card probabilities add up to 200% for each conference because there are two spots, and I'd have to think about whether that is correctly accounted for, but generally it can be done that way. I'd rather have the FO guys do the math ;-), but the necessary information is already available.

by Staubach12 :: Wed, 11/12/2008 - 4:04pm

Dallas has discovered that the difference between positive 70% DVOA and negative -70% DVOA is the difference between Tony Romo and Brad Johnson.
Can we officially dispense with all the "Romo is overrated because of his supporting cast" crap now? I've actually seen comments to that effect on this site this season.

by Bowl Game Anomaly :: Wed, 11/12/2008 - 5:42pm

I think Romo is overrated because of his good receivers, but underrated because of his bad line. So he's rated just about right overall.

(Formerly "The McNabb Bowl Game Anomaly")

by Staubach12 :: Wed, 11/12/2008 - 6:36pm

I think that's a fair evaluation.

by Wanker79 :: Mon, 11/17/2008 - 4:55pm

Last year (or maybe the previous year) Pat and I had a pretty good discussion regarding Divisional rankings (I don't remember if it was related to your article or if it was just brought up in a random DVOA thread). I think the consensus we came out with was that if your goal is to rank the divisions from best to worst, you should probably weight the teams in the division as to reduce the impact of outliers. We felt that the overall quality of a division of three good teams and one really bad team should probably be looked at closer to the three good teams than the one bad one. Afterall, the aforementioned hypothetical division could potentially mean 3 losses for an average team, whereas a division of 4 average teams would be a toss-up for an average opponent.

I'm pretty sure this was last year, because I believe the cause for contention was the AFC East where NE's ridiculous DVOA was making what was otherwise a mediocre at best division look much stronger than it probably should have.

Pat, if I've misspoke please feel free to correct me.