Week 7 DVOA Ratings

Week 7 DVOA Ratings
Week 7 DVOA Ratings
Photo: USA Today Sports Images

by Aaron Schatz

The year of extreme parity continues. The Giants take the number one spot from Pittsburgh this week, by a very small margin. New England is third. All three teams had close victories this week, so none of them really improve upon their overall DVOA ratings. (Yes, it is hard to imagine that the Giants' win over Dallas was close, but that's how it turned out when Big Blue let the Cowboys back into the game in the fourth quarter.) As a result, no team is above 30.0% in total DVOA for the first time since the last week of 2006, when San Diego finished the year on top at 29.5%. If we want to set the bar a little lower, we get this amazing statistic: This is only the fourth week of the DVOA Era, out of every single regular-season week going back to 1993 -- that's 296 total weeks -- where no team is above 28.0% in total DVOA. The others:

The parity is not quite as equal at the bottom of the league, where there have been plenty of weeks where no team scored below Arizona's current rating of -37.6%. However, the Cardinals are odd because they've managed to actually win three games despite playing so poorly overall. Arizona is only the third team with at least three wins and DVOA below -25%. The others:

  • 3-2 Detroit in Weeks 5-6 of 2007 (Lions had a bye in Week 6).
  • 6-6 St. Louis in Week 13 of 2004.

While we are on historical oddities, San Diego is the first team of the DVOA Era with a special teams rating below -15% any time after Week 4. The record for worst special teams belongs to the 2000 Buffalo Bills at -12.9%. The 1997 Seahawks (-9.8%) are the only other team to finish a season below -8.5%. So yes, the Chargers are obscenely, historically bad on special teams.

Speaking of obscenely, historically bad, I said last night on Twitter that I would try to answer a couple of questions in today's commentary, so let's get to it.

@matthewglidden: Do you track all-time worst DVOA on third down? If so, how do 2010 Bears compare?

The Bears currently have a horrific -117.1% DVOA on third and fourth downs, and if the season ended today, that would indeed be the worst figure we've ever tracked. However, the season doesn't end today, and this is the kind of split where you are going to get both high and low extremes early in the season. In fact, Arizona (-78.3%) would also break the record for worst third-down offensive DVOA if the season ended today. These numbers are going to regress to the mean a bit with more playing time, so we'll have to see if Chicago will in fact finish as the all-time worst offense on third downs, breaking the record set by... themselves.

Worst Offensive DVOA on Third/Fourth Down, 1993-2009
Team Year DVOA
3rd/4th Down
DVOA
All Downs
CHI 2004 -68.8% -36.1%
HOU 2002 -65.5% -41.4%
SF 2005 -61.3% -42.0%
SEA 1998 -58.1% -14.2%
BUF 1997 -55.5% -22.0%
BAL 1999 -55.5% -19.5%
IND 1993 -55.3% -26.1%
CAR 2002 -53.4% -23.2%
OAK 2008 -51.2% -21.3%
OAK 1998 -50.3% -28.2%

@sfckoski: Does DVOA validate the common knowledge that the NFC West is the worst division? What is the worst division of the DVOA era?

Yes, the NFC West once again looks like the worst division in football based on DVOA ratings. The four teams have an average DVOA of -15.8% after Week 7. The NFC South is the next-worst division with an average DVOA of -7.8%. The NFC East is the best so far this year, with an average DVOA of 11.9%.

The NFC West has been the worst division in football for years and years now, which is sort of odd -- you figure this kind of thing should be cyclical and move around from division to division as different teams went up and down the development cycle, but the four NFC West teams have consistently awful records over the past few years. There were a couple of years there where Seattle was pretty good, but San Francisco made up for it by being extra-pathetic. Since the NFL went to eight divisions in 2002, the six worst divisions (based on average DVOA of the four teams) have all been the NFC West. The four teams had an average DVOA below -20% in both 2004 (-23.5%) and 2008 (-21.6%). The seventh-worst division -- probably the eighth-worst after this year is over -- was the 2002 NFC North, with an average DVOA of -14.1%.

From 1993-2001, the worst division by average DVOA was the 1998 NFC East, with an average DVOA of -12.6%. Dallas won the division and ranked 11th in DVOA. Arizona actually snagged a wild card with a 9-7 record despite ranking 27th out of 30 teams in DVOA. The Giants ranked 17th, Washington ranked 28th, and Philadelphia was dead last. Unlike the current NFC West, however, these teams didn't struggle for a whole span of years -- the average DVOA in the division was much closer to zero in 1997 and 1999. The other pre-2002 division with an average DVOA below -10% was the 1999 NFC West, where the champion Rams were outweighed by teams ranked 20th, 26th, 28th, and 30th.

If we're going to do the worst divisions, we should also look at the best divisions. Post-realignment, the best division was the 2004 AFC East, with an average DVOA of 20.3%. The Patriots finished second in DVOA and won the Super Bowl, while the Bills finished third but didn't even make the playoffs when they choked against the Steelers backups in Week 17. (I don't normally like to use the term choked, but I can't think of another description for blowing a win-and-you're-in game at home against a team resting its starters.) The 2008 NFC East is second, with an average of 18.8%, and this is currently the sixth straight year where the NFC East teams have an average DVOA above 10%, although the NFC East hasn't been quite as good as the NFC West has been bad.

Prior to realignment, the best division was the 1998 AFC East, with an average DVOA of 12.4%. The Jets, Dolphins, and Bills were all in the top six with the Patriots 11th. The Colts sort of ruined things by finishing 26th, although that wouldn't be a problem in future seasons, once their rookie quarterback got a hang of the NFL game. The 1993 AFC West, with an average DVOA of 11.8%, is the second-strongest division from the 1993-2001 period.

Returning to 2010... Stats pages, playoff odds, and DVOA premium database should all be updated through Week 7. Enjoy the ratings.

* * * * *

These are the Football Outsiders team efficiency ratings through seven weeks of 2010, 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 to consider all fumbles, kept or lost, as equal value. SPECIAL TEAMS DVOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season.

Opponent adjustments are currently at 70 percent strength and will steadily grow stronger until Week 10. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.

DAVE is a formula which combines our preseason projection with current DVOA to get a more accurate forecast of how a team will play the rest of the season. Right now, the preseason projection makes up 8.5 percent of DAVE for teams that have played seven games and 19 percent of DAVE for teams that have played six games. In addition, DAVE slightly discounts the results of the first three weeks of the season, using the beginning stages of what will eventually become WEIGHTED DVOA.

