Writers of Pro Football Prospectus 2008

Most Recent FO Features

HarvinPer09.jpg

» Impact of the NFL's Kickoff Rule Change

After three NFL seasons of kicking off from the 35-yard line, what has been the impact on touchbacks, returns, field position, scoring and injuries? Also, is this rule responsible for a record number of big comebacks?

17 Oct 2006

Week 7 DVOA Ratings

by Aaron Schatz

Here's a look at this week's DVOA and DAVE ratings, and here's a link to the commentary on FOXSports.com.

Even more than in 2005, the NFL is dramatically split between good teams and bad teams, with lots of moderately good teams bunched very close together in our numbers. Note that the difference in DVOA between Atlanta (20) and Buffalo (21) is actually bigger than the difference between Atlanta (20) and St. Louis (9).

I meant to write to write a bunch of extra interesting tidbits here related to the Chicago-Arizona game, but I still have to finish an article for the New York Sun, so folks will have to wait until later in the week to find out:

  • 1) Did Rex Grossman have the lowest DPAR for any winning quarterback since our numbers begin in 1997?
  • 2) Was this the biggest spread ever between the single-game DVOA ratings of two teams in a game where the team with the lower rating actually won?

I'll hit these topics on the FO FOX blog later this week. Make sure you check out the FOX blog post on Edgerrin James, particularly if you want to find out the three running backs since 2000 who actually managed worse PAR ratings in a single game.

By the way, while DVOA is no longer the "official" FOXSports.com power rankings, we have created a keyword at FOX that will allow you to always access the latest DVOA commentary. Simply use keyword "DVOA" at FOXSports.com.

A quick note: I tweaked DAVE a little bit this week to count the preseason projection a little less for the six teams with a bye in the upcoming week. To be honest, I made a mistake in creating a set of coefficients based on weeks rather than games, and when I upgrade DAVE next summer, I'll fix that. But we're at the point where the use of the projection drops significantly each week, which would lead to a weird situation where St. Louis and New Orleans -- two teams where the projection and Weeks 1-6 differ substantially -- both ranked low again this week, then suddenly jumped up next week, even though neither team actually played a game. So for the six teams on bye this week, I counted the projection 20%, much closer to the 15% that we'll use for every team after Week 7. (Next week, I'll do something similar for the teams on bye in Week 8, and then it won't matter anymore because DAVE will go away for the year.)

The stats pages are now fully updated, including offense, defense, special teams, offensive line, and defensive line.

* * * * *

These are the Football Outsiders team efficiency ratings through Week 6 of 2006, measured by our proprietary Defense-adjusted Value Over Average (DVOA) system that breaks down every single play and compares a team's performance to the league average based on situation in order to determine value over average. (Explained further here.)

OFFENSE and DEFENSE DVOA are adjusted based on strength of opponent as well as to consider all fumbles, kept or lost, as equal value. Opponent adjustments are currently set at 60% and will increase each week until they are full strength after Week 10. SPECIAL TEAMS DVOA is adjusted for type of stadium (warm, cold, dome, Denver/Mexico City) and week of season.

DAVE is the new early-season formula that combines early-season performance with our preseason projection to get a more accurate picture of how well teams will play over the course of the entire season. This is the rating used to rank teams at FOXSports.com. After Week 6, the preseason projection counts for 32.5% of the rating for most teams, and 20% of the rating for teams with a bye in the coming week. (DAVE stands for "DVOA Adjusted for Variation Early.") The use of the preseason projection will end after Week 8.

To save people some time, please use the zlionsfan template for all complaints:

<team> is clearly ranked <too high/too low> because <reason unrelated to DVOA>. <subjective ranking system> is way better than this. <unrelated team-supporting or -denigrating comment, preferably with poor spelling and/or chat-acceptable spelling>


TEAM TOTAL
DVOA
LAST
WEEK
DAVE RANK NON-ADJ
TOT VOA
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
1 SD 47.3% 3 39.1% 1 60.7% 4-1 20.9% 2 -22.9% 5 3.4% 7
2 CHI 46.7% 1 39.0% 2 55.2% 6-0 5.4% 9 -28.6% 2 12.7% 1
3 PHI 39.9% 2 31.1% 3 34.2% 4-2 26.1% 1 -14.6% 9 -0.8% 18
4 JAC 31.7% 4 23.0% 4 33.9% 3-2 4.8% 10 -27.4% 3 -0.6% 17
5 NYG 26.5% 7 18.8% 5 23.1% 3-2 20.2% 3 -5.6% 13 0.7% 13
6 DAL 21.1% 9 13.7% 9 23.0% 3-2 4.0% 11 -19.5% 6 -2.4% 22
7 BAL 18.6% 6 15.0% 7 30.8% 4-2 -17.6% 27 -32.8% 1 3.4% 9
8 PIT 18.1% 20 17.5% 6 6.8% 2-3 0.9% 13 -23.4% 4 -6.2% 31
9 STL 13.2% 8 5.8% 15 26.3% 4-2 13.9% 5 0.6% 18 -0.2% 14
10 NO 11.0% 11 4.0% 17 16.2% 5-1 7.2% 8 1.4% 19 5.2% 4
11 NE 10.1% 10 9.3% 13 20.7% 4-1 12.0% 7 5.5% 22 3.7% 5
12 DEN 9.0% 15 12.3% 11 10.3% 4-1 -5.6% 19 -14.9% 8 -0.3% 15
13 IND 7.3% 16 14.1% 8 13.2% 5-0 19.5% 4 11.1% 28 -1.1% 20
14 CAR 5.3% 18 7.4% 14 4.5% 4-2 2.7% 12 -5.3% 14 -2.7% 24
15 CIN 4.4% 13 9.4% 12 8.6% 3-2 -3.1% 15 -4.1% 16 3.4% 8
16 SEA 4.3% 19 13.5% 10 -1.8% 4-1 -1.4% 14 -5.0% 15 0.7% 12
TEAM TOTAL
DVOA
LAST
WEEK
DAVE RANK NON-ADJ
TOT VOA
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
17 KC 4.3% 5 5.5% 16 0.2% 2-3 -9.1% 21 -5.8% 12 7.5% 3
18 WAS 3.8% 14 3.8% 18 -0.1% 2-4 12.6% 6 9.8% 26 1.1% 11
19 MIN 1.4% 17 -5.1% 20 -0.7% 3-2 -9.4% 22 -13.9% 10 -3.2% 26
20 ATL -0.2% 12 2.9% 19 7.7% 3-2 -13.7% 26 -19.0% 7 -5.4% 28
21 BUF -13.8% 21 -16.7% 24 -13.9% 2-4 -12.8% 25 4.6% 21 3.5% 6
22 DET -18.3% 24 -14.9% 23 -19.7% 1-5 -5.2% 18 10.4% 27 -2.6% 23
23 TB -18.3% 26 -14.3% 22 -27.4% 1-4 -11.1% 24 2.6% 20 -4.6% 27
24 MIA -20.0% 27 -13.4% 21 -10.3% 1-5 -20.0% 29 -3.0% 17 -3.0% 25
25 ARI -20.7% 29 -20.7% 27 -16.7% 1-5 -22.4% 30 -12.0% 11 -10.3% 32
26 GB -21.0% 22 -18.3% 26 -28.4% 1-4 -7.0% 20 8.3% 23 -5.8% 29
27 NYJ -22.0% 28 -23.3% 28 -18.0% 3-3 -5.2% 17 15.9% 29 -0.9% 19
28 CLE -22.5% 23 -17.8% 25 -25.3% 1-4 -23.0% 31 9.2% 25 9.7% 2
29 SF -26.5% 25 -26.1% 29 -31.3% 2-4 -3.5% 16 22.7% 31 -0.3% 16
30 TEN -39.8% 32 -35.3% 31 -35.8% 1-5 -19.4% 28 18.9% 30 -1.5% 21
31 HOU -42.1% 30 -35.0% 30 -42.5% 1-4 -9.4% 23 34.7% 32 2.0% 10
32 OAK -51.7% 31 -38.0% 32 -59.0% 0-5 -37.0% 32 8.8% 24 -5.8% 30

  • ESTIMATED WINS uses a statistic known as "Forest Index" that emphasizes consistency as well as DVOA in the most important specific situations: red zone defense, first quarter offense, and performance in the second half when the score is close.  It then projects a number of wins adjusted to a league-average schedule and a league-average rate of recovering fumbles.  Teams that have had their bye week are projected as if they had played one game per week.
  • PAST SCHEDULE lists average DVOA of opponents played this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • FUTURE SCHEDULE lists average DVOA of opponents still left to play this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • VARIANCE measures the statistical variance of the team's weekly DVOA performance.  Teams are ranked from least consistent (#1, highest variance) to most consistent (#32, smallest variance).


TEAM TOTAL
DVOA
W-L ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VARIANCE RANK
1 SD 47.3% 4-1 5.0 2 -13.5% 30 -6.0% 25 6.9% 21
2 CHI 46.7% 6-0 4.9 3 -13.6% 31 -6.8% 26 29.3% 5
3 PHI 39.9% 4-2 5.3 1 -6.2% 26 3.7% 10 2.8% 32
4 JAC 31.7% 3-2 4.0 5 8.4% 6 -11.0% 31 34.1% 4
5 NYG 26.5% 3-2 4.2 4 11.0% 3 7.3% 4 8.1% 17
6 DAL 21.1% 3-2 3.5 13 -1.1% 18 6.3% 7 17.1% 11
7 BAL 18.6% 4-2 3.7 7 -6.2% 25 -1.5% 19 15.0% 13
8 PIT 18.1% 2-3 3.6 10 13.5% 2 -4.4% 23 15.1% 12
9 STL 13.2% 4-2 3.7 8 -12.2% 29 1.4% 16 11.2% 15
10 NO 11.0% 5-1 3.5 12 -2.8% 22 5.3% 9 8.8% 16
11 NE 10.1% 4-1 3.7 9 -7.1% 27 -9.0% 30 5.6% 25
12 DEN 9.0% 4-1 3.5 11 -0.9% 17 1.2% 17 11.4% 14
13 IND 7.3% 5-0 3.8 6 -7.6% 28 0.4% 18 5.9% 24
14 CAR 5.3% 4-2 3.2 17 -1.7% 20 11.9% 3 6.9% 20
15 CIN 4.4% 3-2 3.0 18 -1.4% 19 6.1% 8 7.9% 18
16 SEA 4.3% 4-1 3.4 15 7.9% 7 -9.0% 29 7.5% 19
TEAM TOTAL
DVOA
W-L ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VARIANCE RANK
17 KC 4.3% 2-3 2.9 19 -3.1% 23 2.3% 14 46.8% 1
18 WAS 3.8% 2-4 3.3 16 -0.3% 15 13.2% 2 17.3% 10
19 MIN 1.4% 3-2 3.4 14 4.7% 12 -6.8% 27 4.1% 29
20 ATL -0.2% 3-2 2.7 20 0.6% 14 6.3% 6 45.2% 2
21 BUF -13.8% 2-4 2.4 23 -0.3% 16 -3.0% 22 22.2% 6
22 DET -18.3% 1-5 2.0 26 6.1% 10 -2.8% 21 4.6% 28
23 TB -18.3% 1-4 2.4 22 3.1% 13 15.4% 1 6.7% 22
24 MIA -20.0% 1-5 1.8 27 -17.9% 32 2.4% 13 5.3% 26
25 ARI -20.7% 1-5 2.5 21 8.4% 5 -1.9% 20 22.1% 8
26 GB -21.0% 1-4 2.2 25 15.4% 1 -5.7% 24 5.0% 27
27 NYJ -22.0% 3-3 1.8 28 -4.1% 24 -13.1% 32 22.2% 7
28 CLE -22.5% 1-4 1.5 29 -2.5% 21 3.4% 11 3.6% 30
29 SF -26.5% 2-4 2.2 24 6.4% 8 2.7% 12 40.5% 3
30 TEN -39.8% 1-5 1.3 30 6.3% 9 6.8% 5 21.0% 9
31 HOU -42.1% 1-4 1.1 31 10.4% 4 -7.5% 28 3.5% 31
32 OAK -51.7% 0-5 0.5 32 5.2% 11 1.8% 15 5.9% 23

Posted by: Aaron Schatz on 17 Oct 2006

223 comments, Last at 24 Oct 2006, 10:20am by Wanker79

Comments

1
by zip (not verified) :: Tue, 10/17/2006 - 6:34pm

Detriot is clearly ranked too low because they beat the Bills last week. Beatpaths are way better than this. Teh Bills make me cry :'(

2
by Pat (not verified) :: Tue, 10/17/2006 - 6:41pm

The Bills just haven't had a chance to form a beatloop around that win yet. They will. Or, they would, if the season was infinitely long.

I really like that DVOA didn't kill the Eagles for that loss - I didn't expect it to, of course. New Orleans is playing like a top-10 team. Philly lost to a good team, at home, by a last-second field goal. That's not exactly unexpected, in my mind. I'm kindof amazed Philly's defense didn't take more of a dive, though.

3
by Independent George (not verified) :: Tue, 10/17/2006 - 6:45pm

NYG past sched: #3
NYG future sched: #4

Ugh. At least we're not Washington.

4
by Lou (not verified) :: Tue, 10/17/2006 - 6:46pm

That Spawn guy on the blog has some serious reading comprehension issues

5
by Oswlek (not verified) :: Tue, 10/17/2006 - 6:48pm

Chicago's DVOA drops nearly 20 points after a win.

How often does that happen?

6
by aurumaeus (not verified) :: Tue, 10/17/2006 - 6:48pm

NYG is nothing- expect the Packers to go 12-4, with #1 past sched, #24 future sched.

7
by Matthew Furtek (not verified) :: Tue, 10/17/2006 - 6:49pm

Washington's offense is clearly ranked too high because Mark Brunell Sucks!. Flipping a coin is way better than this. Brunell is going to force Gibbs to keep Campbell on the bench another year.

It wouldn't be shocking if they beat the Colts.... although it would be shocking to the upperground media.

8
by Ray (not verified) :: Tue, 10/17/2006 - 6:49pm

Looks like the NFC East has decided to show up (except for Washington). DVOA ranks 3, 5, and 6 are all from the NFCE.

9
by Chris S (not verified) :: Tue, 10/17/2006 - 6:58pm

San Diego is 4-1, yet their estimated wins is 5.0. What exactly does this mean? I saw the game against baltimore where they lost, and clearly SD dominated for most all of the game except for the winning drive by McNair where their defense fell apart for 5 minutes. Is this saying that statistically they won that game?

10
by Purds (not verified) :: Tue, 10/17/2006 - 7:00pm

Ray:

I noticed that too, and I was wondering which is the worst:

AFC South: 4, 13, 30, 31 (total 78)
NFC West: 9, 16, 25, 29 (total 79)
AFC East: 11, 21, 24, 27 (total 83)

11
by BB (not verified) :: Tue, 10/17/2006 - 7:01pm

Names heard on Chicago sports radio (670) for the Evil Rex that showed up yesterday:

the Rextard
Dyslexy Rexy

12
by David (not verified) :: Tue, 10/17/2006 - 7:04pm

#9: Estimated wins are pro-rated over the length of the season thus far, disregarding bye weeks. So the Chargers are estimated at 5-1, not 5-0.