To save people some time, please use the following format 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
TOTAL
DAVE
RANK W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
1 NYG 27.9% 2 26.9% 2 5-2 12.2% 11 -22.9% 1 -7.1% 31
2 PIT 27.8% 1 27.4% 1 5-1 7.0% 15 -17.8% 3 3.0% 12
3 NE 26.2% 3 25.5% 3 5-1 29.0% 2 9.9% 27 7.1% 5
4 KC 26.2% 7 20.8% 6 4-2 17.6% 5 -4.4% 9 4.1% 9
5 TEN 25.9% 4 23.4% 5 5-2 0.7% 18 -20.9% 2 4.3% 8
6 PHI 25.0% 6 24.1% 4 4-3 23.9% 4 -3.9% 10 -2.7% 25
7 NYJ 20.5% 5 18.6% 7 5-1 14.8% 9 0.5% 15 6.3% 6
8 SD 13.9% 13 12.6% 10 2-5 15.6% 7 -15.6% 4 -17.4% 32
9 ATL 12.5% 10 12.8% 9 5-2 15.0% 8 3.0% 18 0.5% 15
10 IND 12.3% 8 13.6% 8 4-2 25.6% 3 7.8% 23 -5.6% 29
11 GB 11.5% 12 11.5% 11 4-3 16.8% 6 -1.1% 13 -6.4% 30
12 BAL 8.4% 11 10.6% 12 5-2 7.2% 14 2.3% 16 3.5% 11
13 MIA 5.9% 16 7.3% 13 3-3 8.0% 12 -1.1% 12 -3.3% 26
14 NO 4.6% 9 4.9% 14 4-3 7.5% 13 0.3% 14 -2.7% 24
15 HOU 3.7% 15 -0.3% 18 4-2 32.0% 1 26.6% 30 -1.7% 22
16 SEA 2.1% 14 0.1% 16 4-2 -9.8% 25 -2.0% 11 9.9% 1
TEAM TOTAL
DVOA
LAST
WEEK
TOTAL
DAVE
RANK W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
17 DAL 2.1% 17 1.9% 15 1-5 13.9% 10 11.7% 28 -0.1% 16
18 MIN -3.4% 21 0.0% 17 2-4 -13.0% 28 -7.0% 7 2.6% 13
19 CIN -4.0% 18 -3.8% 19 2-4 4.2% 16 6.7% 22 -1.5% 21
20 CLE -4.1% 25 -5.5% 21 2-5 -4.5% 22 3.3% 20 3.6% 10
21 DET -4.5% 20 -9.6% 24 1-5 -3.6% 21 8.2% 24 7.3% 4
22 SF -7.1% 23 -7.4% 22 1-6 -11.3% 27 -5.2% 8 -1.0% 19
23 WAS -7.3% 19 -5.4% 20 4-3 -1.6% 19 3.8% 21 -1.9% 23
24 CHI -11.2% 24 -9.4% 23 4-3 -28.9% 30 -9.9% 5 7.8% 2
25 TB -14.9% 26 -14.6% 25 4-2 -6.0% 24 8.5% 25 -0.5% 18
26 OAK -19.8% 30 -20.1% 26 3-4 -10.7% 26 8.8% 26 -0.2% 17
27 STL -20.8% 28 -20.8% 27 3-4 -14.1% 29 3.2% 19 -3.5% 27
28 JAC -25.5% 27 -24.1% 28 3-4 -5.6% 23 27.6% 31 7.6% 3
29 DEN -25.9% 22 -24.5% 29 2-5 2.0% 17 24.0% 29 -3.9% 28
30 BUF -28.4% 29 -25.0% 30 0-6 -1.7% 20 31.3% 32 4.6% 7
31 CAR -33.3% 31 -28.2% 31 1-5 -39.6% 32 -7.6% 6 -1.4% 20
32 ARI -37.6% 32 -31.6% 32 3-3 -35.7% 31 2.5% 17 0.6% 14
  • 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 most consistent (#1, lowest variance) to least consistent (#32, highest variance).



TEAM TOTAL
DVOA
W-L NON-ADJ
TOT VOA
ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VAR. RANK
1 NYG 27.9% 5-2 25.9% 5.1 1 -0.7% 20 2.5% 14 22.2% 26
2 PIT 27.8% 5-1 28.9% 4.9 6 5.6% 8 -3.4% 25 10.9% 15
3 NE 26.2% 5-1 24.8% 4.9 5 2.7% 12 2.6% 13 9.5% 9
4 KC 26.2% 4-2 23.9% 5.1 3 -1.1% 23 -13.6% 31 11.5% 16
5 TEN 25.9% 5-2 27.0% 5.1 2 1.7% 15 5.0% 10 10.8% 14
6 PHI 25.0% 4-3 29.5% 5.0 4 0.8% 16 6.0% 7 6.9% 6
7 NYJ 20.5% 5-1 32.0% 4.7 7 -2.9% 25 2.3% 15 6.9% 5
8 SD 13.9% 2-5 15.3% 4.1 13 -7.0% 29 -1.6% 23 19.4% 23
9 ATL 12.5% 5-2 8.6% 4.3 11 0.7% 17 -10.1% 30 12.7% 17
10 IND 12.3% 4-2 14.0% 4.3 10 -0.1% 18 7.3% 3 14.2% 18
11 GB 11.5% 4-3 16.0% 4.5 8 -3.4% 26 7.0% 4 8.9% 8
12 BAL 8.4% 5-2 7.7% 4.4 9 1.7% 14 -0.2% 19 4.1% 2
13 MIA 5.9% 3-3 3.1% 4.2 12 9.0% 1 0.9% 16 15.1% 19
14 NO 4.6% 4-3 7.7% 3.9 14 -12.6% 31 -2.2% 24 9.7% 11
15 HOU 3.7% 4-2 1.1% 3.5 16 6.9% 5 5.5% 8 23.5% 27
16 SEA 2.1% 4-2 7.1% 3.8 15 -14.8% 32 -6.2% 27 18.1% 22
TEAM TOTAL
DVOA
W-L NON-ADJ
TOT VOA
ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VAR. RANK
17 DAL 2.1% 1-5 -5.8% 3.4 18 5.9% 7 3.1% 12 15.8% 21
18 MIN -3.4% 2-4 -10.3% 3.3 19 6.7% 6 -1.0% 22 3.7% 1
19 CIN -4.0% 2-4 -0.7% 3.0 24 -0.9% 22 8.9% 1 6.3% 4
20 CLE -4.1% 2-5 -4.4% 3.3 21 8.7% 2 -0.3% 20 7.8% 7
21 DET -4.5% 1-5 -6.4% 3.0 23 4.8% 9 0.1% 17 9.7% 12
22 SF -7.1% 1-6 -9.3% 3.4 17 2.5% 13 -14.5% 32 27.8% 30
23 WAS -7.3% 4-3 -3.3% 3.2 22 3.2% 11 6.7% 6 4.5% 3
24 CHI -11.2% 4-3 -10.7% 2.6 25 -0.2% 19 5.5% 9 23.8% 28
25 TB -14.9% 4-2 -9.9% 3.3 20 -5.0% 27 -5.0% 26 20.4% 24
26 OAK -19.8% 3-4 -7.2% 2.2 28 -6.8% 28 7.0% 5 33.2% 32
27 STL -20.8% 3-4 -14.3% 2.4 27 -9.7% 30 -7.3% 28 15.4% 20
28 JAC -25.5% 3-4 -29.9% 2.5 26 7.0% 4 4.9% 11 20.7% 25
29 DEN -25.9% 2-5 -22.3% 1.8 30 3.4% 10 -0.2% 18 24.6% 29
30 BUF -28.4% 0-6 -27.0% 1.6 32 7.8% 3 7.9% 2 9.5% 10
31 CAR -33.3% 1-5 -34.3% 1.6 31 -0.8% 21 -1.0% 21 10.2% 13
32 ARI -37.6% 3-3 -36.1% 1.9 29 -1.2% 24 -8.2% 29 32.7% 31

Comments

185 comments, Last at 30 Oct 2010, 3:18am

1 Re: Week 7 DVOA Ratings

is DVOA drunk? No way canChiefs be 4 and Raiders 26.

Doesn't matetr anyway, becaus Raiders goign to scalp Chiefs soon enough and show who real boss in AFC West

19 Re: Week 7 DVOA Ratings

All I care about is that the Chiefs - Raiders rivalry resumes living up to the quality I became used to in the 60's and 70's. I'll be happy with a split as long as KC and Oakland finish 1 - 2 and Denver and San Diego return to the 3 - 4 slots like the good old days. Looking forward to November 7th!

2 Re: Week 7 DVOA Ratings

Dallas is clearly ranked too high because they have no fire and playing behind pathetic offense line Tony Romo could only manage to "finish" his season sixth in DYAR and third in DVOA (and one in rushing DVOA), and also this is the worst season ever.

4 Re: Week 7 DVOA Ratings

As an Eagles fan, my main question into regards to Dallas is, "Will the players give up on the season?" I'd love for Dallas, which is actually playing far above their record, to just throw in the towel and stop trying. Those games looked like the big ones on the Eagles last season schedule, but I'd love for the Eagles to actually be going up against a bunch of guys who genuinely don't care. And my initial reaction is that this doesn't seem like a "Play for Pride" bunch of fellows...

8 Re: Week 7 DVOA Ratings

As a Cowboys fan who cannot stand Wade, I'd love for them to end up with a top five draft pick and a new regime. We got Aikman rather than Mandarich because Green Bay won a meaningless week 16 game trying to fight rather than play out the string. This team has a lot of areas of strength, but it isn't their year. Might as well make the best of it

152 Re: Week 7 DVOA Ratings

There will definitely be a draft next year (unless a new CBA abolishing the draft is signed before then), as it's the last thing covered by the current CBA. It's after the draft that things get uncertain.

- Alvaro

58 Re: Week 7 DVOA Ratings

I think the offense will definitely keep playing. And on defense, we know Igor Olshansky will keep trying to tackle RBs for no gain so he can celebrate regardless of score or record. Spears and Bowen are in contract years (Hatcher, too, but he's hurt). Ware and Ratliff will always give their all. And there will be a lot of young guys getting playing time looking to prove themselves. Still the Eagles should be able to beat them pretty easily.

3 Re: Week 7 DVOA Ratings

The Giants' defense being #1 is a lock if they continue to injure opposing quartebacks

9 Re: Week 7 DVOA Ratings

Well both Philly guys look injury prone behind that line, so you never know. Depends on when they catch them, anyway, as Vick doesn't look like he'll last no matter the competition.

6 Re: Week 7 DVOA Ratings

How naive us Bear fans were to think that we couldn't possibly get any worse on offense than under Ron Turner. It seems remarkable that this team will quite possibly break its own record for offensive futility - with this offense, at least on paper, being far more talented than the Jonothan Quinn, Craig Krenzel, Chad Hutchinson monstrosity of 2004. In my opinion it comes down to arrogance -

The arrogance of Jerry Angelo thinking that he could replicate the success he had with veteran FA offensive lineman in 2005/6 with Tait, Brown and Miller and subsequently neglecting the line in the draft

The arrogance of Mike Martz thinking that he can make his scheme work with the horrific talent on the offensive line, and then when its painfully obvious it isn't working, failing to adjust. Ironically, this was seen my many Bears fans as being Ron Turner's biggest fault, yet Martz has looked even more clueless than Turner in this respect (bar the Dallas game)

And last of all, the arrogance of Mike Tice thinking he can transform the non-NFL talent on the line into something that resembled a competent group. IMO he should've realised in camp that there was signifcant problems with the talent and urged Angelo to go and get a McNeil or Mankins, despite the high price they would've demanded. The amount of mental errors the line has made also should be blamed at least partly on Tice.

How sad that the best Bear defense since 2005/6 will go to waste, as will the brilliant special teams.

This leads me to another point. When does Dave Toub start to get chances to rise up the coaching ranks? There is precedent for special teams coaches to get head coaching jobs now with John Harbaugh in Baltimore, and Toub is clearly the best in the league. Even though he's had the luxury of having Devin Hester, he's still had outstanding results from Knox, Manning, Earl Bennett etc. If (when) Lovie is fired, I'd be all for Toub getting a look in as Head Coach

17 Re: Week 7 DVOA Ratings

Call me naive (or just stupid), but I think the offense is going to start turning things around. The current 5 guys they have actually block pretty well when they make contact. Continuity and experience will increase and they could look like an average unit in a few weeks.

I also have faith in Lovie Smith being able to stop the Bills offense. So I think they'll be 5-3 plus the bye week. Minnesota could easily give up on the season by then.

Things could turn around.

24 Re: Week 7 DVOA Ratings

I thought the same thing about the Seattle game. I can unfortunately see us being Buffalo's first win of the year.

Even if there is improvement on the O-Line, Cutler continues to be a worry. We all heard how Martz will be hard on Jay and improve his mechanics, yet they have regressed even further than last year and he continues to make these mistakes (throws of the backfoot primarily)

Things could turn around, I agree, maybe the Martz offense will start to click, but I remain sceptical :(

96 Re: Week 7 DVOA Ratings

Don't worry Bears fans, Chester Taylor is the new goal line back! All your worries have been washed away!

Disclaimer: I own Forte in two leagues and am more than annoyed to see him be a scapegoat for short yardage struggles. He doesn't have a prayer in Hell most of the time as the line is getting blown the F UP!

154 Re: Week 7 DVOA Ratings

I couldn't agree more. After the last two weeks, and the worst coaching call ever (not challenging the goal-line fumble) that directly cost them the game (and killed Cutler's confidence or yet another week. If it was the Skins playing from a 9-point deficit, the way they were putting the ball on the ground and in the DB's hands, instead of Cutler tryign to play catch-up, that game would have ended MUCH diferently), I wouldn't even take the Edwards-led Bills for granted. As it is, the Fitzpatrick-led Bills actually worry me...

- Alvaro

29 Re: Week 7 DVOA Ratings

"The current 5 guys they have actually block pretty well when they make contact."

You could say the same thing about a fire hydrant.

39 Re: Week 7 DVOA Ratings

"The arrogance of Mike Martz thinking that he can make his scheme work with the horrific talent on the offensive line, and then when its painfully obvious it isn't working, failing to adjust. Ironically, this was seen my many Bears fans as being Ron Turner's biggest fault, yet Martz has looked even more clueless than Turner in this respect (bar the Dallas game)

"

How exactly does one adjust to their offensive line not being able to block anyone?

42 Re: Week 7 DVOA Ratings

Shorter drops, as per the Dallas game and do away with the 7 step drops that Martz keeps trying to use. It isn't fullproof but its better than inviting the big blitz and leaving Cutler a sitting duck in the backfield as Omiyale and Webb watch DBs come through untouched

105 Re: Week 7 DVOA Ratings

Did you not watch the last game? They did do shorter drops and D Hall jumped all the routes cause he knew that's what they had to do.

170 Re: Week 7 DVOA Ratings

I'm still not sold the D'Angelo Hall jumped the routes because he knew that is what he had to do... He jumped those routes because he is D'Angelo Hall and that is what he does, usually he will get burned down-field with an offense that can protect their corner-back.

I think the under-rated thing is how much they avoided Carlos Rogers. If they had attacked him at least he would have dropped those interceptions

50 Re: Week 7 DVOA Ratings

In reply to by RichC (not verified)

Yeah, Martz has developed his QB-killer reputation because he simply won't abandon his offensive strategy; pass plays that take time to develop. If he simply changed his plays (or playcalling - maybe he simply doesn't 'do' 3-step drops and short outlet passes), then the Bears would be looking a lot better. And Cutler would be more likely to actually survive the whole season.

60 Re: Week 7 DVOA Ratings

Conventional wisdom is if Martz wasn't so arrogant and stuck on his system, his team would have won the 2001 Super Bowl.

55 Re: Week 7 DVOA Ratings

Technically, Harbaugh had been promoted to Eagles DB coach for at least one season before he was hired by Baltimore, so it's still unlikely that a coach would go directly from coaching special teams to the head coach position. The manager making the hiring decision would be taking a big risk to bring in somebody who's never coached either an offensive or defensive position - if both units decline, the manager would be blamed for the bad decision to hire such an "inexperienced" coach.

89 Re: Week 7 DVOA Ratings

Mike Martz's offenses failing to adjust is one of their hallmarks.

He'll keep doing what he wants to do. If it works, great. If it doesn't, he'll keep doing it until it does.

123 Re: Week 7 DVOA Ratings

I can directly contradict that generalization with how he reacted here in SF. His play-calling changed drastically as the year went on and his QBs changed.

111 Re: Week 7 DVOA Ratings

Living in Chicago, it always amazes me to hear how badly people want Lovie's head. He is a damned good head coach, and an astonishingly good defensive coach. My great fear is that Lovey gets fired and winds up in Dallas, where his professionalism turns a young & talented, but mentally weak, team and turns them into perennial contenders.

118 Re: Week 7 DVOA Ratings

He has some flaws as a head coach, but I'm one of those who would rather wait and see if he can learn from his mistakes and grow than throw him out and start over.

A lot of teams get praised for keeping a head coach and GM together for a long time as promoting continuity. When the Bears do it it's because they're cheap or something.

153 Re: Week 7 DVOA Ratings

To be fari, this particualr Bears fan called hiring Martz a horrible decission form the moment it was made. And although we were the minority I do seem to recall there were a few others who realized that Martz + Cutler + the Bears O-Line = unmitigated dissaster.

- Alvaro

7 Re: Week 7 DVOA Ratings

Wow. SD stands out like a sore thumb up there at 2-5. But in the AFC West, they're not out of it, are they?

13 Re: Week 7 DVOA Ratings

I'm curious if they had an even "normal" bad ST DVOA (like say -2.3%) if they would be the top ranked team in DVOA.

129 Re: Week 7 DVOA Ratings

That's not how it works; it's 3/7ths * off + 3/7ths * def + 1/7th * ST. They'd still be significantly higher than they are now, but it's not taking off so much of their score.

138 Re: Week 7 DVOA Ratings

The special-teams-is-only-33%-as-important-as-offense-or-defense factor is already accounted for in the numbers. A -17% ST DVOA is the equivalent of a -51% offensive DVOA!

134 Re: Week 7 DVOA Ratings

No, actually, the ST DVOA is scaled to be equal to offense/defense. That's why you virtually never see "-17.4%" ST DVOA. Seriously, WTF, San Diego. That -17.4% ST DVOA is the equivalent of them having a (-/+) 50% offense/defense DVOA. They're just not hurt so monstrously bad because you don't have that many special teams plays in a game. (You can tell this is the case by just doing off. dvoa - def. dvoa + st. dvoa)

I mean, you normally get some outliers early in the season. But San Diego's actually bad even by those standards - Minnesota was at -14.4% in Week 7 in 2007. San Diego's significantly past that.