13
by Manteo (not verified) :: Tue, 10/17/2006 - 7:04pm

At least the Hawks' schedule strength goes from 7th ranked (past) to 29th (future). Now if Shaun Alexander can just pray a little harder...

#7 - So is FO now considered underground, too? Do we have secret meetings with the Deadspin folks to plot insurrection and/or rock and roll? Cool.

14
by Scott C. (not verified) :: Tue, 10/17/2006 - 7:06pm

2) Was this the biggest spread ever between the single-game DVOA ratings of two teams in a game where the team with the lower rating actually won?

Wouldn't VOA, not DVOA be more appropriate for such a comparrison?

Lets assume the hypothetical "mirror image" game.
In this game, both teams play identically well. Every play is equal to a counter on the other side. Team A returns the first kickoff from the 5 yard line to the 25. First play is a 3 yard run, second a 8 yard pass.... etc. Team B's first drive is identical, kickoff and everything. One team wins because either it had the ball last or maybe a missed field goal separates them.

In this case, the VOA of the two teams would be nearly identical. (though could be below or above 0% based on luck factors).

But now lets assume Team A has one of the top defenses and offenses outside of the game in question, its one of the top rated DVOA teams in the league. Team B is mediocre. DVOA, for the mirror-image game would have Team B with a positive DVOA score and Team A with a negative one -- even if the performance in game was IDENTICAL. Team B didn't perform better, than A, it just performed better compared to expectations. Both teams were identical in the particular game, so VOA is what should be used.

It makes more sense to me to use VOA, not DVOA, to look at wacky outcomes, since VOA is what correlates with winning in any single game. DVOA just adds the perspective of opponent adjustment and expectations. It makes Chicago look worse and Arizona look better, than they actually played in that game relative to each other.

15
by B (not verified) :: Tue, 10/17/2006 - 7:07pm

9: It means SD would have gone 5-0 against an average schedule.

16
by Pat (not verified) :: Tue, 10/17/2006 - 7:14pm

Wouldn’t VOA, not DVOA be more appropriate for such a comparrison?

Depends what you're trying to compare: the team who played the worst, and won, in absolute terms, or the team who played the worst, and won, in relative terms.

Arizona played fantastic on offense - not just because they played better than average, but because they were playing the Bears. Likewise for Chicago.

9: It means SD would have gone 5-0 against an average schedule.

Bye teams are prorated to have a bye. So 5.0 estimated wins is a 5/6 winning percentage.

17
by DenverFan (not verified) :: Tue, 10/17/2006 - 7:17pm

Denver is clearly ranked too lowbecause I like them a lot. Ranking the Broncos #1 no matter what is way better than this. Faiders frikkin suk, and that makes me happy. GO BRONCOS!!!!!!

18
by Independent George (not verified) :: Tue, 10/17/2006 - 7:22pm

Hmmm... The Raiders are almost 10 points worse than Houston, but the Texans have been so much more consistently awful... Oh, how does one choose?

19
by KNSD (not verified) :: Tue, 10/17/2006 - 7:29pm

San Diego is clearly ranked too low because they have been playing above themselves lately and deserved to be ranked higher than they actually are. My ranking system is way better than this. Go Chrgers!

20
by Luz (not verified) :: Tue, 10/17/2006 - 7:30pm

i think dvoa is broken. it has the steelers at #31 in special teams. that can't be correct. #35 would be more appropriate.

21
by Rollo (not verified) :: Tue, 10/17/2006 - 7:35pm

Jacksonville:

Past schedule: 6th
Future schedule: 31st

Sounds alot like last year.

Indianapolis:

Past schedule: 28th
Future schedule: 18th

It will be interesting to see if McFarland improves Indy's run defense enough and/or in time for the real schedule.

22
by Richard (not verified) :: Tue, 10/17/2006 - 7:38pm

As a Charger fan, seeing them ranked #1 worries me. Damn Chargers...

23
by Pacifist Viking (not verified) :: Tue, 10/17/2006 - 7:39pm

Oakland is clearly ranked too high because Ohio State could beat them. Using Tarot cards is way better than this. USC might beat Oakland too.

24
by Orothar (not verified) :: Tue, 10/17/2006 - 7:44pm

Is anyone else as suprised as I am to see the Rams as high as they are? They've played ok, but have their record mostly due to recovered fumbles (once again, against Sea, they recovered a late fumble to make it a close thing).

On the other hand, they are CAUSING all those fumbles - but eventually the luck will even out right?

25
by CaffeineMan (not verified) :: Tue, 10/17/2006 - 7:46pm

In the suckitude race between the AFCE, NFCW, AFCS, the AFC East wins! Average DVOA's are:

AFCE: -11.4%
AFCS: -10.7%
NFCW: - 7.4%

26
by admin :: Tue, 10/17/2006 - 8:00pm

Re: 14/16, it also depends what I can compile in an hour because I've already got most of the data and what I can compile in four hours because I don't. Such is life.

Commentary now up:
http://msn.foxsports.com/nfl/story/6072082

27
by turbohapy (not verified) :: Tue, 10/17/2006 - 8:12pm

Re: 21

I saw that. Why did TB give up McFarland? Can some of the TB fans on here offer some insight?

28
by michael (not verified) :: Tue, 10/17/2006 - 8:13pm

*snicker* Not even Simmons could find an excuse to reference Ballistic: Ecks vs. Sever. Well played, sir; well played.

29
by Kevo (not verified) :: Tue, 10/17/2006 - 8:14pm

The Raiders and Titans having byes this week might kill me in my survival pool.

30
by Xian (not verified) :: Tue, 10/17/2006 - 8:14pm

Nice. Now I don't have to feel like a fool for believing in the Chargers. Heh.

Also, on the Fox commentary, there are two instances of the Cardinals logo. The 49ers have it as well for some reason. Was the editor giggling over the commentary a little too hard?

31
by michael (not verified) :: Tue, 10/17/2006 - 8:17pm

also, re: "most groundskeeping talk on local sports radio", they mentioned on PTI today that NESN said the Pats will switch to FieldTurf *this* year, by the Chicago game. (Link on my name).

Has *any* team ever switched field surfaces mid-season (either from natural to artificial, artificial to natural, or Astro- to Field-)?

32
by Larry (not verified) :: Tue, 10/17/2006 - 8:18pm

I'll weigh in that VOA is the right thing to discuss the "outplayed really badly, but still won" question. DVOA will better reflect the extreme interest in discussing last night's game, though, as it brings out Arizona's defying of expectations in outplaying Chicago. I'm guessing that looking at VOA, it won't seem nearly as remarkable. Still, I think VOA spread is the right number to look at here.

Of course, if Aaron does differently, I can always ask for my money back...

33
by Ben B. (not verified) :: Tue, 10/17/2006 - 8:20pm

The Chargers are clearly ranked too high because dear god please don't let what happened to the Bears happen to us. Cowering in terror is way better than this. *shudders and slowly backs away*

34
by wr (not verified) :: Tue, 10/17/2006 - 8:24pm

Aaron, your NEXT line for the Cardinals at Fox is just priceless!!!

35
by BillWallace (not verified) :: Tue, 10/17/2006 - 8:24pm

Sunny D vs. the Purple Stuff!!!!!
Yet another can't stop from laughing at the office joke.
Man that takes me back.

36
by Andrew (not verified) :: Tue, 10/17/2006 - 8:25pm

Am I the only one here that really dislikes DAVE? I don't like seeing teams like New Orleans getting no credit for way outperforming expectations just because the expectations were so low and teams like Seattle getting ranked so high just because they went to the Super Bowl last year. New Orleans is playing well and the DVOA shows it. Seattle isn't and the DVOA shows it. Why bring in preseason projections?

37
by Scott C. (not verified) :: Tue, 10/17/2006 - 8:27pm

Depends what you’re trying to compare: the team who played the worst, and won, in absolute terms, or the team who played the worst, and won, in relative terms.

If the topic of discussion is the most improbable victory, its VOA then.

If the topic is most unexpected result based on pre-game assumptions, DVOA makes more sense. However this is not only restricted to games where the winner had the lowest DVOA, but simply, biggest DVOA game differences. If both teams play to expectations, the favorite wins and DVOA for the game almost = DVOA for the season, for both teams.

Since the context of this seems to me to be the most improbable victory according to statistics, we want the absolute, game only analysis that VOA gives.

38
by Richard (not verified) :: Tue, 10/17/2006 - 8:37pm

36: With regard to the "why," I believe they determined that they're more accurate looking forward. The issue with the Saints would seem to be that they crapped in the projection system.

39
by Travis (not verified) :: Tue, 10/17/2006 - 8:39pm

Has *any* team ever switched field surfaces mid-season (either from natural to artificial, artificial to natural, or Astro- to Field-)?

You can't switch from natural to artificial mid-season, by NFL rule.

40
by Jerry (not verified) :: Tue, 10/17/2006 - 8:41pm

Aaron: Why are the ratings so backloaded? I don't understand the reasoning for still having the projection numbers in the ratings. In fact, I don't even understand why games beyond the previous 4 are used. I really don't care what happened 4 or 5 games ago. In football time, that's forever given the nature of the game.

For example, The Ravens are currently ranked 7th yet the two teams that have beat them, Denver and the Panthers, are only ranked 12 and 14 respectively. I am pretty sure both wins were not flukes and would venture to guess, either is better than Baltimore yet by your numbers, they are worse. And if you look at the Panthers, not only did they beat the Ravens (ranked 6th at the time) but they also beat NO (ranked 5th at the time) and yet they are still ranked below both by a considerable amount. Do you really believe both NO and Baltimore are currently better then the Panthers? Do you really believe the Ravens are the 7th best team in the NFL?

I realize the ratings are a lagging indicator but it seems to me that you could get a better indicator of teams true value with a greater concentration on recent performances. Most people don’t care how a team performed 5 or 6 games ago; they just want to know who stands the best chance of winning next Sunday.

Just curious.

41
by Seattle Doug (not verified) :: Tue, 10/17/2006 - 8:44pm

The Burger King line is hilarious, but we've been calling Truffant Quizno's - because he's always freshly toasted.

42
by Travis (not verified) :: Tue, 10/17/2006 - 8:45pm

For some reason, with 45 seconds left and the lead, in a situation where a first down ends the game, the Jets decided to throw instead of run on third-and-2 from their own 33-yard line.

I'm sure it's just a typo, but there were 2 minutes and 45 seconds left before that 3rd down play.

43
by JonL (not verified) :: Tue, 10/17/2006 - 8:48pm

NEXT: Self-flagellation, then at OAK

I laughed really hard at this.

44
by BadgerT1000 (not verified) :: Tue, 10/17/2006 - 9:04pm

What am I missing in the evaluations? Because in the last 3 games the offensive line for GB hasn't been the issue.

vs. Detroit Favre had plenty of time to throw as Tony Moll of all people handled Shaun Rogers. The running game wasn't much but it was later discovered that Ahman Green played hurt.

vs. Philly Favre also had time to throw but stunk it up in the second half. Again the running game was ok when Morency held onto the ball

vs. Rams Favre had time (until the last play of the game--sigh) and Noah Herron had huge opportunities against the Rams.

The numbers keep declaring that the Packers are worse then last year even though they have played a tougher schedule then last year's start of the season. And GB was in the game til the final play against both the Rams and Saints. Losing to Philly and Chicago isn't exactly a disgraceful event.

So again, educate me. Because I don't understand the offensive line comment and I don't understand folks telling me the team is regressing as the season moves along.

Thanks.

45
by David (not verified) :: Tue, 10/17/2006 - 9:13pm

Sucks to be the Jets. 3-3, and yet ranked behind no less than five one-win teams (DAVE says 6 , DVOA says 5)? Ouch.

46
by MitchW (not verified) :: Tue, 10/17/2006 - 9:13pm

Okay, got a poll question - Who is prettier -

(A) the newly slimmed down Dan Marino

OR

(B)the (not-quite-possessed)Catholic chick?

47
by Matthew Furtek (not verified) :: Tue, 10/17/2006 - 9:15pm

Seriously about the Redskins. Why is their offensive Rank 6? Brunell is even ranked in the top 15 of DPAR?!

Tanier's comment is right on the money. Reports from fans at the game have indicated Lloyd and ARE are getting open but Brunell is not able to get them the ball, especially over the middle.

I think a lot of us fans might be unfairly critical to Brunell, and overlooking the defense. I would expect both offense and defense ranked in the bottom half of the league. It could just be a bit of improvement in red zone performance for the offense.

I expect another shootout with the Colts this week.... Brunell vs. Manning... what a hoot.

48
by Rodafowa (not verified) :: Tue, 10/17/2006 - 9:20pm

Thoroughly entertaining and informative, as ever. And all power to you for the Hellblazer reference.

Alan. Moore. Is. God.

49
by Jesse (not verified) :: Tue, 10/17/2006 - 9:33pm

Denver is clearly ranked too low because they have, by far, the best defense in the NFL. Ranking teams by the elevation of their home stadium is way better than this. Raidas sukkk! 0-16 worst teem evaaa!!

50
by BroncoMan (not verified) :: Tue, 10/17/2006 - 9:35pm

Oakland is clearly ranked too high because being ranked last in the NFL cannot convey how horrible they are. The BCS is way better than this. Raaaaaideeeerrrsss SUUCKKKKKKK!!

51
by D (not verified) :: Tue, 10/17/2006 - 9:49pm

As a Bear's fan that has been saying for weeks that Grossman was a fundamently unsound QB who was putting up great numbers because of a weak schedule and incredible luck I just want to say that I really hope this scares him into learning that he can't go for broke on every play and just assume that any bad pass will end in a simple incompletion. That or Lovie Smith simply deciding its time to start Griese.

52
by D (not verified) :: Tue, 10/17/2006 - 9:52pm

P.S.
I wouldn't be too woried Ben B., no one has started comparing your team to the '72 Dolphins, '85 Bears, or '99 Rams.

53
by Kevin (not verified) :: Tue, 10/17/2006 - 9:58pm

Seeing the Giants ranked number five overall is especially gratifying when you see the past schedule rankings of the top three teams. It's also nice to see Tiki Barber ranked as the top overall back. Gotta love the fact his backup Brandon Jacobs (no Troy Aikman, he doesn't stink) has the top success rate in the league and is number seven overall.

54
by admin :: Tue, 10/17/2006 - 10:08pm

Hi everybody. All stats pages are now updated, including offensive and defensive lines.

Bill Barnwell gets credit for the Ecks vs. Sever mention in the Dallas comment.

I'll try to answer some of these questions in the blog later this week, but often the answer to your question is going to be one of the following:

A) I know, it seems like the system should work that way, but I've spent a ton of hours trying different things over and over to get the most accurate system possible, and it ended up like this.

B) It's very difficult to write 32 comments about 32 teams every week, and usually one or two of them ends up not quite being as interesting or funny as the others. That's usually the last one I wrote. If that comment happens to be about your team, it's not because I have anything against them.

55
by Becephalus (not verified) :: Tue, 10/17/2006 - 10:11pm

The rating are backloaded because they are more correlative that way. A good evidence for this might be PIT or SEA winning games this weak impressively, both of which normal DVOA was not a huge fan of, but DAVE gave a boost to.

If you cannot understand sticking with a stastical system when it works better despite it conflicting with conventional wisdom, this site just isn't for you.