San Diego's an interesting test case, though, because my gut feeling is that bad special teams hurts you, in terms of your ability to win, more so than bad offense or defense (that is, you'll lose more games than you might expect based on the decrease in points scored/increase in points allowed) because special teams plays tend to be viewed as "safe" options by coaches - plus, they're used to playing field position if their offense is struggling, they're used to playing aggressive if their defense is struggling, but I don't think coaches are used to changing their playcalling when their special teams are struggling.

141 Re: Week 7 DVOA Ratings

That's a really good point, coaches don't really game plan around special teams (except very specific things like FG range or not punting to Devin Hester).

This also explains why Belichick gets so much criticism when he does, since it seems so odd.

172 Re: Week 7 DVOA Ratings

Interesting theory, my DVOA database doesn't contain the variance data. FO could run a quick correlation between the two to see if there is a relationship. Using this week, the r = .15 and is ns. I know small sample size.

137 Re: Week 7 DVOA Ratings

It's trivially easy to check your statement by using the numbers at the top of this page, which would confirm that I'm right and you're wrong, nanny nanny boo-boo. :)

12 Great post

enjoyed the mailbag format.

15 Re: Week 7 DVOA Ratings

in the new Playoff Odds report, Pats have better chance to get bye, but worse chance to win division than the NYJ.any idea why?

25 Re: Week 7 DVOA Ratings

I believe that this is the case because the Jets beat the Pats. In the event of a two way tie, the Jets are more likely to win the division but somehow based on the simulations, if the Jets, Pats, and another team were in a multiple team tie, the Pats would be more likely to win, thus getting a bye.

The other possible explanation is that the Pats have a higher variance than the Jets meaning that their inconsistency could make them more prone to dominating and faltering.

44 Re: Week 7 DVOA Ratings

From what I figure, This has everything to do with the team they lost too.

Not only do the Jets have to win the division to get the Bye, they have to also have more wins then Baltimore (unless Baltimore is First and the Jets are 2nd in the AFC).

For instance Say Titans or KC or Indy wins their division with a 14-2 and gets the 1st spot in the playoffs, then Jets and Baltimore end with 13-3 record, in this scenario Jets miss the bye. However if 13-3 Baltimore and 13-3 NE are vying for spot NE gets the nod since NE won head to head.

Also factoring in who gets Byes is the NE/Pitt Game, and Jets/Pitt game. Because Jets have less Dave then NE they probably lose to Pitt in more scenarios then what NE loses to them in.

16 Re: Week 7 DVOA Ratings

Aaron, I was wondering if this weekends turnoverthon between the Bears and Redskins set any records. At the minimum it seems like the Redskins recovering 7 out of 8 fumbles has to be pretty unique

18 Re: Week 7 DVOA Ratings

Any Arizona fans out there? The Whiz seems to be showing his true worth now that he doesn't have Warner's arm and experience to cover his limited abilities.

34 Re: Week 7 DVOA Ratings

Not exactly a Cards fan, but I do live in Tucson, and root for them
to have a good team since we are force fed their games...

I don't think the Whiz is a bad coach overall. After all Shanahan lost
several IQ points when Elway retired, Landry lost some when Staubach
retired, etc. etc. The problem I have is how the QB situation has been
handled. If Leinart was so untrustworthy that he could be released
just before the season started, why wasn't a QB taken with a higher
(2nd or 3rd) round pick? Why wait so long to release Leinart?
IMHO Whizzy just didn't like the guy, and treated him accordingly.

FWIW

23 Re: Week 7 DVOA Ratings

The Colts run-DVOA is surprisingly good. 8th, for a team that usually is in the bottom half of the league. Hopefully this is a trend, as the running game needs to pick up with Dallas Clark gone.

65 Re: Week 7 DVOA Ratings

A lot of that is probably from the Giants game, when the current-#1-defensive-DVOA team decided that they should use Dime as their base defense and consequently got run over.

27 Re: Pats Defense

So, NE has the same defensive ranking as they had coming off the Buffalo game, but the actual DVOA value has improved from 20% to 9.9%. Sounds about right to me. If they keep getting a little better each week, I think they could be a top 10-12 unit by year's end.

36 Re: Pats Defense

In reply to by Anonymous1 (not verified)

Solid improvement by the defense but the offensive DVOA is in serious decline since the Moss trade. It's only two games so take it with a grain of salt until there's more data, but the Pats' offensive DVOA for the season has declined from 42% after week 5 to 35% after week 6 to 29% after week 7. Weeks 6 asnd 7 have been Brady's worst performances of the season by DYAR. (Not that the performances were "bad" but rather they were just ok...middling.) Time will tell whether this trend will continue or its just a random blip. I've always had an outsize view of Moss's effect on an offense, so I suspect the trend will continue.

41 Re: Pats Defense

In reply to by Led

While I agree that Moss is having an effect, they've also played 2 of the best defenses in the league in the last 2 weeks, and DVOA's opponent adjustments aren't even good, let alone perfect.

76 Re: Pats Defense

In reply to by RichC (not verified)

Could be. But Baltimore's defense has been average so far and they just got manhandled by the Bills. And I'm not sold on SD's defense. They've played the Pats and KC (whose offense I'm not sold on yet despite a good DVOA) and then 5 of the 9 worst offenses in the league. Time will tell.

28 Re: Week 7 DVOA Ratings

I tried to post this in last weeks DVOA ratings, but got to them rather late and assume it went unread (or at least I hope as it got no replies) so I'll repost it here:

The way that DVOA calculates SOS is clearly flawed in some ways. We can all imagine a fake scenario where you can play the middle teams in the league (9th-24th) or the top and bottom 1/4 of the league (1st-8th and 24th-32nd). However, you always have to consider what kind of team would want a certain schedule.
-If Team A is the 8th best team in the league they would want a schedule that was full of average teams. Those ranked 9th through 24th. That should lead to them winning the most games, with the idea that all of their opponents are worse than them.
However, Team B, which is the 23rd best team in the league would probably want to play the second schedule type as they get games against 8 teams that they are better than.
Team C which is in the best team in the league would want the schedule against the mediocre teams because they have a reasonably good chance of winning all 16 games, where if they had to play the top and bottom they could very easily lose some close games to a couple of the top teams.

The point of this long post is that I believe that the SOS should be modified to be more similar to what is using in Fremeau's (sp?) FEI index where, if memory serves me correct the SOS of schedule is based upon the chance that an elite team would go undefeated. In college teams going undefeated is fairly common (generally at least 1 or 2 a year). However, in the NFL this is not as common, but winning 12 or 13 games is kind of the equivalent. Perhaps SOS should be based a teams likelihood of winning some number of games (perhaps 11, 12, 13; an amount usually guaranteed to get you into the playoffs).

45 Re: Week 7 DVOA Ratings

I don't see why you would change the baseline. Yeah going undefeated is virtually unheard of in the NFL, but that doesn't mean it's not a good way to measure SoS.

47 Re: Week 7 DVOA Ratings

The way that DVOA calculates SOS is clearly flawed in some ways.

That depends on what you expect from DVOA's SOS ratings. If you want to know how good a team would have to be, on a play-by-play level, to post a VOA of 0.0% against a particular schedule, then the way DVOA calculates SOS is basically perfect. Now, you obviously want something else from SOS, but some of us find it more useful/interesting the way it is now. Calling it "clearly flawed" basically assumes that everyone wants exactly what you want from it. Not everyone does.

We can all imagine a fake scenario where you can play the middle teams in the league (9th-24th) or the top and bottom 1/4 of the league (1st-8th and 24th-32nd). However, you always have to consider what kind of team would want a certain schedule.
-If Team A is the 8th best team in the league they would want a schedule that was full of average teams. Those ranked 9th through 24th. That should lead to them winning the most games, with the idea that all of their opponents are worse than them.

Not necessarily. Just because they are favored in more of their games doesn't mean that they will have more wins at the end of the season. For instance, let's assume that they would win 60% of their games against the middle teams of the league, 45% against the top quarter, and 80% against the bottom quarter. Then, against the middle teams, they'd get 11+ wins about 33% of the time. Against the top and bottom quarter of the league, on the other hand, they'd get 11+ wins almost 40% of the time.

Now, the probabilities in that hypothetical might be off - maybe by enough to swing things in favor of a schedule of middling teams - but there's no reason to assume that a priori. The probabilities I used are entirely consistent with your description of them as the 8th best team in the league - they would still be favored in all of their games when playing a middling schedule, while only being favored in half of their games when playing the top and bottom quarters of the league. And yet they would win more games against the latter schedule.

The same argument applies equally well to the other examples you give.

Perhaps SOS should be based a teams likelihood of winning some number of games (perhaps 11, 12, 13; an amount usually guaranteed to get you into the playoffs).

This is the one part of your post that I agree with, at least partially. I would like to see what the probability of winning, say, 11+ games against each team's schedule would be for an elite team. That'd be pretty cool. Of course, I'd also want to see what the chances of an average team would be, or even a below average team, to see how likely it would be for a mediocre team to wind up with a playoff slot because of an easy schedule. But I'd also like them to keep the normal DVOA SOS ratings. There's no reason that it has to be one or the other. They each contain valuable information.

77 Re: Week 7 DVOA Ratings

I don't know if you or xtimmygx have read the FO Almanac (ore previous Pro Football Prospectus) before, but one of the features for each team is calculating a percentage chance of finishing with the following record ranges: 0-3, 4-6, 7-8, 9-10, 11+.

Of course this is all based on the preseason DVOA projections. During the season we don't have numbers for finishing with a particular record, but the playoff odds report basically considers all of the specific "DVOA matchup" possibilities into consideration. This doesn't yield a direct SOS like xtimmygx wants, but taking all the information together can paint a pretty good overall picture.

My personal opinion is that a SOS for calculating specific records and taking large variations in team quality into account is much more important for college football than the NFL. In the NFL the SOS between all teams is much closer and there is much less variance in team quality than there is across FBS.

81 Re: Week 7 DVOA Ratings

I don't know if you or xtimmygx have read the FO Almanac (ore previous Pro Football Prospectus) before, but one of the features for each team is calculating a percentage chance of finishing with the following record ranges: 0-3, 4-6, 7-8, 9-10, 11+.

I have read those, and really enjoyed that feature. It would be pretty awesome to see it updated throughout the season. I think that would solve everyone's problems.

My personal opinion is that a SOS for calculating specific records and taking large variations in team quality into account is much more important for college football than the NFL. In the NFL the SOS between all teams is much closer and there is much less variance in team quality than there is across FBS.

I agree, which is why I want the original DVOA SOS to stay.

80 Re: Week 7 DVOA Ratings

Yeah, it would be cool if they added another page to stats section that focuses solely on SOS, and measures it in a variety of ways. They could also show the average offensive and defensive DVOA of past/future opponents. Or even a schedule page for each team that lists their entire season, with opponent DVOA (total/off/def/st).

139 Re: Week 7 DVOA Ratings

"The way that DVOA calculates SOS is clearly flawed in some ways."

You can't take a random distribution of numbers and reduce them to one number. Duh. There's no one "good measure" of strength-of-schedule. It just depends on what you want it for.

You could measure strength-of-schedule by the number of wins added/lost, for a team of that strength, versus an average schedule. You could measure strength-of-schedule by the number of wins added/lost, by an average team versus that schedule, compared to an average schedule.

"The point of this long post is that I believe that the SOS should be modified to be more similar to what is using in Fremeau's (sp?) FEI index"

Yeah, that's because in college football you only really care about comparing the top few percent of the league. No one cares if the reason that Northwestern won 6 games was because they played really easy teams.

That's not really true in the NFL.

"Perhaps SOS should be based a teams likelihood of winning some number of games "

What team? An average team? That team? An elite team? They're all different questions. Again, in college, no one cares why a team won, say, 6 games instead of 8 (or vice versa). In the NFL, we do. We care why a top-ranked DVOA team won only 9 games instead of 12. We care why a nearly-bottom ranked team won 8 games instead of 5. We care why an average team won 10 games instead of 8.

There's no one number that could explain all of those things.

As an aside, part of the reason you don't want to ask "likelihood of winning some games" as a measure of something connected to DVOA is that DVOA is linked to scoring, not winning. What you're talking about is something linked to winning - so really, you're interested in something like an opponent-adjusted estimated wins.

173 Re: Week 7 DVOA Ratings

I think you are breaking down what I am saying to much. What I am saying is that I come on to this website, which I have been doing for at least 5 years now, and conversations of what the SOS number means are constant. I find the number provided to be only marginally enlightening, as do many others, hence the recurring discussion.

The issue here is that you can get the same value for SOS while playing vastly different schedules. If I told you that a team had a SOS value of 0%, you have gained very little information on the median quality of opponent, which is for all intents and purposes more useful to know. The years when the Colts and Pats had obscenely high DVOA there were SOS values that were high in the AFCE and AFCS due just to those two teams.

You say that in college no one cares why a team won a certain number of games but in the NFL we do. I agree, that is why I think it is important that we get a SOS measurement that accounts for this. 2 teams with a SOS value of 0% can very easily have schedules which when analyzed game by game would very obviously lead to very differnt records for teams of the same quality. In looking at a team under my method we could look at the 10th best team, or the average 6th seed in the playoffs, the options are a plenty. In fact, it can be calculated based on all kinds of teams, whether it be elite, great, good, average, poor, terrible. The wonders of computers lets the owners of this site churn this sort of data out pretty fast.

Also, to say that DVOA is linked to scoring, not to winning is just ridiculous. Points are directly related to wins and we can use that to estimate wins. We are given an estimated number of wins during the season and we are given how likely a team is to win a certain number of games in the preseason.

The type of SOS that I proposed would clearly be valuable. Maybe I was being harsh in saying that the SOS that FO gives us is clearly flawed, but it is imperfect and can be aided by providing more information.

174 Re: Week 7 DVOA Ratings

"What I am saying is that I come on to this website, which I have been doing for at least 5 years now, and conversations of what the SOS number means are constant."

Right, and when you change it again, you'll have the same discussions. Again - trying to represent a random distribution with 1 number doesn't work.

"Also, to say that DVOA is linked to scoring, not to winning is just ridiculous."

Really? Too bad it's true. DVOA's tuned to correlate to points and to be consistent year-over-year. A team with an offensive DVOA of +20% scores about 25.8 points per game (with an average defense and an average number of drives). Estimated wins, however, is based on the 'forest index':

"Forest Index: A response to a claim that total team DVOA rankings "missed the forest for the trees," the forest index (developed near the end of 2003) spits out an estimate of wins based on a formula that combines DVOA ratings in offense, defense, and special teams, as well as red zone defense, offense and defense in the second half of close games, offense in the first quarter, and variance (see VARIANCE)."

Points are not *directly* linked to wins - it's a timed game, and so certain *kinds* of performance (as mentioned above) lead to wins more often than you might think. Also, the game is a certain average length (~12 possessions, ~3-4 scores) so offense tends to be more relevant than defense because an equivalent percentage reduction isn't as beneficial as an equivalent increase given the small numbers involved. Now, granted, those kinds of performance might not be statistically relevant, but estimated wins is intended to be explanatory rather than predictive (which is what strength-of-schedule is anyway).

So in actuality, the SCHEDULE measure here isn't that bad when looking at DVOA to pseudo-"back out" the opponent adjustments. I don't think many people use it for that, though.

"The type of SOS that I proposed would clearly be valuable. Maybe I was being harsh in saying that the SOS that FO gives us is clearly flawed, but it is imperfect and can be aided by providing more information."

It's valuable for *some* cases. In college football, facing (and being) an elite team is rare, whereas in the NFL, it's common, and in college football, you're only ever discussing the most elite teams anyway, which is why you frame SOS in that context. The 'average' team in college football is Northwestern, Fresno State, Minnesota. Even the weakest schedule in history couldn't boost a ~50th-ranked team up to the point where anyone would be talking about them for a BCS game.

In the NFL, though, you'd want to frame it in the context of an average team, since, well, on average, that's who you're talking about. The easiest way to do that is to do something like "estimated wins for an average team given this schedule."

You could also frame it in the context of the team *in question* and ask how many wins they would expect based on, say, an average schedule based on a multi-year average. That's yet another measure, though. As an example, Philly, in 2004, probably wouldn't've looked like it had that easy or hard a schedule (despite the -5% or so average ranking) by "estimated wins for an average team." It probably would've been something like ~9 wins or so. Most of their opponents were around -5% or so (MIN,WASx2,GB,CAR, etc.), and the dogs on their schedule were mostly balanced out, so the "gimme" wins are balanced with losses.