56
by Fnor (not verified) :: Tue, 10/17/2006 - 10:25pm

#46: (c) John Elway

He's hauntingly beautiful, after all.

57
by Pat (not verified) :: Tue, 10/17/2006 - 10:38pm

Why is their offensive Rank 6? Brunell is even ranked in the top 15 of DPAR?!

San-freaking-tana Moss. I have to say, as I've said elsewhere, I think Santana Moss is getting heavily screwed by all the attention lavished on Steve Smith. If ever there was a clone of the Carolina Panthers offense, it's the Redskins.

In fact, I don’t even understand why games beyond the previous 4 are used. I really don’t care what happened 4 or 5 games ago. In football time, that’s forever given the nature of the game.

Actually, you're wrong. Studies have actually shown that games don't become useless for predicting future behavior until about 10-12 weeks later, if memory serves. It's definitely not 4 or 5 games ago.

It's true that in "football media world time" it's forever, but not in reality.

58
by mm (not verified) :: Tue, 10/17/2006 - 10:40pm

31-- I believe the Saints switched from astroturf to one of the new nicer surfaces with a few games remaining in the season a few years back. Of course, that's a switch from one type of artificial surface to another.

59
by Duff Soviet Union (not verified) :: Tue, 10/17/2006 - 10:44pm

"If you want an indicator of how good a team is in the long run, try Brian Urlacher tackling the running back for no gain roughly 40 bazilllion times. THAT's an indicator." Hey, remember when Brian Urlacher was the "most overrated player in the NFL"? Good lord, that was stupid.

60
by morganja (not verified) :: Tue, 10/17/2006 - 10:46pm

I don't think that his point was that this is a statistical system that works better but conflicts with conventional wisdom. His point is that it is not a good indicator because it counts games at the beginning of the season equally to recent games. Agee with it or not, he has a valid point and your arrogant condescending attitude is a very weak counterargument. Statistics aren't Black Magic. They have to be constantly reassessed and used by those with an honest pursuit of the truth. The biggest weakness of statistics is when people fall in love with their model and forget what they are trying to do in the first place. This model may or may not be the best model out there, but it is still weak, and questions like his need an honest answer if the model is going to be improved.

61
by CaffeineMan (not verified) :: Tue, 10/17/2006 - 10:50pm

I like DAVE a lot, because together with regular DVOA it gives you a better picture of what's actually going on with a team. Of course, I really don't care about whether the teams are listed in order of DVOA or DAVE, so maybe that's why DAVE doesn't bug me. More data (especially data that correlates well with full-season performance) is good. SEA and PIT are both good examples of the value of DAVE, as mentioned before. And as Aaron has already mentioned, NO is doomed this year to be a lousy example of how DAVE/DVOA work because their circumstances last year were so weird.

62
by DGL (not verified) :: Tue, 10/17/2006 - 10:51pm

I'm waiting for the week when Aaron writes all the Fox comments using the zlionsfan template.

63
by Pat (not verified) :: Tue, 10/17/2006 - 10:52pm

His point is that it is not a good indicator because it counts games at the beginning of the season equally to recent games. Agee with it or not, he has a valid point and your arrogant condescending attitude is a very weak counterargument.

I agree.

Here's a better counterargument: look up Weighted DVOA. There is a version of DVOA which counts older games less important than current games. It takes a lot more than 4 or 5 games for that to happen, though.

Maybe that should get added to the FAQ.

64
by Becephalus (not verified) :: Tue, 10/17/2006 - 10:57pm

moreganja, there is weighted DVOA, I assume that will be making an appearance later on in the year.

65
by Pat (not verified) :: Tue, 10/17/2006 - 10:57pm

Here's the link to the weights for weighted DVOA. You can see that weighted DVOA basically won't become useful for two more weeks. Conventional wisdom is just wrong - teams just don't really evolve that much over timescales less than about two months.

66
by young curmudgeon (not verified) :: Tue, 10/17/2006 - 11:24pm

Re 60: to completely misconstrue what morganja is saying, but to make what I think might be a valid point, I'd like to suggest that games at the beginning of the season count equally to recent games in one very important way, viz. determination of who makes the playoffs (i.e., the standings.) Games in which your quarterback, your top receiver, and your best safety are all playing injured (or like they're injured) don't necessarily predict how well the Steelers are going to play the rest of the year (or I sure hope not!), and I dearly hope that recent performance is a more accurate indicator, but alas, those losses still count in the standings.

67
by Will Allen (not verified) :: Tue, 10/17/2006 - 11:36pm

Unless, Pat, the evolution involves the healing of a player like Steve Smith. This may be an area, injury recovery, or injury occurence, where improvements to the model may be made. I think the model may be incorporating qb injuries already, which makes sense, since that is the postion where injury effects are most easily quantified. The injury component to the model has always been a little opaque to me, and perhaps it is meant to be; it is a propietary model, after all.

Anyways, I know the Vikings DVOA would drop somewhat if their victory over Carolina was weighted for the substantial effect of Smith being absent, but I know how fiendishly complex it would be to try to incorporate the effect of all injuries in a sound manner.

68
by Vince (not verified) :: Tue, 10/17/2006 - 11:51pm

And Atlanta's passing attack is STILL not quite the worst in the league!!! YES!!!

69
by Jersey (not verified) :: Tue, 10/17/2006 - 11:52pm

Ok guys the bolded ridiculous stuff about teams was funny like the first 50 times but the fact that it comes up 10 times a week is just annoying at this point. We all get the point of stupid comments... it doesn't need to be throw in my face every single time i read ANY thread on this site, even extra points threads. christ almighty....

70
by Adam, VA (not verified) :: Wed, 10/18/2006 - 12:09am

Re: #60

No complaints when you're posting using the name "Jersey"

71
by DGL (not verified) :: Wed, 10/18/2006 - 12:27am

In other words...

The zlionsfan template is clearly ranked too high because we all get the point of stupid comments. Comments on the Fox Blogs are way better than this. christ almighty...

Or to recurse another time,

Complaints about the zlionsfan template are clearly ranked too low because we all appreciate self-denigrating and indirectly self-referential humor. Sportspickle.com is way better than this. If you don't like it, don't read it.

72
by luz (not verified) :: Wed, 10/18/2006 - 12:27am

i just noticed that the steelers defense ranks highest, 3rd, on passes to running backs, while being 10th, 17th and 11th against #1 WRs, #2 WRs and TEs. as i recall from last year, it was noted that passes to the running backs were the steelers biggest defensive weakness and, i believe, a general criticism of the 3-4.

my question is, will the steelers defensive numbers flip by years end or did something more subtle happen with the defense/defensive players that i missed?

73
by AHBM (not verified) :: Wed, 10/18/2006 - 12:45am

Interestingly, if you use the central mean method (cf. Sagarin ratings) to compare the three worst divisions, you get a slightly different result if you compare numeric rankings and numeric ratings. AFC South edges out NFC west if you apply central mean to their ranking positions, but by DVOA central mean AFC South is much closer to AFC East.

Whichever way you go, though, AFC East still scrapes the bottom. (central mean DVOA -14.3)

74
by Becephalus (not verified) :: Wed, 10/18/2006 - 1:20am

And the best part about this is that despite listing 3 divisions as easier than NFCN you still constantly hear it refered to as the weakest division in football. Lazy writers.

Anyone notice that during the MLB playoffs the best pitcher in the major leagues has made about 20 apperances. Maybe he is the best because he can pitch every game? Oh thats right its because he is 12 different guys...

Superlatives must die!

75
by manning,eli (not verified) :: Wed, 10/18/2006 - 1:42am

did anyone else noticed how large john elways hauntingly beautiful head got in the last few years of his career-think coors light commercials in those comfy chairs in vail-due to all those "cortisone" shots to deal with the pain? he was still the best qb ever though.

76
by Pat (not verified) :: Wed, 10/18/2006 - 1:47am

The injury component to the model has always been a little opaque to me, and perhaps it is meant to be; it is a propietary model, after all.

There is no injury component to in-season DVOA. The injury components are in the season projections.

Losing a player for a week or two is just adding a bit of variance. There's nothing you can really do for that.

Besides, it says something about a team if they play significantly worse without one player.

77
by Countertorque (not verified) :: Wed, 10/18/2006 - 1:49am

The Raiders have been in almost every game they've played into the 4th quarter (meaning down by less than 2 TD's). They must be doing something right.

78
by Fnor (not verified) :: Wed, 10/18/2006 - 2:02am

It's good to see that Becephalus has declared war on a portion of our language.

79
by David (not verified) :: Wed, 10/18/2006 - 2:50am

The Raiders have been in almost every game they’ve played into the 4th quarter (meaning down by less than 2 TD’s). They must be doing something right.

I think most of what they're doing is inspiring the opposing coach to start running out the clock at the second-half kickoff. I don't agree with Madden much, but when he pointed out that the Raiders seemed to have no chance even though they were within 10 with five minutes to go and driving inside the Denver 30, he was saying exactly what I was thinking. Granted, the odds are that sooner or later lightning will strike and someone will regret that gameplan. But it's worked so far.

80
by Jake (not verified) :: Wed, 10/18/2006 - 2:55am

Someone from Jersey complaining? In an annoying manner? SHOCKING!

81
by manning,eli (not verified) :: Wed, 10/18/2006 - 3:00am

i think you meant SHOCKEY

82
by Bobman (not verified) :: Wed, 10/18/2006 - 3:03am

Aaron, first things first: your Bills commentary is misplaced--it references the team ranked right above them, Atlanta, but that's not the case. Left over from last week?

Second, this is a real head-scratcher, but I'll chalk it up to early-season-ness. As an Indy fan, I am happy to see the Broncos and Pats (two upcoming foes) ranked near the middle of the pack. But I am pretty sure they're better than that. I'm also sure Indy is better than #8, but that's more an article of blind faith at this point.

Regarding the Colts commentary, does that mean that Indy's OL is good, despite what everyone here says? (By everyone, I refer to the reader comments, not staff) Again, Indy RBs have a high success rate and get the first few yards, in theory because of the holes, and then when they have to make yardage themselves, there is a bit less there...?

83
by Moridin (not verified) :: Wed, 10/18/2006 - 3:13am

Bobman, go by 2006 only numbers. Buffulo is 21 and Atlanta is 20.

84
by Becephalus (not verified) :: Wed, 10/18/2006 - 3:19am

Fnor, in regards to sports journalism they must die. They are shield to hide ignorance/laziness, why say anything meaningful about someone when I can just throw a bunch of meaningless superlatives out there.

85
by Bobman (not verified) :: Wed, 10/18/2006 - 3:20am

Moridin, okay, so I need reading lessons. Thanks for clearing it up for me. I criss-crossed the columns and thought it was last week. Maybe it's not a good time to try to balance the old checkbook....

86
by Jake (not verified) :: Wed, 10/18/2006 - 3:51am

A few of the more interesting developments that got squashed by the Bears comeback:

1. Cinci. Are they a playoff team? How did they manage to lose to a Gradowski helmed Tampa team?

2. Tampa. Of course, maybe Tampa's just good. They almost beat NO and Carolina (first game with Smith back). I'm inclined to think they aren't, and look at their sked, they'd have to play amazingly to do anything, but I still want to know how good they are.

3. Steelers. Are they the dominant team they used to be when helmed by Big Ben? Stomping the Chiefs is a start, but not enough to make people forget about the stinkers in previous weeks. Mediocre or dominant? I guess we'll find out.

4. NO. Ok, this is getting a tad ridiculous. I'm not a believer in this emotional lift thing, but now they've beaten the mighty Eagles. Their pre-season projection is epicly bad. Their DVOA is Top 10. So confusing!

5. Screw it. There are too many teams that are hard to get a read on. Is Denver a defensive powerhouse? Are they even good? Is Jacksonville a Superbowl contender? Dallas killed them. Are the Giants a playoff team? Is Mark Brunell good or bad? Will Seattle get its running game back? Are the Bears really that good? Are the Colts the best AFC team or not? Are the Chargers the best? What's the deal with the Chiefs? This is too much. I wonder what clear ideas the experts have on the league.

I'm thinking the Colts end up on top. I still think Peyton is the best player in the league by a lot.

87
by Bobman (not verified) :: Wed, 10/18/2006 - 4:31am

Becephalus, do you recall the old Steve Martin comedy routine in which, if he ever took hostages, he would request, in addition to money and a helecopter, that "the letter M be stricken from the English language?" You have a better case than he did.

Jake, I'm not sold on Indy yet and won't be til the second week of January. But they may be in the mode of "practicing" for close, last-minute, pressure-packed wins like the playoffs, rather than the 30 point stops of yesteryear. If that's the case, great. It flies in the face of FO wisdom that STOMPS are a better predictor of team quality. But they'll be "battle tested" as was the argument when NFC East teams beat up on each other en route to NYG, DAL, and WAS winning a bunch of SBs back in the 80s and 90s.

You know how people often talk about how come-from-behind 4th quarter wins are the mark of a great leader? Elway? Marino? Brady? I always figured that if their teams were better, they wouldn't need to. But Manning has 3 (or 2) this season and is on track to challenge Elway's career total. That includes a couple seasons where their typical margin of victory was about 2 TDs, meaning few opportunities for this type of win. So he clearly has the game well in hand once more; but we'll see if the OL and D hold together just enough to make it happen. Maybe Booger is the key, maybe keeping Sanders healthy at safety, maybe getting Stokely back.

88
by David (not verified) :: Wed, 10/18/2006 - 6:08am

You know, DVOA is cool! I'm so happy to see a list that actually notices when the Bengals lose two in a row, that the Colts are just barely winning most of their games, Jacksonville has played an incredibly tough schedule so far, and Dallas was absolutely dominating in all three of their wins.

I know that the answer to this is either Answer A or Secret Answer C:

C) I'll do it when I have the time.

But in addition to VOA, the Defense adjustment, and the preseason projection adjustment, there should be an injury adjustment.

When Steve Smith comes back from injury, the Panthers should immediately shoot up the rankings just because Delhomme is now passing to Smith instead of Carter. Teams would also lose points immediately whenever a player gets injured.

The hardest part would be determining the replacement's value (often being forced to use the average.) Also, rookie starters on injury would be forced to be compared to an average rather than their value from the previous seasons. So if Reggie Bush were injured tomorrow, there'd be no real way of knowing how he compares to Aaron Stecker (there's a name you don't hear too often.)

89
by cowfez (not verified) :: Wed, 10/18/2006 - 6:09am

Anyone else notice that Keenan McCardell and Eric Parker are second and fourth, respectively, in DVOA after finishing last season second and third? I don't think the chargers wideout situation is bad at all.
Also, i'm curious as to why I don't see more criticism of the Running Back rankings, since they appear to be the least accurate of FO's stats.

90
by cowfez (not verified) :: Wed, 10/18/2006 - 6:32am

I believe that the strength of the NFC East is an aberration. Is it possible that DVOA favors a powerful offense?

91
by Rick (not verified) :: Wed, 10/18/2006 - 6:43am

Hmmm...I don't quite get how the Patriots have the 22nd best defense when they rank 7th in points/game and 13th in yards/game. Pittsburgh, who give up 2 more points per game, have the 4th best D? I wonder what their relative performances were against common opponents - let's ask the Bengals and the Dolphins! (Yes, they both scored more points against the Steelers than they did against the Pats, and the Bengals actually beat the Steelers.)