But Philly in 2004 was an elite team (using the ~30% DVOA number excluding weeks 16-17), so really, those -5% teams were pretty much "gimme" wins rather than, say, 55% wins, and so "estimated wins for this team based on an *average* schedule, compared to estimated wins for this team based on *this* schedule" could actually be positive (meaning the schedule was *hard*) simply because for a 30% DVOA team, an "average" schedule wouldn't include a 35% DVOA team.

But that's a different (and weird) metric. The one I've suggested before (in addition to the SCHEDULE measure) is what I mentioned above: estimated wins for an average team based on this schedule.

185 Re: Week 7 DVOA Ratings

The issue here is that you can get the same value for SOS while playing vastly different schedules.

That would still be true no matter how you calculate SOS. Always.

If I told you that a team had a SOS value of 0%, you have gained very little information on the median quality of opponent, which is for all intents and purposes more useful to know.

Using median quality of opponent means you lose any effects of outliers, and anything else that skews the distribution. For instance, if you look at median quality of opponent, the following schedules look identical:

A - Best team in NFL history, 2nd Best team in NFL history, 16th Best team of 2010, 17th Best team of 2010, 18th Best team of 2010.

B - 14th Best team of 2010, 15th Best team of 2010, 16th Best team of 2010, 2nd worst team in NFL history, worst team in NFL history.

Yet they aren't even close to being the same difficulty.

If you look at just about any collection of numbers, and try to represent them with just one value, you'll lose information somewhere. If you use median or mode, you lose information about the outliers. If you use mean, you lose information about how the numbers are distributed.

In looking at a team under my method we could look at the 10th best team, or the average 6th seed in the playoffs, the options are a plenty. In fact, it can be calculated based on all kinds of teams, whether it be elite, great, good, average, poor, terrible. The wonders of computers lets the owners of this site churn this sort of data out pretty fast.

Yes, but then you're not talking about one SOS number. You're talking about dozens of distinct SOS numbers to describe the same schedule. If you take any one of them, they are going to have just as many flaws and problems as the current SOS measurement. So, saying that replacing the current SOS measurement with your SOS numbers would provide more information is like saying that upgrading the O-line would help a team's offense more than upgrading the RB. Of course it would - upgrading 5 players is likely to help more than upgrading one - but that doesn't make O-linemen more valuable than RBs.

The type of SOS that I proposed would clearly be valuable.

Well, sure. I don't think anyone's saying it wouldn't be nice to have. But it would also be a lot more work than the current one is, and it wouldn't fit as neatly into the current format. They'd have to make a page dedicated to nothing but various types of SOS. Now, I'm not against this, but it would be more difficult/cumbersome than the current SOS ratings.

Maybe I was being harsh in saying that the SOS that FO gives us is clearly flawed, but it is imperfect and can be aided by providing more information.

Once again, that depends entirely on what you're looking for from SOS. If what you're trying to find out is "why is Team A's DVOA so much different from their VOA?", then the current SOS ratings are pretty much perfect, and more information about what record a particular team could expect against that schedule would be superfluous. As far as I can tell, the only fair criticism of the current ratings from that perspective is that they don't account for the possibility that, say, Team A had many more plays in its games against easy opponents than it did in its games against hard opponents, which led to a higher VOA than expected, but an identical SOS rating. But that could be easily corrected by weighting opponents' DVOA in the SOS calculations by the amount of plays that occurred in the games against them. If you modified them in that way, the current SOS ratings would be a perfect answer to the question, "How much of the difference between Team A's DVOA and their VOA is due to the difficulty of their schedule?".

Now, maybe you think that the current ratings are too focused on VOA and DVOA as opposed to wins and losses. And I guess that's a fair point. But then we're talking about the broader issue of what the focus of this entire site should be - whether it's more important to get a better understanding of the statistics that give us insight into why teams win, or whether it's more important to get a better understanding of why certain teams have won/will win as much as they have/will. Both of those are worthy goals, and FO has always worked toward both of them to some extent, but the former seems like it's always been the main focus of this site, and I don't see that changing anytime soon.

31 Re: Week 7 DVOA Ratings

Consider the CLE rushing offense:
Against BAL they are successful on 14 of 30 rushes including 4 rushes of over 10 yards (12,19,25,48) with 1 rush for a loss. Average 5.97 YPC. Game rushing DVOA=19.5.
Against PIT they are successful on 5 of 21 rushes including 1 rush over 10 yards (14) and no rush for a loss. Average 3.14 YPC. Game rushing DVOA=-19.4.
Against NOR they are successful on 8 of 23 rushes including 2 rushes of over 10 yards (12,12) with 4 rushes for a loss. Average 3.26 YPC. Game rushing DVOA=16.6.
It seems that the the PIT and NOR games are more similar than the BAL and NOR games yet DVOA says the opposite. And since PIT is #1 in DefRushDVOA and NOR is #14 (BAL is #9) it would seem that VOA was even farther apart. 1 fumble in each game so no real difference.
What factor is so strong as to force this result?

38 Re: Week 7 DVOA Ratings

That's what I did. I used 45% of needed yardage on first down, 60% on second down, and made on 3rd and 4th down to count successful plays. That seems to be what the site says they are doing at a basic level with additional points added in an undisclosed manner. This is from the playbook and doesn't include kneeldowns.

70 Re: Week 7 DVOA Ratings

I agree those raw stats don't seem to line up with the DVOA rushing numbers. Off the top of my head (and IANAS, I've just read the official stat explanations) I think the big adjustment may be situational based on score.

In the NO game the Browns were up 20-3 at halftime. Teams with a two-score lead in the second half are going to run the ball a lot, and I expect they get less per-play production from their runs. Since DVOA would compare the Browns runs to rushing plays from teams in similar situations that probably bumps it up.

In the PIT game the Browns were losing from the 2nd quarter onward. I expect teams that are losing in the second half get more per-play production from their runs, so that would depress the Browns ratings.

I'm not sure that my explanation is correct, it's just the one that makes sense to me both statistically and in football terms. 3 yards and a cloud of dust is fine when you're chewing clock to protect a lead, it's unproductive when you're losing and need to catch up.

68 Re: Week 7 DVOA Ratings

The game situation (I think) plays a role here. In the NO game, for example, the Browns were well ahead for a large portion of the game. Being successful running in that situation may be more dificult and more valuable than in a situation like the Pittsburg game where they were well behind for a good bit.

140 Re: Week 7 DVOA Ratings

You're looking at more of a 'V' than 'VOA'. With New Orleans, those 8 successful rushes were 2nd & 2 for 3 yards, 1st and 10 for 8 yards, 3rd and 2 for 6 yards, 1st and 4 for 4 yards (TD), and then ( 1st and 10 for 9 yards, 2nd and 1 for 2 yards, 1st and 10 for 12 yards, 1st and 10 for 12 yards ). Those last four plays are important - they're in the fourth quarter, and teams rush *worse* in the fourth quarter with a lead (since they rush more often - you don't often see drives consisting of like, 75% runs). Cleveland ran almost exclusively on that drive in the fourth quarter, bled 7:30 off the clock and still scored 3 points. That's way better than an average team would do.

That's where they racked up all of the value. A 12 yard rush on 1st and 10 in the 4th quarter is much more valuable than a 14 yard rush on 2nd and 7 in the 2nd.

Also, that "no rush for a loss" bit is misleading. They had something like 5 or 6 rushes for no gain, which isn't really significantly better than a rush for a loss of 1, and the "1 fumble in each game" is wrong - the Browns didn't fumble on a rushing play in the New Orleans game. They did in the Pittsburgh game.

62 There really are some

There really are some muddled W-L records so far this year. According to DVOA, 7 of the worst 10 teams in the league have won at least 3 games. A lot of it must simply be down to the schedule, that has pitched a lot of these teams against each other so far. But some of these teams must have had an awful lot of luck in achieving their record so far.

33 Re: Week 7 DVOA Ratings

Please help me out here. The Colts and Pats are looking pretty similar on the surface, except for ST, which is, IIRC, about 1/7 of the total. (and about which, I'll make my annual appeal to say: it's overweighted! Okay, done.)

Pats O is ranked #2 at 29.0% and Colts are #3 at 25.6%. Pats D is ranked #27 at 9.9% and Colts 23rd at 7.8%. Based on that alone, you'd expect the Pats DVOA to be about 2.0% superior to the Colts.

So is the heart of the pretty wide gap all ST--where the 5th ranked Pats are 7.1% and the 29th ranked Colts are -5.6%???

If that's the case--the entire DVOA case against the Colts, I'll take it. They've already cut the returner who fumbled twice in one game. Their historical MO has been to secure the ball and hand it over to Manning with mediocre position, rather than risk fumbles trying for big returns. So their philosophical baseline is below average in the return game to begin with and has been for years. I'm relatively confident in their punter (when he sobers up and dries off) and kicker, which leaves coverage units.... hmmmmm. Always a sore spot. And with so many injured safeties and such, not likely to improve.

Assuming O and D are what they are and will be, and the ST stabilizes to a more respectable "gentleman's C" of about 0% to negative 2.0% (ranked 16-20), would that really move them up to the low-20s from their current DVOA of 12.3 and DAVE of 13.6?

I know that wins are not part of DVOA, so I won't snipe about NYG or KC being ahead of Indy, but something looks amiss. SD being so highly-rated looks eerily similar to the traditional NFL rankings based on yards--we know they are relatively meaningless and that points are a bit closer to the truth.

Just in the top 3rd of the league, Jets and Colts look too low, Giants, Phi, SD, and KC too high. Granted, the top six are so close in DVOA that it's essentially a tie.... but I'd put the 7th ranked Jets ahead of at least half of them, and the Colts and Ravens ahead of a few as well. Time will tell.

51 Re: Week 7 DVOA Ratings

The Colts and Pats are looking pretty similar on the surface, except for ST, which is, IIRC, about 1/7 of the total. (and about which, I'll make my annual appeal to say: it's overweighted! Okay, done.)

I'm sure Norv Turner will be comforted by the fact that someone else thinks special teams are overrated.

I doubt the Colts coaching staff really considers better or longer returns to have some supremely high turnover risk. Unfortunately, I don't have the data to see if longer returns are more prone to turnovers (from what I have seen, turnovers on special teams are primarily from a muffed reception, and distant second to that from a well-played tackle/forced fumble). I think the Colts just don't pay much attention to improving the special teams talent (or coaching staff).

San Diego is highly rated because their defense and offense typically perform pretty well. And then their special teams do them in.

150 Re: Week 7 DVOA Ratings

San Diego is making a really good case that while normal special teams might be worth 1/7 overall, really exceptionally bad special teams can be worth as much as your offense and defense combined.

Devin Hester a few years back made a pretty good case that exceptionally good special teams play can be worth as much as your offense, if that offense is average.

That said, any kind of system is going to falter when you hit the edge of the bell curve. Beating probability like that almost requires circumstances that violate the normal patterns.

162 Re: Week 7 DVOA Ratings

DVOA shows San Diego's special teams to be more important than either their offense or their defense, so I don't think it's off by as much as I think you think it's off: OFF +15.6%, DEF -15.6%, ST -17.4%.

35 Re: Week 7 DVOA Ratings

off-topic
Trent Dilfer is saying that it's not Brad Childress's call whether Favre starts or not. Apparently, even if Favre's left leg falls off, it's still his call, since Favre has a long career of accomplishing whatever he sets his mind to.
I wish ESPN would fire Dilfer. "Irrational" doesn't begin to cover it.

52 Re: Week 7 DVOA Ratings

I think the problem may be that you're watching games on ESPN with the sound on. Find your 'Mute' button, you'll be much happier for it.

Having said that - as a Vikes fan, I wouldn't object to Favruh's leg falling off. I just wish the management had made more of a long-term planning effort, rather than 'rehire the crusty jeans model, and go with Tarvaris 'Headlights!' Jackson as the backup', without even sniffing at a new QB to groom and prepare. One guy's 41, the other is obviously not starting-caliber.

40 Re: Week 7 DVOA Ratings

For masochistic purposes I would really like to see which performance by the Broncos was worse in DVOA terms: this week's home loss to Oakland, or the 1994 home loss to the LA Raiders ( http://www.pro-football-reference.com/boxscores/199409180den.htm ). Worst single game performances of the DVOA era, by team, would be a fun article.

59 Re: Week 7 DVOA Ratings

I don't know about DVOA, but I do know that the final result of the Den/Oak game fell 52 points away from the Vegas line (Denver were generally a 7 point favourite), which may well be some sort of record.

46 Re: Week 7 DVOA Ratings

These numbers are going to regress to the mean

Ahem, Aaron, that would be regress toward the mean. Didn't you say all the writers here would make that a point of emphasis? Please be better than NFL officiating!

74 Re: Week 7 DVOA Ratings

It's really :

"These numbers are going to regress to the mean a bit..."

The "a bit" part would imply toward the mean without really reaching it.

49 Re: Week 7 DVOA Ratings

This is an extremely odd line:

30 BUF -28.4% 0-6 -27.0% 1.6 32 7.8% 3 7.9% 2 9.5% 10

53 Re: Week 7 DVOA Ratings

In reply to by kamchatka (not verified)

I don't know which part you're calling 'odd'. Beau Fleuve is the 30th ranked team in terms of DVOA, their non-adjusted VOA is close to their DVOA. The only thing that might be somewhat odd is the Forrest index projecting them with 1.6 wins, based on 7 games played (remember that teams that have had their bye are projected as if they had played that week).

They have had a tough schedule (3rd toughest), they have a tough remaining schedule (2nd toughest), and their generally consistent in their suck (10th overall in variance).

54 Re: Week 7 DVOA Ratings

Okay, it really is time to revise the 'A' in DVOA.

There are now 25 passing offenses above average. Twenty-one of those are above 12%, including the Browns and Lions. You sacrificed DPAR for DYAR to make the stat easier to understand; you simply have to do the same for DVOA. Nobody looking at Cleveland's +14% DVOA is going to think "Ah, that's average!"

And it's not as if we're seeing a great improvement in passing offense this year. The top teams are not as lights-out as they were in 2008-09 (QB interception rates have returned to earth, except in Indianapolis); the mass of teams in the middle has improved fractionally in stats like completion percentage and yards per attempt; and there's a huge gap down to the three or four teams with really horrible DVOA, which brings down the mean.

If you were using a 2010-only baseline, with the mean of passing offense set at 0% (rather than 14.1%), you'd still end up with more teams above than below average. Without the DVOA formula it's hard to tell, but probably none would be above +35% (which fits the impression that there is no extraordinary passing offense this year), although upwards of a dozen teams would be squashed between +10% and +25%. At the bottom end, Chicago, Arizona and Carolina would rank as some of the worst pass offenses of all time — relative to the league average.

63 Re: Week 7 DVOA Ratings

I think there has actually been a great improvement in passing offense this year, and that's what's driving the change in the "A." I agree that a re-adjusted "A" would be nice, but I wonder if the last couple of years is enough sample size to make it meaningful. I expect that if we go back farther than that, we'd start seeing the same "A" we have now.

67 Re: Week 7 DVOA Ratings

Where is this improvement? Completion percentage is up by 0.1%, and yards per attempt by an even smaller fraction. Interception and sack rates are the same. The only significant increase is in attempts (34.0 per team per game, most since 1995), which would affect DYAR but not DVOA.

The change in pass offense this year seems, on the face of it, to be that there are no outstanding units and fewer really sucky ones (see: Cleveland, Oakland, Tampa Bay, etc).

79 Re: Week 7 DVOA Ratings

Others have written about the increase in offensive passing DVOA the past 5 years ago (perhaps it was longer, I haven't checked the numbers recently). This rise interests me as well; maybe I can find some of the threads later on today.

90 Re: Week 7 DVOA Ratings

A reconsideration of the baseline we use as "average" is coming this offseason. Reconsideration of how we calculate SOS is probably coming as well. However, to the people who want to complain about it every week (not you, Ammek), I will note that complaining about it will not make it come any faster.

100 Re: Week 7 DVOA Ratings

That's good news — thanks. And yeah, I complain about it maybe twice a season, so I only feel *averagely* a bore for doing so.

I'll be interested to see what you find when you do the reconsideration. It's hard, using only conventional stats, to get a grip on the nature of how pass offense has progressed over the past few years. The consensus opinion seems to be that "rule changes" have led to passing becoming "easier". But all of the trends in conventional stats — improved completion percentage; decreases in yards per completion, pick rate and sack rate — are long-running, many of them have trended steadily since the early 1990s or even prior to that. Will it be enough to divide DVOA according to a 'pre-2007 baseline' and a '2007-and-after baseline'? Or will it be necessary to draw up a rolling baseline (and if so, what will you do with the outlying years, 1995 and 2002?)? And how will rush offense be affected by a change in baseline: is the 'average' in Run DVOA an average of running plays only, or is it based on success rates for all offensive plays, including passes?