So yeah, you've got the Steelers too high. I'm thinking they should be at least 7 notches lower. They've only had one impressive game so far. A 2-3 team is the 6th best team in the NFL? And that doesn't cause concern with you guys?

92
by Yaguar (not verified) :: Wed, 10/18/2006 - 6:50am

89: San Diego wide receivers constantly post high DVOA with comparatively mediocre DPAR. I don't think that's at all surprising. In words, not four-letter acronyms, that means that San Diego's receivers aren't involved in so many plays, but when they are thrown the ball, they do very well.

Why is that? Two names. The first is Antonio Gates, the second is LaDainian Tomlinson. The Chargers' offense centers on these two players, both arguably the absolute best at their positions. Defensive efforts center around stopping these two players as well. The receivers have a lot of room to work with.

Our stats tell us that McCardell and Parker are more than suitable for the Chargers' needs at receiver. We can surmise that adding a third big-time weapon to their offense probably shouldn't be as high a priority as, say, improving the secondary.

However, if you took away Torry Holt from the Rams offense and asked Eric Parker to fill his role, I'm betting the results would not be pretty.

93
by Sam! (not verified) :: Wed, 10/18/2006 - 8:48am

#86:

Dallas killed Jacksonville? When?

94
by Tom Kelso (not verified) :: Wed, 10/18/2006 - 9:38am

#40 Jerry --

Whatever the problem with the ratings, it isn't due to the projections being included. Carolina projected higher thatn both baltimore and New Orleans, so their DAVE only helps them.

As to why they are lower, then, it has more to do with the bad performances against Minnesota, Tampa and Atlanta than it does their current streak. As the season goes on, the weighting will make those fall away, and the Panthers will more accurately reflect their current state. That's been shown in every season DVOA has measured -- outlying early performances eventually lose their effect.

95
by Goathead (not verified) :: Wed, 10/18/2006 - 10:04am

NE: Past Schedule: 27 Future: 30.

See, that's coaching GENIUS. Most coaches are focused on 'positions' and 'plays', the NE staff was more intent on making sure they had an easy schedule :-) Wait till next year when they play the raiders 16 times...

96
by Rich Conley (not verified) :: Wed, 10/18/2006 - 10:41am

95: Did you see the Patriots schedule the last 2 or 3 years? They deserve a year like this. Now if we could only get Indy a tough schedule, for once.....

91: DVOA seems to be in love with the Patriots offence for some reason. They, according to DVOA, have had the 7th best offence in the league so far. Anyone who's watched thier games knows better. Theres something about their slow moving, ball control passing game that DVOA loves. To contrast that, DVOA hates their defence. I think its for the same reason, giving up a lot of 4 and 5 yard plays, yet stopping teams from scoring.

97
by Adam, VA (not verified) :: Wed, 10/18/2006 - 10:46am

RE: 96

What's not to love about lots of 4 or 5 yard plays for the offense? That leads to a lot of first downs, one of DVOA's biggest boosters is for teams making positive plays that lead to easier situations of getting first downs. Vice versa for the Pats D.

98
by Pat (not verified) :: Wed, 10/18/2006 - 11:16am

To contrast that, DVOA hates their defence. I think its for the same reason, giving up a lot of 4 and 5 yard plays, yet stopping teams from scoring.

Preventing scoring isn't everything a defense does. Giving up yardage means that you're losing field position, which makes things harder for your offense. Even if you stop someone from scoring on one possession, if they drive down to the 1 yard line and you stop them, their team is still more likely than yours to score again.

If a defense allows a team to drive down the field, and then gets an interception at the 1 yard line, the offense starts at the 1. What if the other team's offense fumbles the ball shortly after, and then the other team scores a touchdown? (See Philadelphia vs Green Bay, this year) Whose fault is that? The offense or the defense?

A defense's job, just like an offense's job, is to help the team win by increasing the point differential. A defense can do that by both limiting opposing scoring, and increasing their own scoring through beneficial field position. Likewise for offense - offenses don't just score. They also help prevent opponents from scoring.

You can't judge a defense solely by points allowed. You can't judge an offense solely by points scored.

Note New England's average defensive starting field position, by the way: best in the league. A lot of that is because the offense is better than average. Which means that opposing drives have to be longer in order to score. Which means you can give up more yardage without giving up points. But that doesn't mean that the defense isn't still worse than average on those drives.

99
by Pat (not verified) :: Wed, 10/18/2006 - 11:27am

4. NO. Ok, this is getting a tad ridiculous. I’m not a believer in this emotional lift thing, but now they’ve beaten the mighty Eagles. Their pre-season projection is epicly bad. Their DVOA is Top 10. So confusing!

The Saints are also really hurt by the fact that there are so many good teams this year. Their actual DAVE isn't bad: 4.0%. That would've been good for 12th in the league last year, ahead of New England. But now it's 17th.

I think that's the one real surprise this year more than anything else: probably about half the league is very good. Not just above-average good, which is obvious, but deep into the playoffs-good.

100
by Goathead (not verified) :: Wed, 10/18/2006 - 11:46am

96: Yes, the Pats had it rough last year. Except they got to compete with who for a playoff spot? Buffalo/NYJ/Miami. LOL, it almost doesn't matter who your out of division games are against when the other divisional teams have zero chance of making the playoffs.

101
by Jerry (not verified) :: Wed, 10/18/2006 - 11:51am

Becephalus: It was a simple, honest question. Your flippant, condescending answer was not needed.

I have been visiting this site since 03 and have done nothing but praise it: both here and elsewhere. However, just because it works, doesn't mean it can't be better. And just because it works, doesn't mean things can't be questioned!

102
by Jerry (not verified) :: Wed, 10/18/2006 - 11:55am

Pat: Thank you for the link to the weights.

Looking at them, it seems to me the system rewards the consistently good teams but punishes the surprisingly good ones. I don't need a book of stats and metrics to tell me Indy, NE and Philly are good based upon past performance. I do want to know though if NO really is the 10th best and if SEA really is 16th though.

For example: In 2003, the Panthers were 5-0 and had beaten TB (ranked 1st) and Indy ( ranked 5th). But the system, because it is so back loaded, only ranked them 20th. The Panthers went on to barely lose the SB that year but at no time did the system ever indicate they were better than 18th or –5.8 DVOA which meant they were a below average team.

In 2005, the Bears started off the season ranked 7th according to the FO site but were only ranked 21st on Fox because of their previous year and projections. It wasn’t until the latter part of the season that the system caught up and recognized their actual value.

So is NO really as good as the number indicte: my guess is yes. Was KC really a top 5 team before the Pitt game: no way. Is Chi really still a 2nd ranked team; I would say no. In all these cases, the past performance of a team is either artificially inflating or deflating their value.

There is no doubt that you have to utilize past performances in valuing a team. But giving a game that occurred 8 and 9 weeks ago a .99 and .93 value respectively doesn’t allow for very rapid movements in the rankings. Especially given that the season is only 16 games long.

Sorry for the long rant!

103
by RCH (not verified) :: Wed, 10/18/2006 - 11:59am

RE: 100 - if the goal is simply making the playoffs you would be right. However overall record is what drives playoff seeding and homefield advatange in the playoffs. Sure the Pats made the playoffs last year, but would have been in much better shape with a bye or 2nd round game at home.

104
by Wanker79 (not verified) :: Wed, 10/18/2006 - 12:00pm

Here's my take on ranking the divisions. I took the average DVOA for each division and then removed the team with the greatest differential between thier DVOA and the divisional average DVOA (I wanted to remove the outliers). Then I re-averaged the divisions using only the three remaining teams. Here's what I got:

NFCE: 29.17% (PHI/NYG/DAL)
AFCW: 20.20% (SD/DEN/KC)
AFCN: 13.70% (BAL/PIT/CIN)
NFCS: 5.37% (NO/CAR/ATL)
NFCN: -12.63% (MIN/DET/GB)
AFCE: -18.60% (BUF/MIA/NYJ)
AFCS: -24.87% (IND/TEN/HOU)

There are clearly 4 good divisions (2 from each conference) and 4 poor conferences. It's a remarkably steady decline (~8% each) between the NFCE /AFCW / AFCN / NFCS (m = 0.3658, R^2 = 0.9966). Then there's a significant drop between the NFCS and the NFCN (18.00% drop). And then a steadily increasing difference (1.67%, 4.30%, 6.27%) between the NFCN and the AFCS.

105
by JHA (not verified) :: Wed, 10/18/2006 - 12:01pm

Rick: The Steelers D has dominated in 2 wins and been pretty good in 2 losses. Their stats look bad in the loss to SD. (Miami and KC wins were dominating. Cinci scored 2 TD's on drives of 25 yards or less or so off of turnovers. Gave up only 3 FGs in loss to Jacksonville and not many yards). Against Jacksonville and SD the defense was on the field way too much in the 2nd half, with only SD good enough on offense to take advantage. That's why they rank so high on D.

106
by Wanker79 (not verified) :: Wed, 10/18/2006 - 12:04pm

Obviously, that should have been "There are clearly 4 good divisions (2 from each conference) and 4 poor divisions."

107
by Wanker79 (not verified) :: Wed, 10/18/2006 - 12:08pm

Re: 102

Those weights weren't just randomly guessed at. Those weights were shown the produce the best correlation to a team's actual quality. Just because it didn't work perfectly with every single team, doesn't mean that any change that would "correct" that particular team wouldn't make the correlation for the other 31 teams worse. There will always be exceptions to any statistical measurement. The goal is to be as accurate as possible for the greatest number of teams. We'll just have to live with the occational exception.

108
by Pat (not verified) :: Wed, 10/18/2006 - 12:10pm

In 2005, the Bears started off the season ranked 7th according to the FO site but were only ranked 21st on Fox because of their previous year and projections.

DAVE didn't exist last year. It's not the same system as it was last year.

But giving a game that occurred 8 and 9 weeks ago a .99 and .93 value respectively doesn’t allow for very rapid movements in the rankings.

Yeah. That's because history has shown that teams don't just radically improve over one week. Not even over four weeks. It's human nature to forget things rather quickly, but that's just media.

Those weights aren't random. See the commentary in the article I linked.

109
by Pat (not verified) :: Wed, 10/18/2006 - 12:15pm

Just because it didn’t work perfectly with every single team, doesn’t mean that any change that would “correct� that particular team wouldn’t make the correlation for the other 31 teams worse.

Yeah. I don't think a "if (team(year-1) == natural disaster && if (team(year-1) == homeless for most of the year) && if (team(year-1) == no home field advantage throughout the year))" clause would be all that helpful for other teams.

As for Seattle, DAVE is actually doing the right thing. The preseason projections told us "Seattle is a good team". Seattle looked really bad after that Chicago game. But Seattle's been trending upwards since then. And when Alexander comes back, DAVE's projection will probably be dead on.

110
by Jim Maron (not verified) :: Wed, 10/18/2006 - 12:23pm

I really like the work you outsiders are doing here. I read the 15 principles and they all seem bang on to me.

However, I'm having trouble understanding the vast discrepency between the rankings of teams with respect to their basic yards per play differential and the DVOA rankings. In many cases they seem to match up (SD, Philly, Chic at the top) but in some cases there is a wide discrepency (Minn 6 in yards per play differential at +.7, 19th in DVOA, Jacksonville 12th in yardage diffential, 4th in DVOA).

Is DVOA a better predictor of future success than yards per play differential?

111
by Jerry (not verified) :: Wed, 10/18/2006 - 12:37pm

107:
I understand that and I realize the best indication of future performance is past performance regardless whether it is football, baseball or interviewing for a new hire. I also know that the numbers do eventually catch up with a team toward the end of the season so use them then.

But at the beginning of a season, I just think there has to be some way to adjust the weights to better reflect new trends: whether they be up or down.

112
by Sandman (not verified) :: Wed, 10/18/2006 - 12:46pm

Cowfez (#89) raises a question I was thinking about too. Aaron, does the performance of Edge in Arizona and his replacements in Indy suggest that the DVOA formulas underestimate the impact of the OL on the running game and overestimate the running back's influence? Unless you think Edge got a lot worse in the offseason, the formula seems to have been giving him too much credit for his numbers on the Colts, no?

113
by Will Allen (not verified) :: Wed, 10/18/2006 - 12:49pm

Pat, maybe I misremembered, but I thought I read Aaron stating that in- season DVOA had an injury component for quarterbacks. Yeah, I understand that being so dependent on one player is indicative of a team's quality(although I think most good teams are very, very, dependent on their starting quarterback; how do you like Philly without McNabb, or Indy without Manning I?), but the fact of the matter is beating a visiting Carolina team when they don't have Smith in the lineup isn't even close to being the same sort of quality performance as doing so when he is in the line-up. Oh well, the world ain't perfectable, but it's something to think about.

114
by Pat (not verified) :: Wed, 10/18/2006 - 12:49pm

Is DVOA a better predictor of future success than yards per play differential?

Yes. It's in the FAQ, just after the 15 points. In the "Does DVOA really work?" section.

115
by Pat (not verified) :: Wed, 10/18/2006 - 12:52pm

Aaron, does the performance of Edge in Arizona and his replacements in Indy suggest that the DVOA formulas underestimate the impact of the OL on the running game and overestimate the running back’s influence?

The numbers for James (DPAR, specifically) are really "James + Offensive Line". There's no attempt to separate out the offensive line from James's performance. The offensive line stats are an attempt to do the other way (separate the offensive line from the running back).

DVOA is still far away from the point where we can use it to represent the value of a player separate from the performance of his ten teammates that are also involved in each play. That means that when we say, "Larry Johnson has a DVOA of 27.6%," what we are really saying is "Larry Johnson, playing in the Kansas City offensive system with the Kansas City offensive line blocking for him and Trent Green selling the fake when necessary, has a DVOA of 27.6%."

116
by dbt (not verified) :: Wed, 10/18/2006 - 12:53pm

#113: I think you're thinking of the DAVE projections, which can be adjusted based on replacement-level QB and so on. The Chiefs projections were altered after Green went down, but their DVOA is based solely on play on the field.

117
by Travis (not verified) :: Wed, 10/18/2006 - 1:12pm

Seattle looked really bad after that Chicago game. But Seattle’s been trending upwards since then.

This is not a criticism, but how much of this Seattle trend is the result of Seattle's previous opponents playing better in the last two weeks (specifically the Giants), rather than Seattle's own performance? Seattle's only played one game over that stretch, the last-second win over the Rams.

118
by Jerry (not verified) :: Wed, 10/18/2006 - 1:35pm

"That’s because history has shown that teams don’t just radically improve over one week. Not even over four weeks."

I both agree and disagree with that statement.

If everything stays the same - coaches and key players - then yes, teams don't radically change over a course of a couple of weeks or even months. But if there are major changes, than a teams performance HAS TO change. It is inevetable.

With a team like Indy or NE, the weights make perfect sense as there is very little change either with the coaches or players (at least the key players). But with teams like the NO, KC, STl etc, there is change...major change. And using numbers from previous games with a different coach or QB just doesn't make sense.

Given that change is so prevelent in the NFL, I would think it would make sense to adjust the weights according to any changes. If a team, such as NE, has pretty much remained intact, than use the numbers as is. If a team has replaced a QB, than use past games but weight them at a .9 or something and put a greater value on the more, immediate games he has played in. And if a coach has been replaced, than weigh the past games at a .5 or something.

Now those are just quick thoughts and I do not mean to be crtical of the system. I still think it is the best thing available. But I also think it can be improved and am just wondering - out loud - how it might be.

Thanks for the link and explanations Pat.

119
by Wanker79 (not verified) :: Wed, 10/18/2006 - 1:41pm

Hey, can someone with better math skills than me (I'm looking at you, Pat), help me out. I'm trying to use the "10+ Yards" (boom) and "Stuffed" (bust) offensive line stats to figure out the boom-or-bust quality of a team's running game, but I'm having trouble coming up with a suitable formula. I thought a simple ratio of (boom / bust) would work, and I guess for the most part it makes intuitive sense. But there are some odd results. Like Oakland being the 4th largest boom-or-bust team in the league even though they're 29th in Stuffed rank and 2nd in 10+ Rank. If you're getting big chunks of yardage at a time and stuffed very little, that doesn't really scream boom-or-bust to me.

Am I just over thinking this?

120
by David (not verified) :: Wed, 10/18/2006 - 1:58pm

I think you're misreading the rankings - teams with lower Stuffed ranks are being stuffed more, not less. It's operating on the principle that #1 should be the best at whatever the stat is measuring, regardless of whether more of it is good or less is. So the Raiders get most of their rushing yards in 10+ increments, which is partly due to the fact that a lot of their rushes go nowhere. Sounds like boom-or-bust to me.

Honestly, your first tip-off that the stats don't work like that should have been when they led you to say something good about the Oakland offense.

121
by David (not verified) :: Wed, 10/18/2006 - 2:02pm

Also, now that I've looked at the offensive line ratings for the first time this season:

WTF is going on with Seattle? 100% power success, with no Shaun Alexander? Are they throwing all the time on third-and-short, or is their running game/o-line doing better than advertised, at least in that part of the job?

122
by Marko (not verified) :: Wed, 10/18/2006 - 2:18pm

"You know how people often talk about how come-from-behind 4th quarter wins are the mark of a great leader?"

I was thinking about this yesterday regarding the Bears' come-from-behind win on Monday night (or, perhaps more accurately, the Cardinals' come-from-ahead loss). Will Rex Grossman be credited with a come-from-behind win, despite the fact that he contributed absolutely nothing to the 4th quarter comeback?

123
by David (not verified) :: Wed, 10/18/2006 - 2:20pm

I swear this will be my last post in a row: I was just reading over the Angry Troll Hatred posts on the blog, and realized that anybody who found the blog through Fox gets no indication that FO is anything but esoteric, over-analyzed blog posts on various football plays, in-jokes, stats and/or minutia. A prominent link back to the main site might help with the trolls.

124
by Travis (not verified) :: Wed, 10/18/2006 - 2:20pm

WTF is going on with Seattle? 100% power success, with no Shaun Alexander? Are they throwing all the time on third-and-short, or is their running game/o-line doing better than advertised, at least in that part of the job?

By my quick count, Seattle has had 12 power situations (3rd or 4th and less than 2, or goal to go from the 1 or 2). They've run on 9 of them, all successful, and passed on 3, 2 successful and 1 incomplete.

125
by Travis (not verified) :: Wed, 10/18/2006 - 2:23pm

A prominent link back to the main site might help with the trolls.

Do you really want them coming over here?

126
by admin :: Wed, 10/18/2006 - 2:23pm

FOX blog now features posts on how bad Edge and Grossman were this week. Click link.

127
by Wanker79 (not verified) :: Wed, 10/18/2006 - 2:32pm

Re: 120

You're right, I think I was reading the Stuffed stat backwards. But now the boom-bust-ratio makes even less sense. It puts SD at the #1 boom/bust team in the league even though they rank 1st in Stuffed Rank?!? How is it possible that the team that get's stuffed the least out of all the teams in the league also grades out to be the biggest boom/bust team?!? Something is definitely screwie with my calculations.

128
by Pat (not verified) :: Wed, 10/18/2006 - 2:33pm

If everything stays the same - coaches and key players - then yes, teams don’t radically change over a course of a couple of weeks or even months. But if there are major changes, than a teams performance HAS TO change. It is inevetable.

You're being a bit too simplistic. It's not that things don't change. They do. It's just that you can't get a good read on how they change from just one week - not two weeks, not three weeks, not four weeks. Any small change in a team will just look like normal variation until you build up enough data. So if you arbitrarily weight the recent data more, all you really do is just decrease the accuracy. And that's borne out by the actual data itself.

129
by Moridin (not verified) :: Wed, 10/18/2006 - 2:35pm

Ryan Leaf was 1/15... wow. I knew he sucked but that's horrid.

130
by Sophandros (not verified) :: Wed, 10/18/2006 - 2:36pm

Aaron,

In the Edge article, you say that the Saints lost the game at Jacksonville in 2003 after several laterals on a kickoff return. That play was from scrimmage.

131
by Sophandros (not verified) :: Wed, 10/18/2006 - 2:43pm

Wow. 7INTs. Who was Detmer's backup in that game? Me?

Seriously, I'm curious what the average fan would rate in PAR. At some point, we have to be PBR (and not the beer)...

132
by Rich Conley (not verified) :: Wed, 10/18/2006 - 2:44pm

re:98

Again, go back and watch the Patriots games this year. Thats not whats happening. The offense is consistently getting 20 yards and punting, and then putting together one long 15 play 85 yard drive at some point in the game. Whereas the defence is putting up a 3 and out anytime the opponent gets the ball past the 50.

Theres something that DVOA doesnt take into effect, or doesnt portray, and thats defensive design. Its pretty god damn obvious (if you watch the games) why the Pats give up the yards between the two 40s so easy. They play their corners like 15 yards off the LOS in the middle of the field. Its almost like they think that the middle of the field doesnt matter. Now, yes, that hurts the DVOA, but does the DVOA then accurately reflect what the defence is capable of? I dont think so.

Now, most "bend not break" defences are just mediocre defences, but I think thats just how the Pats play. IE the middle of the field doesnt matter all that much. The point is not to give up a huge play. It think DVOA has been underestimating the Pats D and overestimating their offense since 2003.

Remember those 2004 playoffs? When Indy with their +40% Offensive DVOA played NE with their -10% Defensive DVOA? And scored 3 points? Thats not supposed to happen.

Something still isnt right here. Either he model is still WAY too simple, or the fact that it treats all defensive schemes the same just doesnt work. I really think the model just doesnt agree with the Patriots defensive philosophy: stop the big play, and make teams beat you on 14+ play drives. Make them keep grinding and grinding, and chances are, one of the stud front 3/4 will make a big play, and force a bad throw, or a fumble, or a big sack before the other team scores.

A good example is the Buffalo/NE game at the beginning of the year. The Pats offense was AWFUL. GOD AWFUL. MIND BLISTERINGLY AWFUL. The defense wasnt bad. They gave up 10 points, and scored 2,and yet VOA shows the offense was better than the D. Anyone who watched that game knows otherwise.

DVOA is missing something here. To just say that DVOA is right without looking at specific games and ratings is ignoring what this site is all about. DVOA is becoming common sense around here, and while its a very good metric, its not nearly even close to perfect.

133
by B (not verified) :: Wed, 10/18/2006 - 2:52pm

129: Not just 1-15, but 1-15 for 4 yards with 2 Ints and 3 fumbled snaps. If he simply spiked the ball after every snap, he would have had better numbers.

134
by Pat (not verified) :: Wed, 10/18/2006 - 2:55pm

How is it possible that the team that get’s stuffed the least out of all the teams in the league also grades out to be the biggest boom/bust team?!?

Yeah, that ratio isn't going to work. First, dividing by "stuffed" isn't the right way to go - a low "stuffed" rank should make a team rank lower, since they bust less. Dividing makes that go higher.

I'd just subtract. Say, "(10+yards-league average)/(league average)" - "(league average - stuffed)/(league average)". I'd actually be a bit more complicated, and normalize it by standard deviations. So ((10+yards-league average)/(league average))/stdev of 10+yards, etc. 10+yards has a much higher standard deviation than stuffed, so you'd need to normalize for that.

Rankings like that should come out looking similar to just "stuffed rank - 10+rank".

135
by Pat (not verified) :: Wed, 10/18/2006 - 3:05pm

The offense is consistently getting 20 yards and punting,

20 yards and punting: way better than less than 10 yards and punting. The former gets you good defensive field position. The latter does not. 20 yards and punting is not a bad offense. 3 plays and 0 yards is a bad offense.

When Indy with their +40% Offensive DVOA played NE with their -10% Defensive DVOA? And scored 3 points?

1) You do realize -10% DVOA on defense is good, right?

2) You can't judge a defense's performance by points allowed! Yeah. They allowed 3 points in the first half. New England, with a 24% offensive DVOA, and Indy, with a 3.5% defensive DVOA, scored 6 points in the first half.

40%+-10% = 30%
24%+3.5% = 27.5%

Basically the same. Results in the first half: basically the same. Second half was a bit different, but that had more to do with Indianapolis's defense collapsing.

136
by GlennW (not verified) :: Wed, 10/18/2006 - 3:42pm

I didn't know where else to ask this question, but the NY Times apparently has the information (linked) on the flex TV scheduling for Weeks 10-15, i.e. in that period the four games each already protected by CBS and FOX from re-schedule to Sunday night (although the NYT doesn't list the games specifically), with follow-on projections of the more likely Sunday night games. Has more detailed information on the "protected" games been released (or leaked) elsewhere?

137
by Rich Conley (not verified) :: Wed, 10/18/2006 - 4:08pm

"1) You do realize -10% DVOA on defense is good, right?

2) You can’t judge a defense’s performance by points allowed! Yeah. They allowed 3 points in the first half. New England, with a 24% offensive DVOA, and Indy, with a 3.5% defensive DVOA, scored 6 points in the first half.

40%+-10% = 30%
24%+3.5% = 27.5%"

Again, you're not refuting my point. My point isnt that the overall DVOA is wrong. My point is that the components arent what they should be. In both those situations, the teams should have been doing 30% better than average on every play offensively. It should have been a shootout, not a 6-3 game. IE, if the average NFL play is 4 yards, they should have both been getting 5.3 yards plus on each play. In the 11 drives in the first half, there were 6 three-and-outs. According to DVOA, that game should have looked like KC/INDY the year before

KC O (27%) + Indy D (-1.3) = 25.7
Indy O (17.3%) + KC D(6.7) = 24

That game was exactly what DVOA says it should have been, a shootout.

DVOA is doing something, and consistently doing something, that artificially makes the Pats offense look better than it is, and the defense worse than it is.

138
by Jeff (not verified) :: Wed, 10/18/2006 - 4:22pm

Rich Conley, check out the drive stats page (linked from my name).

On offense, NE's drive success rate is 4th in the league at .725. This means (if I understand it correctly) that 72.5% of their drive series result in first downs or TDs. This leads to long drives and usually scores. And their yds/drive is 6th. At the same time, their DVOA is 7th. If anything DVOA might be underrating their offense.

On defense, their drive success rate is 11th - better than their DVOA (22). But they don't get turnovers (TO/drive is 22nd) and they give up some long drives (yds/drive is 22nd.) But check out their line of scrimage/drive - best in the league. Their defense is successful (points-wise) because it can afford to give up long drives. On average, the opponent has to go 75 yards to score a TD. But this isn't due to the defense. This is because the offense and the special teams keeps their opponents pinned at their own end.

If you look at these stats (all of which are averages of readily measured events) they back up the results DVOA is getting.

139
by Pat (not verified) :: Wed, 10/18/2006 - 4:29pm

It should have been a shootout, not a 6-3 game.

I don't see why it's really important that DVOA predicts the actual pace of the game, rather than the relative point differential. In a close matched game, the pace of a game can frequently be dictated by special teams moreso than anything else, as well as turnovers.

140
by Pat (not verified) :: Wed, 10/18/2006 - 4:35pm

I guess the other problem is that it might be difficult for some people to realize that a 17-3 game can actually be an offensive battle rather than a defensive battle. I've maintained for a long time that that Indy-New England game (which was 20-3, I know) was more about New England's offense beating Indy's defense than New England's defense beating Indy's defense. When you drive 42 yards from the 10 yard line and pin the opposing offense inside their own 20, that has to be considered a win for the offense. Three times in that game, Indianapolis's offense and special teams were able to pin New England inside the 20. And three times their defense gave up drives the length of the field for points. It's not the offense's fault they had to punt - that's going to happen. Even in Indianapolis, the average offensive drive isn't 80 yards long.

141
by Rich Conley (not verified) :: Wed, 10/18/2006 - 4:36pm

Jeff, and there exactly lies my issue, like you said, their drive success rate is 11th in the league, and yet their DVOA is 22nd because they dont cause enough turnovers. I think turnovers are being overvalued here.

The Patriots offense gets a huge boost because of very few turnovers, and their defense a huge penalty for the same.

Yes, turnovers are great, but you dont need a ton of them to win games. That Indy/NE playoff game (20-3), there was 1 turnover, the whole game, but you can't say that NE'd D didnt have a huge game.

IIRC, during the 2004 superbowl run, the Pats redzone (defensive)DVOA was extremely high, whereas between the 20s, it was low. That says to me, that maybe DVOA isnt in line with what teams feel is important. In some games (IE situational,) field position ISNT that important, but points are, as points win games.

142
by Pat (not verified) :: Wed, 10/18/2006 - 4:44pm

IIRC, during the 2004 superbowl run, the Pats redzone (defensive)DVOA was extremely high, whereas between the 20s, it was low.

I really, really don't get what the problem is. New England's defense was -11% in 2004. Sixth in the league.

Situationally, it might be better than that against any given team. So what? You really expect to be able to characterize a defense with one number?

In some games (IE situational,) field position ISNT that important, but points are, as points win games.

That's only if you've got a good offense, though. If you couple that kind of a defense with, say, Oakland's offense, you have a disaster.

It's the same problem Denver fans have. For some reason, they see low scoring games and think it's the defense that's carrying the team. Nope. It's the offense (which is getting them out of crappy field position) plus special teams. It's not a surprise that the two teams we're talking about here (Denver and New England) have extremes on one side of the drive stat line of scrimmage.

143
by Jeff (not verified) :: Wed, 10/18/2006 - 4:50pm

Rich,

Their yards/drive is 22nd. Their TO/drive is 22nd. But their defense has the best starting field position in the league.

Yes, they don't give up many points. How much of that has to do with giving up a lot of yards and not getting TOs and how much of that has to do with the opposing offense always having a long field?

If the NE defense started every drive with the opponent having to go 95 yards and it gave up no points, is that the defense being good or is that the special teams/offense getting good field position for the defense?

144
by Pat (not verified) :: Wed, 10/18/2006 - 4:51pm

Or, stated another way, pundits would consider ROBO-PUNTER's defense would be the best in the league.

145
by Rich Conley (not verified) :: Wed, 10/18/2006 - 5:02pm

"If the NE defense started every drive with the opponent having to go 95 yards and it gave up no points, is that the defense being good or is that the special teams/offense getting good field position for the defense?"

Its both. In a situation like that, a defence's job is simply to keep big plays from happening.

I think my issue is that too much emphasis is being placed on field position here. Yes, NE's defence benefits from having a good offence, and good starting position. A defence that always starts with good field position, and knows its offense can move the ball has different responsibilities than a defence that is always starting in poor position.

As to Denver, really? DVOA has their offense as below average(-5.6% 19th) and their defence, as much better than average(-14.9% 8th), so if thats the case, then DVOA is wrong about denver.

146
by Miles (not verified) :: Wed, 10/18/2006 - 5:12pm

Curious how Dallas' punts have a -3.5 DVOA, but lead the league in Gross Punting and 2nd in net punting. Does it mean that Dallas punts from places that FO expects everyone to hit 60 yard punts the vast majority of the time? (BTW, this was true even before McBriar had his fumble).

147
by Jeff (not verified) :: Wed, 10/18/2006 - 5:12pm

Ok, Rich. So if I understand you, you're saying

NE's defense is clearly ranked too low because it knows that it's offense and special teams are good so it'll always have good field position.

148
by Pat (not verified) :: Wed, 10/18/2006 - 5:17pm

I just don't get it. Total DVOA in every case so far looks dead on correct with NE, with DEN, etc. The sole difference seems to be that the fans are saying "The offense is the problem, their defense is fantastic". Why is it so hard to believe that your opinion of "offense" and "defense" might be different than what DVOA calls "offense" and "defense"?

Its both. In a situation like that, a defence’s job is simply to keep big plays from happening.

In a situation like that?? The defense has the opposing offense on the 5 yard line, and their job is simply to keep big plays from happening? That just doesn't make any sense. If the defense doesn't give up yardage, that's a guaranteed 3 points for the offense. Giving up 40 yards in that situation is exactly the same as giving up a field goal. The only time that's true is if the defense is in prevent, and the clock is more important than field position.

Not giving up a big play is a perfectly valid strategy, but it has its drawbacks. It hurts your offense. It shortens the game. In some teams' cases, they're better built to handle this. That's fine. But it doesn't mean that the defense is great, in an objective sense. It might mean it's an ideal price/performance benefit for the team, but it's not objectively great.

149
by Pat (not verified) :: Wed, 10/18/2006 - 5:21pm

And actually, let's be a bit saner about this: compare New England's 2006 defense with its 2003 defense, which was ranked 2nd in DVOA.

Is New England's defensive strategy really dramatically different from 2003 to now? Isn't it possible that New England's 2003 defense was, in fact, much better, and it's only their offense, which is significantly higher than it was in 2003, which is hiding the weaker defense?

150
by PatsFan (not verified) :: Wed, 10/18/2006 - 5:27pm

Reading about the Pats having the best defensive starting field position encouraged me to look at the special-teams stats. The Pats are -5.4 points on FG/XP but are +6.0 points on kickoffs. Not surprising given Gostkowski's misses and his booming kickoffs. For grins, I looked at the 2005 final stats. The Pats were +0.1 points on FG/XP and +0.4 points on kickoffs.

So combining the two place-kicker related items, this year so far the Pats are +0.6 points while last year they were +0.5 points.

I'm certainly not going to claim this vindicates letting AV go, but I do think it does point out that the general public seriously underrates the value of good kickoffs re: field position.

151
by Wanker79 (not verified) :: Wed, 10/18/2006 - 5:27pm

As to Denver, really? DVOA has their offense as below average(-5.6% 19th) and their defence, as much better than average(-14.9% 8th), so if thats the case, then DVOA is wrong about denver.

Yes, because Denver is even a more extreme case than NE. The problem is that DVOA is undervaluing field position. Denver's offense is actually probably better than the numbers and their defense probably worse because their opponent's offense begins with the 11th worst field position in the league while Denver's offense begins with the league's worst field position.

Likewise, NE's offense begins with a middle-of-the-pack 15th best field position while thier opponents offense begins with the league's worst field position.

152
by Pat (not verified) :: Wed, 10/18/2006 - 5:37pm

Actually, I think DVOA is valuing it dead on. If you just use Denver's points scored, Denver's third last in the league. Their offense is about 42% below the league scoring average. Saying it's only 5.6% below league average, and only 13th worst in the league is a big, big jump.

153
by Rich Conley (not verified) :: Wed, 10/18/2006 - 5:41pm

Another Example where DVOA has issues with the Patriots:

Last year's Jax game in the playoffs.

Jax O Pats D gives you 6.6 + 10.5 = 17.1
NE O + Jax D gives you 12.8 + (-10.2) = 2.8

Should have been a blowout, and it was, but the wrong way. Theres something about the way the Pats play football that DVOA has problems with. I really think it has to do with the defence being willing to give up yards, at times when yards dont matter. (and before you bring up Leftwich,etc, he wasnt the issue, his line getting killed was)

Does DVOA have a score differential in it? I dont remember, but I think not. It seems like a lot of the time, the Patriots are up 24-6, and end up winning 24-20, when the entire second half was essentially junk time.

I think that really messes with DVOA, and is a problem.

154
by Led (not verified) :: Wed, 10/18/2006 - 5:50pm

This discussion of the relative strength of the Pats offense and defense is similar to the situation with the Jets in 2004. The Jets were 5th and 11th, respectively, in offensive and defensive DVOA, but their ranking in points scored and points allowed was reversed. Both the 2006 Pats and 2004 Jets had consistent, run-heavy, ball control offenses with few turnovers -- which DVOA loves -- and defenses that gave up a lot of successful plays but not a lot of points to poor offenses -- which DVOA likes not so much. I'm not so sure there isn't a minor glitch in the way DVOA allocates that type of team's success as between offense and defense, although it's remarkably accurate in terms of the team's overall quality.

155
by MDD (not verified) :: Wed, 10/18/2006 - 5:50pm

Mike Brown has surgery to repair ligament damage.

Link on name.

156
by MJK (not verified) :: Wed, 10/18/2006 - 6:04pm

Aaron has actually discussed the Pats offense vs defense this year in his Fox blog. He pointed out that the Pats defense is below average overall, but one of the top defenses in the red zone, and one of the top defenses in close games. That means that they've been allowing a lot of yards and first downs when they don't matter as much, and tighten up when the game is on the line. That is why their defense has seemed so good lately.

I think everyone here agrees that "clutch" play is not real but a result of cherry picked statistical anomolies, so maybe the low overall DVOA rating for the Pats defense is telling us that this exceptionally good play in close games is not sustainable.

On the other hand, better play in the red zone IS real and probably is sustainable. So the next issue is whether or not DVOA counts red zone performance enough as compared to performance between the 20's. It already counts red zone play a little more, but maybe not enough? Didn't Aaron comment not too long ago that he tried increasing the weight of red zone play, and it made DVOA's correlation to wins worse?

I guess the key issue is how important is field position? And there is no easy answer to that. Any coach will say that it depends on the game. It's relatively meaningless when the offenses are better than the defenses, and vital when the defenses are better. But including that in DVOA would be a nightmare, and would kind of defeat the point. The point of DVOA isn't to introduce a complex model that optimally lumps factors together into one or two nice numbers that always can be compared between teams. It's to break all the complex aspects of the game apart so that matchups can be analyzed. So you could look at the Patriots DVOA and say "Well, they'll probably give up a lot of yards, but their red zone DVOA is really good, and their opponent's is bad, so don't expect the opponent to score a lot of points. On the other hand, the Patriots offense will probably be starting in poor field position all game, so don't expect them to score as much as you would expect either".

As to turnovers, I think it's a matter of philosophy. The problem with turnovers is that they happen relatively rarely, and often somewhat randomly (although there are teams that are better than average at forcing them), but can have devastating effects, especially in field-position (defense dominant) games. You could leave them out of your ratings and not get such screwy looking, variable numbers, but that would unfairly penalize teams that actually are better than average at causing them. Or you include them, and get some wild swings. Which is the more valuable approach?

By the way, no coach, including Belichick, is willing to simply let the other team march out of the shadow of their own endzone. If you can force a 3-and out from inside the opponent's 10, you do so--it's a huge advantage. The Patriots this season have pinned opponents deep very frequently, and their defense has let them escape every time. That's not a sign of an elite defense, no matter how few points they give up.

157
by Pat (not verified) :: Wed, 10/18/2006 - 6:08pm

I’m not so sure there isn’t a minor glitch in the way DVOA allocates that type of team’s success as between offense and defense,

Isn't it more likely that it's our perception that's wrong? The only way DVOA assigns things to offense and defense is which of them is on the field.

158
by MJK (not verified) :: Wed, 10/18/2006 - 6:10pm

But the detailed DVOA breakdowns that Aaron has tell us all that. The problem, if there is one, is probably not with "DVOA" but rather with the method used to lump the different pieces of DVOA together into a composite defensive (or offensive, or total) ranking.

Actually, I'm not really clear exactly what that method is. If a team has X DVOA against runs up the middle, Y against runs off Tackle, Z against Medium passes, Q against short passes, and R against long passes, what is their total defensive DVOA? I'm sure it's not just X+Y+Z+Q+R. Are the different DVOA's weighted by how often each situation comes up on average? If so that could be a source of some inaccuracies, for teams that don't confrom to average playcalling trends. Are they weighted by how much the different situations actually came up so far over the year? If so, that could cause issues and make teams seem worse, because other teams attack a team's weaknesses so it would over-weight what a team is bad at.

159
by Jerry (not verified) :: Wed, 10/18/2006 - 6:10pm

Pat, I think you are misinterpreting what I am saying. I certainly would never suggest arbitrary changes. I was thinking more along the lines of a value/number system for a coach or QB and adjust weights according to it. If a team gets a new coach, then discount last years numbers by say 40% (.6). Get a new QB, then discount last years numbers by 10% (.9). If a team gets a new coach and QB, then discount last years games by .5.

For example, NO this year has a new coach and QB. Do the numbers from last year really matter this year given all those changes? They do but not nearly to the same degree as they do to say NE. So discount NO’s numbers for last year. Example:

Here are the weight numbers currently being used for NO through week 5

.20 .67 .67 .67 .70 .95 .95 .95 .95 .95 .95 .95 1.0

Here are the weight numbers for week 5 after adjusting for the new coach and QB

..10 .335 .335 .335 .35 .475 .475 .475 .475 .95 .95 .95 1.0

And here are the weight numbers for week 6

.....10 .335 .335 .335 .35 .475 .475 .95 .95 .95 .95 .95 1.0

You are discounting last year’s games while still using your original weights for the games this year. And of course you would do the same thing to any team with a new coach, QB, etc.

You could even carry it further if you wanted. Say Brady gets hurt and misses a couple of games. The team is still basically the same so the past numbers should still be used. But in the games Brady misses, the weights would be discounted by 10% (.9). For example, say Brady goes out in week 8 and misses 4 games. This is what weight numbers would look like.

Here are the weight numbers if Brady did not get hurt through week 12

.20 .67 .67 .67 .70 .95 .95 .95 .95 .95 .95 .95 1.0

And here are the weight numbers if Brady misses 4 games

.20 .67 .67 .67 .70 .95 .95 .95 .95 .855 .855 .855 .90

And here are the weights 2 games after his return

.20 .20 .20 .67 .67 .67 .70 .95 .95 .769 .769 .769 .769 .95 1.0

So basically all the games Brady missed are adjusted to reflect his loss. Opponents do not get as much credit if they beat NE and NE’s ratings aren’t penalized as much as well. If Brady gets hurt and goes on IR however, then the old weights would still be used because the change is not temporary and future performance is more accurately predicated with each game played.

Now I admit, the values I selected -.5 for a coach and .1 for a QB - are arbitrary. I really do not know what they would need to be to accurately reflect a coaching or QB change. But with trial and error, I don't see why the correct numbers couldn't be found.

160
by Jeff (not verified) :: Wed, 10/18/2006 - 6:12pm

"Another Example where DVOA has issues with the Patriots:
Last year’s Jax game in the playoffs."

0-4 Tampa Bay just beat 3-1 Cincinnati.
0-5 Detroit just beat 2-3 Buffalo.
0-5 Tennessee just beat 2-3 Washington.
1-4 Arizona just almost beat 5-0 Chicago.

Upsets happen.

161
by B (not verified) :: Wed, 10/18/2006 - 6:17pm

153: More than just an upset, NE's 2005 DVOA was compiled with a significant number of injured over most of the season, most of whom were healthy and playing in the Jax game.

162
by Jeff (not verified) :: Wed, 10/18/2006 - 6:21pm

re 158:

Total DVOA will be weighted by how often each situation comes up. Total DVOA gets a value for each play and sums it up. The component values (throws to #1 WR, etc) get filtered out later.

When you analyze any single game, you have to look at the matchups, you can't just compare overall DVOAs. (Aaron's mentioned this before, but I can't actually find a quote from him.)

163
by Rollo (not verified) :: Wed, 10/18/2006 - 6:24pm

153: The FO game preview noted that the teams were trending the opposite way - Jacksonville's adjusted DVOA was way down while New England's was way up. This was in part due to Jacksonville injuries late in the season and New England ones early. Based on the adjusted DVOA, which all but ignored September, the result should have been a close pats win after you factored in homefield advantage.

164
by Jeff (not verified) :: Wed, 10/18/2006 - 6:25pm

Found some quotes:

From http://www.footballoutsiders.com/2006/10/10/ramblings/dvoa-rankings/4367...

Comment 32"
"We should all think that nobody is supposed to judge the validity of a statistic by a single game result."

Comment 86:
"I’ll say this again, since it seems I have to say it at least four times a year: DVOA is not meant to be taken as a sure-fire method of predicting that one team will beat another. There are just way too many issues involved in deciding which team should be favored in a game: individual matchups, strategic matchups, injuries, home field, etc. If it was always “higher DVOA should be favored,� I wouldn’t write those weekly game previews that drill down into specifics."

165
by Rich Conley (not verified) :: Wed, 10/18/2006 - 6:34pm

re:156

MJK, again, you're the only one who understands what I'm talking about here.

"Aaron has actually discussed the Pats offense vs defense this year in his Fox blog. He pointed out that the Pats defense is below average overall, but one of the top defenses in the red zone, and one of the top defenses in close games. That means that they’ve been allowing a lot of yards and first downs when they don’t matter as much, and tighten up when the game is on the line. That is why their defense has seemed so good lately.

I think everyone here agrees that “clutch� play is not real but a result of cherry picked statistical anomolies, so maybe the low overall DVOA rating for the Pats defense is telling us that this exceptionally good play in close games is not sustainable. "

Thats exactly the Crux of it. Generally, the Patriots have been phenomenal in the Red Zone, and so/so elsewhere. The problem is, deciding which is more important. Like you said, the Patriots defence seems to be very good when it needs to be, and mediocre otherwise. That "very good in close games, and very good in the redzone" screams to me that the patriots defence giving up lots of yards isnt an inability to stop people from getting yards, its that, for some reason, theres no importance given to it. In the Miami game this year, near the middle of the field, the pats corners were playing 15 yards off the line. As soon as you got to the 30s, suddenly it was all press coverage. Theres a reason for that: the coaches feel that stopping a deep pass there is more important that stopping short plays.

Maybe theres some aspect we're all missing here. It seems like the Pats give up long methodical drives, when theyre up points. I know a stop would be better, but in the 4th quarter, when you're up 14, is giving your opponents a 14 play 7:00 fieldgoal drive bad?

166
by Jeff (not verified) :: Wed, 10/18/2006 - 6:49pm

Rich,

I understand your point. But we're talking about a single number that's supposed to sum up a team's entire defense. They give up a lot of yards, but they have a lot of field position to work with. They don't get many turnovers, but they don't give up many points either. They're bad in the middle of the field, but tighten up at the red zone. But team defensive DVOA is one number to sum up all of that.

If they're up 14 in the fourth quarter, a long drive that nets a field goal is fine. But if they're down 3 in the fourth quarter, a long drive of any sort is not fine. DVOA factors in "down and distance, field location, time remaining in game, and current scoring lead or deficit." You can see the link at my name. All of these things you've mentioned are considered when DVOA is calculated.

But it's one number that needs to sum up their entire performance. So far they just haven't been that good.

167
by Rich Conley (not verified) :: Wed, 10/18/2006 - 7:10pm

166
"These include down and distance, field location, time remaining in game, and current scoring lead or deficit. "

It looks like its just whether you're ahead, or behind, not by how much.

The thing is, everything is weighted by how important they feel things are. For example, plays in the red-zone are given an extra 20% weighting. Who says 20% is correct? They do, because thats what correlates best to the data (Thats what they usually do with weightings, I assume aron did it with this too). The problem is, that weighting is how things correlate to the average, which makes the assumption that teams are trying to do the same thing. Once you start dealing with playcalling, and all the more situational stuff, those weightings may not be accurate. IE Red Zone offensive performance may be more important to a team with a bad defence, than it is to one with a good defence. That changes playcalling, and sometimes makes teams play against thier strengths. That doesnt mean the strength isnt there.

For example, Arizona the other night, once they got inside the 20, they went into a super conservative offensive game plan. It was obvious that getting atleast 3 points was more important to them, than getting 7. DVOA says thier offence sucked there, when in reality, their offence was fine. It got exactly what it was trying to get, a guaranteed 3 points.

You can bundle it together and say playcalling and offense are one, but I think that ignores something very important: Gameplans change on a week to week basis, team quality does not. I think thats part of the reason why Tampa, NE, and a couple other teams seem to show really wierd DVOA results every once in a while. Playcalling/gameplanning that doesnt have the same goals that DVOA assumes teams have.

168
by Jeff (not verified) :: Wed, 10/18/2006 - 7:21pm

I would say Arizona's offense did suck there. If an offense gets inside the 20, they should be expected to score more than 3 points. Had any of those drives ended in TDs, Arizona would have won. Arizona's O-coord has been fired, in part, because of that conservative play calling. Their offense was not fine in that situation.

What is your issue with DVOA? Gameplans change week to week, as does matchups. FO writes game previews because of this. DVOA will not predict every single play, quarter, half or game of the season correctly. Nothing will.

FO continually tries to tweak their system to make it better, but it will never be perfect. You can find any number of examples of where it's "wrong", but that's not going to prove anything.

169
by Jim (not verified) :: Wed, 10/18/2006 - 7:24pm

Question - does red zone performance have a greater variation than normal field performance?

170
by Rich Conley (not verified) :: Wed, 10/18/2006 - 7:31pm

Jeff,

I dont have a problem with DVOA. I just think it could be better, and I'm trying to figure out exactly why.

171
by Cid (not verified) :: Wed, 10/18/2006 - 7:48pm

"Jeff, and there exactly lies my issue, like you said, their drive success rate is 11th in the league, and yet their DVOA is 22nd because they dont cause enough turnovers. I think turnovers are being overvalued here. "

Based on what? Turnovers are consistently named by coaches as the single most important stat when it comes to winning games. And unlike, say, rushing attempts, they are a cause of winning, not just a product of winning.

A defense that turns the ball over has the power to create points for its offense through extra possessions and better field position. Offenses fail to score more often than they succeed in scoring, which means that most drives end up being judged on how they add or take away from field position. One turnover can have the effect of multiple successful (non scoring) drives on offense, or multiple 3 and out situations on defense.

Some defenses (such as Chicago) actively seek a structure which allows them to score points off of turnovers. I think we just saw the ultimate example on Monday Night of how much that can matter in winning a game. If the Bears had merely forced a lot of 3 and out situations, they'd have looked good according to a lot of defensive statistics, but they'd have lost the game. So how much value is too much value when it comes to creating turnovers?

172
by CaffeineMan (not verified) :: Wed, 10/18/2006 - 7:55pm

This is an interesting discussion about the importance of field position and giving up yards/drives. Thanks to Pat/MJK/Rich and others.

Rich Conley says:

That “very good in close games, and very good in the redzone� screams to me that the patriots defence giving up lots of yards isnt an inability to stop people from getting yards, its that, for some reason, theres no importance given to it.

See, I think you have a point right up until you say the above. The problem is, I don't think you can reliably conclude what you've just concluded. I don't think you can disprove that statement from the numbers, either. I think that's the point at which the data stops and the opinion begins. You're right to point out that DVOA can't take into account game planning. But we can't really tell whether the Pats are game planning to stop yards and failing in execution, or deciding to give up yards in favor of preventing big plays, when they could also choose to give up fewer yards.

My opinion is that you've got it right that the Pats D is more deliberately playing in a bit of a "prevent big play" mode, but you've got it wrong that they could give up fewer yards and are choosing not to do so. I think the Pats are doing this because they don't have enough talent on D to refuse the yards. They're not as good as in 03-04, so Belichick et al are doing damage control (at which I think they excel) by attempting to limit big plays until the offense runs out of room to maneuver in the red zone.

One reason that I think that it's not deliberate is that when Belichick ran the Giants D back in the 80's, they didn't give up diddly, with ranks in yards against of 5,11,3,4,8,16,5,2 (1983-1990). They had the talent to stop the yards, so they did. I think the same would be true for the Pats today if the talent would support that approach.

173
by Scott C. (not verified) :: Wed, 10/18/2006 - 8:15pm

Rich Conley:

Another Example where DVOA has issues with the Patriots:

Last year’s Jax game in the playoffs.

Jax O Pats D gives you 6.6 + 10.5 = 17.1
NE O + Jax D gives you 12.8 + (-10.2) = 2.8

Should have been a blowout, and it was, but the wrong way.

OK I had to have a good laugh at this one.

17.1 - 2.8 is approximately 15% DVOA differential. You didn't factor in the special teams bit between the teams, BTW.

Anyhow, where do you get the Idea that a 15% DVOA differential is a blowout?

This does not mean each play gets 15% more yards on average for one team versus the other. It means, that each play is likely to generate 15% more POINTS.

If the average team scores 22 points, that means DVOA predicted the Jags to win by 15% of that.... 3.3 points.

Blowout indeed.

174
by Scott C. (not verified) :: Wed, 10/18/2006 - 8:25pm

To further elaborate, it predicted that on a neutral field JAX would score 17.1% more points than the average, or about 26. It also predicted that NE would score about 2.8% more than average, or about 23. Given the nature of the game and how a lucky bounce or gain of a couple inches can skew a NFL game by 7 points for either team fairly easily, it was giving JAX the edge (on a neutral field) but in no means a blowout or dead-lock for a win. Even a 60% DVOA differential is only predicting about a 13 point difference -- note how rare vegas odds are equal to or greater than this number.

As others have mentioned, there were recent trends and injury analysis to factor in (that DVOA does not do) as well.

175
by B (not verified) :: Wed, 10/18/2006 - 9:17pm

169: There are less red zone plays than non red zone plays, so with the smaller sample size, greater variance is expected.

As a Patriots fan, the thing I don't like about thier defensive performance as they seem to only play well in important situations It's like they know they only have a certian amount of good plays in them, and they save them for when it's most important. If they could play that way all the time, then they would really be dominant, and I'd expect them to go deep in the playoffs. As it is, I see a flawed team that's not good enough to make it past the divisional round.

176
by DavidH (not verified) :: Thu, 10/19/2006 - 12:54am

153, 173, 174:

I think DVOA actually refers to "success points." These can in theory be converted into points, so 173 gets the gist of it right.

But... Several things need to be pointed out.

1) Maybe I missed it, but I don't remember ever reading that you can just combine TeamA's offense and TeamB's defense in some simple way to get an accurate prediction for that pairing. But if you really want to make a simple matchup calculation, my guess would be that it'd be better to center the percentages around 100% and multiply:

Jax O Pats D gives you 106.6% * 110.5% = 117.8%
NE O + Jax D gives you 112.8% * 89.8% = 101.3%

(Think of it the same way you would a price adjustment. Taking 10% off then adding 10%, or vice versa, does not leave you back at the same price.)

2) Even if you have an accurate DVOA advantage for one unit over the other, this doesn't mean it is going to show up in that unit's score. Half of this thread's discussion about the Patriots has been about that. If the offense has a 15% advantage, but keeps starting with terrible field position, they probably still won't score 15% more than average. Instead, some of that 15% margin will be in points that are taken away from the other team's score (by handing THEM bad starting field position).

3) Remember, special teams are 1/7 of the game. (It's in the 15 Things We've Learned, it must be true!)

------

And, on a side note, this stuff is making me think about the attribution of success points. On drives where the offense gets a lot of yardage, but no points, the offense gets some success points, but the real benefit in the scoreline is on the defensive side of the ledger. Just for fun, I'd like to see what the numbers look like if the success points are assigned like so:

a) most drives: credit/blame is assigned in the regular way
b) drives by TeamA starting in their own territory AND that do not result in points for TeamA: reverse - TeamA's DEFENSE gets credit, and TeamB's OFFENSE gets credit.
c) TeamA kicks/punts to TeamB: TeamA's DEFENSE gets credit, and TeamB's OFFENSE gets credit.

This would of course be stupid if you wanted to really evaluate the different units' play. But I'm wondering if it would correlate a little better to the common perception of a team's offense and defense. I think the biggest factor that leads to over/undervaluing of an offense/defense is starting field position. When calculated this way, the offense/defense would actually get a boost in their rating by being handed good position.

Like I said, I think this is stupid for evaluation purposes, but it would be interesting to see if doing this would make the DVOA numbers match public perception a little more. If so, it'd be a pretty damning strike against public perception, and a point in favor of DVOA.

Aaron, if you were my uncle, I'd ask for this for Christmas. If it weren't for that whole "Jewish" thing.

177
by DavidH (not verified) :: Thu, 10/19/2006 - 1:41am

159:
Last year's numbers are not used in this year's DVOA. (They're used in the preseason projection, which is part of DAVE, but that's another matter.)

Other than that, I think you have a pretty cool idea there. Every week a team could have both a "full strength" and "current" version of DVOA. The full strength one would designate the team's ideal/preferred players as the starters, and then adjust down the importance of weeks missing those players. The current one would treate the players currently expected to play as the "true" starters, and adjust down the weeks where they weren't playing. (In other words, Seattle's "current DVOA" would reduce the weight of weeks where Shaun Alexander played, while it's "fell strength DVOA" would reduce the weights of the weeks where he didn't play.)

I've got a sneaking suspicion the FO staff have had an idea like this before and decided it's either too complicated, too difficult, or too subjective. Or, they actually checked it out, and it just plain didn't improve the numbers.

178
by Rick (not verified) :: Thu, 10/19/2006 - 9:04am

Re: 175

"As a Patriots fan, the thing I don’t like about thier defensive performance as they seem to only play well in important situations."

I don't quite follow. Who cares if the Pats let the Jets get a couple TDs after they'd already built a solid lead?

"It’s like they know they only have a certian amount of good plays in them, and they save them for when it’s most important."

Well, if that's what you think...

"If they could play that way all the time, then they would really be dominant, and I’d expect them to go deep in the playoffs."

And maybe they'd make the Super Bowl! That would be a new thing for them!

"As it is, I see a flawed team that’s not good enough to make it past the divisional round."

The Pats' most glaring flaw is at WR, and that alone shouldn't hold them back. I have found more troubling a tendency to give up long passes. But I don't think that will persist if the Pats can field both Hobbs and Samuel at the same time.

I like DVOA but I think some of the people here are a bit _too_ into it. When it says a 2-3 team is the 6th best team thus far in the season, it is taking its own statistical projections too seriously. (And people who say things that diminish the importance of "points allowed" are just barmy.)

The Patriots' D has given up the 7th fewest points in the NFL per game, and is rated at 22nd. Their O has scored the 11th most points in the NFL per game and is rated 7th. When I point out this curiosity, people here starting pulling arguments out of their nether regions, like saying "their D only looks good because their O is doing such a great job." Anybody who had been following the Pats closely would know that their O has been erratic but their D has been carrying the team.

I suspect the reason DVOA doesn't like the Pats' defense is because the Pats have not been getting a lot of turnovers.

179
by B (not verified) :: Thu, 10/19/2006 - 10:17am

Watching the Pats, it looks to me that most drives go like this: Pats get the ball ~20 yard line, and drive to about midfield. On 4th down they put, pinning the opposing defense inside thier 10. If the defense got a 3 and out here, the Pats O would get the ball around midfied, but instead what happens is they give up a couple quick first downs, and don't get a stop till about midfield. Then it's another punt, and the offense starts again in bad field position. If the defense was playing better, the offense wouldn't have as far to go to get a touchdown, and they'd be scoring more points.

180
by Fnor (not verified) :: Thu, 10/19/2006 - 10:20am

Points allowed isn't very useful at all... it completely ignores how good the offences that the team has played are. That's a rather large problem with context, because, while over the course of the season that should even out, we're only in week 7 right now.

As for the steelers, are they rated too high? Possibly. It might be early-season blowout-itis. But the numbers do reflect something we know: a top QB has not been playing well for them should be returning to form, and when he does, that team is one of the best in the league.

Regarding defence DVOA, since it's done on a play-by-play basis, a consistent team that gets a lot of 3 to, say, 6-and-outs is going to be rated more highly than one that gives up a lot of medium-length drives, and a few 80-yard drives before recovering fumbles.

181
by Wanker79 (not verified) :: Thu, 10/19/2006 - 10:47am

Rick, I'm so damned tired of trying to explain to people how incredibly stupid it is to judge a team/unit by points scored/allowed. So I'll just leave it to Pat. He still seems like he has some fight left in him. I'm spent.

182
by Wanker79 (not verified) :: Thu, 10/19/2006 - 10:48am

...or Fnor. ;-)

183
by JHA (not verified) :: Thu, 10/19/2006 - 11:10am

Rick:

If the Patriots points scored is better than their DVOA ranking, then it means that DVOA is predicting that the Pats scoring results are outperforming their play by play analysis. And when Pittsburgh's DVOA ranking is higher than it's win/loss record would seem to show, then that means DVOA is predicting that Pittsburgh is underperforming relative to the play-by-play analysis. It is really that simple. Having watched Pittsburgh this season, that is an accurate assessment. My guess is that DVOA is showing that the Pats performance on play by play basis (2nd tier) is not as impressive as the W/L(top-tier).

I'm pretty sure this is a summary of what others have said here.

And NE is a quality team, that knows how to win, even when they don't play their best game. That's a quality that DVOA will not measure.

Of course at playoff time, or against the highest quality opposition, you have to win the battles on a play-by-play basis to win.

Cheeers!

184
by Horrorshow (not verified) :: Thu, 10/19/2006 - 12:00pm

I've been thinking lately how I'd like to see the "R" in DPAR actually defined by a real player who spent his entire career at or around 0.0 in terms of his DPAR score.

In recent years someone like Rod Gardner might be a decent example.

185
by Sean (not verified) :: Thu, 10/19/2006 - 12:54pm

Re 36, 40: If you are interested in who is more likely to win on Sunday, then the Dave predictions have been the way to go. DAVE has compiled a 50-20 record at picking the winning team, while the strictly 2006 DVOA has gone 46-24 over that same period.

186
by bravehoptoad (not verified) :: Thu, 10/19/2006 - 2:28pm

I like the zlionsfan template. I never get tired of reading those comments. They've turned into their own form, like sonnets or haiku or limericks. I suppose there are people who say, "God, I'm so sick of limericks! I get the point already!" But then there are those of us who appreciate good limericks versus bad limericks.

Go go, zlionsfan template.

187
by B (not verified) :: Thu, 10/19/2006 - 2:31pm

Well, maybe we should give trolls the option of using the zlionsfan template or writing in complaints in haiku or limerick format.

188
by Sid (not verified) :: Thu, 10/19/2006 - 2:47pm

I was looking at the punting numbers, and they're very skewed. Way more negative totals (and more extreme ones) than positive.

189
by Fnor (not verified) :: Thu, 10/19/2006 - 2:47pm

Way to quickly troll/
Unappreciated art/
zlionsfan's form.

190
by Led (not verified) :: Thu, 10/19/2006 - 3:02pm

My team ranked too low
Why? Just because I said so.
DVOA sucks!

191
by DavidH (not verified) :: Thu, 10/19/2006 - 4:01pm

Dear God,

PLEASE do not let this "haiku complaint" fad catch on. At least the zlionsfan template gives some creative prompts. Badku just makes people think that anything they can say with exactly 17 syllables is worth posting.

Also, it would have been REALLY nice for my survivor pool if you'd have let the Cardinals win on Monday. Thanks a LOT, big guy.

Amen.

192
by Independent George (not verified) :: Thu, 10/19/2006 - 5:13pm

The thing about haiku is that the 5:7:5 syllable structure only works in Japanese; supposedly, the English equivalent should be 3:4:3. Personally, I prefer the Hemingway 6-word story (when asked if he could write a story using only six words, Hemingway came up with, "Baby shoes - for sale. Never used.").

So, in 3:4:3 haiku form:

What's wrong with
DVOA?
My team sucks.

Here's my attempt at the six-word troll (assuming that DVOA or FO count as only 1 word each):

Screw DVOA - it sucks. You homo..

I thought about composing a sonnet, but those things are bloody friggin' difficult.

193
by Fnor (not verified) :: Thu, 10/19/2006 - 5:34pm

#192: Just out of interest, where did you get the 5-7-5 conversion to 3-4-3 from? Considering traditional haiku contained fragmented and truncated morphemes as an artistic style, I'd be wary of any type of conversion formula to proper english.

We could always do waka instead!

194
by Pat (not verified) :: Thu, 10/19/2006 - 6:59pm

Personally, I prefer the Hemingway 6-word story

I have no idea why I'm writing this, but...

Someone actually challenged me to write the shortest story possible, akin to Hemingway's challenge, but stronger. I said I could do it in two words - "Nothing happened."

195
by Jim (not verified) :: Thu, 10/19/2006 - 8:58pm

194 - nothing happened??? Life and all existence is about one thing - you can tell a story in one word:

Change

196
by Pat (not verified) :: Thu, 10/19/2006 - 9:19pm

195: Sorry, the main idea (as put to me) required a full sentence. Plus saying "Something happened." doesn't have a conclusion, since it leaves the question of what that something is. But that's enough random literary talk for today. :)

197
by Will Allen (not verified) :: Thu, 10/19/2006 - 10:06pm

A more accurate appaisal of the human condition requires at least three words...

"What the hell?"

...or perhaps one can do it in two...

"Not again!"

...but maybe three are required...

"I'll be damned!"

...it is all dependent on whether one views history as one damned thing after another, or as the same damned thing over and over. Personally, I'm agnostic on the matter.

198
by Terrell Owens (not verified) :: Thu, 10/19/2006 - 11:33pm

I can do it in one word(repeated many times)
memememememememememememememememdememememememememem

199
by Jay B. (not verified) :: Fri, 10/20/2006 - 1:34am

Will, I think you can sum it up in one word:

F***!

200
by DavidH (not verified) :: Fri, 10/20/2006 - 1:39am

Damn it, another thread taken over by discussions of poetry, Hemingway, and the human condition.

But I guess every football site must have this problem.

201
by Fnor (not verified) :: Fri, 10/20/2006 - 8:48am

Sorry, I'll try to make it more mainstream

I want a Hemi/
Though I work in an office/
Silly image thing.

202
by Jim (not verified) :: Fri, 10/20/2006 - 8:53am

Bill James the baseball stats guy summed up player contributions in a stat called win shares. I think these stats would be more easily digested if the +- numbers were converted to how much of a win does such a value contribute to.

203
by Independent George (not verified) :: Fri, 10/20/2006 - 11:16am

The NFL, and the human condition, described in six words or less by two of the greatest philosophers of our times:

Playoffs? Who said anything about playoffs?!!!

Oops. I did it again.

204
by DavidH (not verified) :: Fri, 10/20/2006 - 1:12pm

185-
Were you including any kind of home field advantage in those numbers? And if so, how much?

205
by Wanker79 (not verified) :: Fri, 10/20/2006 - 1:15pm

You play...to win...the game.

206
by Rich Conley (not verified) :: Fri, 10/20/2006 - 1:56pm

205.

That is exactly my point. Teams play to win the game, not to have high DVOA ratings. DVOA doesnt see any difference between a team running on 1st and 10, playing against a normal defense, vs the cardinals situations, where its 1st and 10, and theyre running against 9 men in the box. Thats a playcalling issue, not a skill difference, and DVOA sees them the same.

There are certain HUGE areas where DVOA could have big improvements, but unfortunately, most of them arent really possible because the information available is so sparce. The Play-by-play can't tell you everything. Its the reason DVOA on recievers isnt that great: it can't tell the difference between a ball that Mike Vick bounces off Alge Crumpler's chest, and a ball that Vick launches 15 feet over Roddie White's head. It can't tell the difference between a Drew Bledsoe throw directly at a cornerback, and a play where the ball goes off of TO's hands and gets caught on the ricochet. Only watching the game can tell you that.

DVOA has huge holes that the game charting project will help eventually, but right now, its a tool, and should be used as such.

DVOA can't tell the difference between a defence thats getting burned (up 10 points in the 4th quarter) and one thats playing Prevent, and succesfully running out the clock.

DVOA tells you how well a team is moving the ball (and how well its preventing the other team from moving the ball). The problem is, thats not always a top priority for teams. How many times have you seen a team run up the middle against 8 in the box just to run down the clock. Killing time is the priority there, not yards, not first downs. How many times have we seen a team up/down 15 points trying to work out something that hasnt been working well that season?

To blindly accept DVOA as the absolute truth of how good a team is, is to shit in the face of everything this site stands for: questioning the commonly percieved notions of why teams are good, asking why things are as they are, and looking at what is actually happening.

207
by Wanker79 (not verified) :: Fri, 10/20/2006 - 2:45pm

Re: 206

Um...wow. I was just contributing to the "six word story" tangent.

But anyway, you've been saying the exact same thing for the past week and every time someone has to remind you that DVOA does in fact take into account the game situation:

Every single play run in the NFL gets a "success value" based on this system, and then that number gets compared to the average success values of plays in similar situations for all players, adjusted for a number of variables. These include down and distance, field location, time remaining in game, and current scoring lead or deficit.

So that answers your incesant gripe about running out the clock. And as to you other point about play-calling and skill being two completely isolated qualities, you can't separate the two. They are not mutually exclusive. DVOA penalizes a team/player because the play-calling put that team/player in a piss poor position BECAUSE IT'S THE RIGHT THING TO DO. Does it really matter how much friggin talent a team has if the play-calling is shitty?!? And does it really matter how tremendous the play-calling is if there isn't enough talent on the field?!? Arizona's rushing unit should take a hit because their OC is/was a moron.

208
by Wanker79 (not verified) :: Fri, 10/20/2006 - 2:48pm

I'm not suggesting that healthy conversation about possible improvements to the system or possible blind-spots in the system are bad. On the contrary, they are some of my favorite discussions around here. But to keep saying the same thing over and over again and to keep receiving the same explaination over and over again just isn't productive.

209
by billvv (not verified) :: Fri, 10/20/2006 - 3:43pm

Success value has to consider the points scored. Maybe if you multiplied the value by points scored, so zero would zero the success value, you'd closer to what matters most, scoring.

210
by Rich Conley (not verified) :: Fri, 10/20/2006 - 4:02pm

Wanker, my point is that what is success in a team's eyes, isnt the same as what is success in DVOA's perspective.

"So that answers your incesant gripe about running out the clock. And as to you other point about play-calling and skill being two completely isolated qualities, you can’t separate the two."

I completely disagree. Arizona wasnt TRYING to get yards there, they were trying to run out the clock. The players were doing exactly what the team was trying to do. They lost that game because they fumbled twice, and had a kick run back, not because Edge couldnt get any yards. They weren't trying to get yards with Edge, they were trying to get 45 seconds.

Whether you think Denny Green was making the right decision, his decision was that they were essentially going to just chew clock, and they were doing that pretty succesfully, until the fumbles.

You may not like the gameplan, but that doesnt mean that the team wasnt successful at what it was trying to do.

The fumbles/ returned kick are what lost that game, not running edge into the line.

DVOA uses league average for things. The problem is, league average doesnt say shit about playcalling decisions. League average, teams winning, start running the clock out with X minutes left in the game. That doesnt mean that running the clock out with X+5 minutes left is a bad idea, but DVOA punishes teams for it, because your average team is still looking for first downs.

Maybe thats why the Patriots O rates so high via DVOA? They routinely march down the field at the end of the game, when they hold a big lead. They'll take the ball with 6:00 left, and march it down the field, and then kneel at the 10. Most teams would run the ball into the line as many times as they could, and punt away with

211
by Rich Conley (not verified) :: Fri, 10/20/2006 - 4:10pm

And heres my issues with not being able to seperate team intent, and team performance:

Team intent can change rapidly, team talent can not.

Look at the Chargers. Their game at Baltimore, they were trying to basically run out the clock, and it cost them the game. That doesnt mean their running game isnt good. Marty may have learned something there: maybe running up the score was a better idea than running out the clock. Two weeks later, they hang 40 on SF. Theres a huge change in strategy/philosophy, in a short time, with no change in talent.

My point is, without looking at what a team is trying to do, then DVOA is much less predictive than it could be.

212
by Wanker79 (not verified) :: Fri, 10/20/2006 - 4:29pm

You know what, you're right. Without the ability to read the minds of the coaching staff, DVOA is much less predictive than it could be.

213
by Pat (not verified) :: Fri, 10/20/2006 - 4:57pm

That doesnt mean that running the clock out with X+5 minutes left is a bad idea, but DVOA punishes teams for it, because your average team is still looking for first downs.

DVOA takes into account the score and time. Quoth I:

These include down and distance, field location, time remaining in game, and current scoring lead or deficit.

Plus, teams who are trying to run out the clock would do a hell of a lot better job if they actually got a new set of downs. I don't see the problem in saying a team is below average if they go 3-and-out 5 times trying to bleed out the clock (hello, Arizona) if an average team against that defense would do better.

214
by Rich Conley (not verified) :: Fri, 10/20/2006 - 5:10pm

"DVOA takes into account the score and time"

Exactly. It does. It says basically, if you're at 15 minutes left in the game, and up by 10, the average team gets 5 yards on this play. You only got 4, so you get a VOA of -20% for that play.

The problem is, the average team there is still trying to score. A team with a great defence may not be. Yes, they got less yards than the average team in that situation, but that doesnt mean that their offense is worse than the average team.

Yes, they got 20% less yards than the average there, but thats because the average team is still trying to score there, this team isnt.

In the opposite situation, lets say 2 minutes left in the game, your team is up by 10. The average play is a kneel, in that situation, for 1 yard. You throw a bomb down the field, and get 60 yards. You get big DVOA/DPAR boost, but that play really didnt matter, and doesnt tell you anything about how good a team is.

Again, my point is this: DVOA is a good tool. Its not even close to perfect though. It doesnt tell you how good a team is, it tells you how effective against average a team has been so far. Those two things are related, but not the same thing.

215
by Pat (not verified) :: Fri, 10/20/2006 - 5:51pm

The problem is, the average team there is still trying to score. A team with a great defence may not be.

5 yards versus 4 yards isn't a 20% difference in VOA if it's, say, on 3rd and 4. The important thing in an offensive drive is whether or not you're helping to continue the drive.

And running out the clock and scoring both require continuing the drive. It's not going to be that big a difference.

For instance:

Most teams would run the ball into the line as many times as they could, and punt away with

But that's all the Patriots are doing as well. They're just doing it better than other teams - good enough to earn first downs and continue doing it.

A team that has the lead, runs into the line three times and punts, is not a good offense. Regardless of how good their defense is, that's not a good offense. And it's not a particularly safe plan, either, as the Cardinals found out.

216
by Rich Conley (not verified) :: Fri, 10/20/2006 - 6:10pm

Pat, I agree.

But thats the thing, the Cardinals werent succesfull because their plan wasnt sound, not because they didnt execute the plan. They knew they were getting 1 yd on most of those runs.

The thing is, game plans can change. Saying the Cardinals lost because they run poorly, says they'll continue to lose because they run poorly, when thats not necessarily the case. Thats what I'm saying: DVOA can only say which team played better, not which team IS, or WILL BE better.

As to running the ball into the line, the pats dont do that at the end of the game when theyre up: They pass more than they run.

217
by Bencoder (not verified) :: Fri, 10/20/2006 - 10:44pm

'.....Am I the only one here that really dislikes DAVE?'

I'd probably feel better about the guy if he would bring my damn lawnmower back.....

218
by DGL (not verified) :: Fri, 10/20/2006 - 11:03pm

Personally I'd go with iambic pentameter rhyming couplets for the more literate troll format. The format forces a certain discipline of thought.

A colleague of mine and I once proposed, in the midst of one of the periodic "process over project" mindset swings in our company, that all requirements specifications be written in iambic pentameter, for that very reason.

And hey, "DVOA" is two iambs right there.

Let's see:

"The Redskins offense rated sixth? You're nuts/Mark Brunell sucks, you mindless geekish putz."

Hmm.

"The Chargers surely cannot be in first/The FOMBC must be reversed."

Hmm. Maybe it requires more discipline of thought than I am capable of bringing to bear at 10 PM on a Friday.

219
by Diane (not verified) :: Sat, 10/21/2006 - 12:05pm

Birth. School. Football. Death.

220
by Mike (not verified) :: Sun, 10/22/2006 - 11:58pm

Rough day for the DVOA top 10!

Currently sitting 0-4 and will finish the week at 1-5 once Dal/NYG play.

221
by Dan (not verified) :: Mon, 10/23/2006 - 2:40am

Lucky teams send us unfortunate trolls.

222
by Dan (not verified) :: Mon, 10/23/2006 - 2:56am

zlionsfan template is ranked too high
because it lacks meter, rhythm, and rhyme.
Fine poetry is way better than this.
Rexy is the 2nd coming of Jesus,
who's retirement from Green Bay is nigh.

223
by Wanker79 (not verified) :: Tue, 10/24/2006 - 10:20am

Wow, Dan. The this/Jesus rhyme was impressive.