Now isn't the time to answer these questions, but I hope that you'll have a chance to explain your findings when you go through the formula in the offseason.

136 Re: Week 7 DVOA Ratings

Can I ask, is part of the SOS reconsideration looking into whether there is any sort of feedback effect? It's a good time to bring up, considering that NFC West question. Which has been quite bad through the DVOA era, unquestionably. But the severity sometimes seems amplified.

56 Correction?

The special teams DVOA numbers are different on the Special Teams page compared with the Team Efficiency table (ie, on this page). San Diego, for example, is a historically bad -17.4 in the chart above and a mind-bogglingly even worse -19.3% on the special teams page.

72 Re: Correction?

In reply to by ammek

I noticed the same thing last week and sent an email to Aaron about it. Almost all the teams had very different ratings in the two tables.

86 Re: Correction?

In reply to by ammek

San Diego's special teams DVOA was 19.3% last week. Maybe it hasn't been updated yet?

93 Re: Correction?

In reply to by Andrew Potter

Sorry. For some odd reason, this is the second straight week I forgot to hit the button for special teams when I was going through the program to create all the HTML, so I re-posted last week's by accident. I'll go fix now.

57 Re: Week 7 DVOA Ratings

Wait, the 49er's DVOA improved? Is that opponent adjustments, or does DVOA think they played decently in Carolina?

61 Re: Week 7 DVOA Ratings

Looks like a number of teams DVOA shifted in the opposite direction from the outcome of the game. San Diego lost and moved up 5 spots. Miami lost and moved up 3. SF, as noted, moved up 2 despite a loss.

Meanwhile, Washington won and dropped 4 while Seattle won and dropped 2.

Opponent adjustments or on field performance? Probably a little from column A and a little from column B. Having watched both the Chicago/Wash and Sea/Ari games my eyes saw two teams win with very lacklustre play. Arizona self-destructed with some terrible special teams play and horrendous quarterback play from Max Hall. Seattle had the fumble luck and the special teams advantage yet was completely unable to put the game away. 1 TD in 7 red zone trips. Yuck.

Meanwhile, the Bears/Washington game may have been the most pathetic display of 'professional' football I've ever seen. I actually missed the first half and turned it on just in time for the worst/stupidest comedy of fumbles and interceptions I've witnessed in almost 30 years of Sunday games. Washington stunk and its probably safe to say that fumble luck was the difference in the outcome of the game. I mean, sure, DeAngelo Hall had a pretty great individual game but a couple of those balls were thrown right to him. Hall notwithstanding, Wash dropped 4 spots and totally deserved it. Meanwhile, Chicago was completely wretched and didn't move in the standings at all. I guess they played to their level. Yikes. For a team leading its division: double yikes.

Maybe we'll see a divisional round playoff game between the Bears and Seahawks that will feature 8 interceptions, 3 fumbles, 3 punts returned for touchdowns, and 9 field goals. The winner will face the Giants who will promptly knock the starting QB out of the game, and despite no drop off on offense for Sea/Chi, the Giants will prevail. The Superbowl will feature the Giants outplaying the Steelers but losing the game on a series of questionable calls by the officals and the worst season in NFL history will come to an end.

78 Re: Week 7 DVOA Ratings

I was watching from the beginning, and actually sent a text to my friend that "McNabb is looking great against a pretty good Chicago defense" in the 2nd quarter, shortly before the wheels came off.

120 Re: Week 7 DVOA Ratings

Is it just me or does Washington look great early in every game and get progressively worse as it goes along?

107 Re: Week 7 DVOA Ratings

San Diego and Miami lost to two of the top three teams in very close games. An increase in the respective DVOAs is reasonable.

We don't have the same explanation for the 49ers, who lost to the awful Panthers.

126 Re: Week 7 DVOA Ratings

Honestly, do people remember 2004, where there were two NFC Wild Cards at 8-8, and the second seed Falcons had a +3 point differential, and the sixth seed Rams had a -73 point differential.

130 Re: Week 7 DVOA Ratings

Yeah, but the AFC was loaded. The Pats, Colts, and Steelers *all* arguably hit a peak that year, and the Chargers, Broncos, Bills, Jets, and Chiefs were all good enough to contend.

142 Re: Week 7 DVOA Ratings

The AFC Was loaded, but the NFC was arguably worse than this year. When people claim that "this is the worst year in history" they are generally talking about the NFC teams, as the AFC teams do have some quality to them this season, and I would put the NFC batch of 2004 (or even 2006) right along with this one.

64 Re: Week 7 DVOA Ratings

Wait, Ben Roethlisberger is #1 in DVOA? There must be some mistake, Aaron was very clear that Ben is not a top 5 QB.

108 Re: Week 7 DVOA Ratings

The full quote was something along the lines of "I don't care what the statistics say, Ben Roethlisberger is not a top 5 QB."

A statistically-based rebuttal is not all that relevant.

Now you could take issue with Aaron for ignoring his own stats when convenient, but that's a different argument.

115 Re: Week 7 DVOA Ratings

@Vicious - where are you pulling that stat from - a subscriber part of the stats, I'm not seeing it on this page.
(I'm asking because I was the one that started the "who are the Top 5 QB's if Ben is not one of them" thread on this weeks Quick Reads)

Steeler fan trapped in Houston!
Six Time SB Champs! ;-)

133 Re: Week 7 DVOA Ratings

He's #1 based on all of two games.

Roethlisberger is an above average quarterback, right around Donovan McNabb territory only with more goofball interceptions and fewer goofball incompletions.

73 Re: Week 7 DVOA Ratings

It's hardly unexpected, but the top teams on the 'worst 3rd/4th down DVOA' list were an unholy mix of already dreadful offense + rookie quarterback (David Carr, Kyle Orton, Alex Smith … and Max Hall?). Further down the list we get some creaky old QBs, once known for their athleticism, hobbling into retirement: Warren Moon in his last year, Donald Hollas, and the decaying corpse of Rodney Peete. The rest are QBs who were inexperienced, but not rookies: Jon Kitna, Stoney Case, Tony Banks, the sorely missed JaMarcus Russell, and Todd Collins. Collins may find himself on the list anew as a Bear, but the most worrying match for Jay Cutler must be Jeff George, who pops up twice, in Indianapolis and Oakland.

116 Re: Week 7 DVOA Ratings

Orton was a rookie in 2005. The 2004 Bears QBs were Chad Hutchinson, Craig Krenzel, Jonathan Quinn, and Rex Grossman. Yeah, it was bad.

83 Re: Week 7 DVOA Ratings

"Since the NFL went to eight divisions in 2002, the six worst divisions (based on average DVOA of the four teams) have all been the NFC West."

This is absolutely boggling. Given the limited number of years we're looking at, this means the NFC West's third best performance since 2002 is still worse than any performance by any other division.

Yet hilariously, the division has still represented the NFC in the Super Bowl twice, and has racked up 10 playoff wins over that period. That's four more than the NFC North and AFC West, and only one less than average.

84 Re: Week 7 DVOA Ratings

DVOA is cumbersome for comparing teams, and frankly I don’t know what the number really means. Therefore, I’ve taken the liberty to provide the following.

I’ve converted this weekend’s DVOA into numbers that make more sense (to me). Caveat: There is error in these numbers, I’ve provided the correlations for you to assess the error and value.

For TOTDVOA, I’ve converted to Wins in a season. This conversion utilizes the following formula WINS = 13.785*(TOTDVOA) + 7.9307. The formula is the regression equation for the correlation between TOTDVOA and Wins (r = 0.854)

RNK TM WINS
1 NYG 11.8
2 PIT 11.8
3 NE 11.5
4 KC 11.5
5 TEN 11.5
6 PHI 11.4
7 NYJ 10.8
8 SD 9.8
9 ATL 9.7
10 IND 9.6
11 GB 9.5
12 BAL 9.1
13 MIA 8.7
14 NO 8.6
15 HOU 8.4
16 SEA 8.2
17 DAL 8.2
18 MIN 7.5
19 CIN 7.4
20 CLE 7.4
21 DET 7.3
22 SF 7.0
23 WAS 6.9
24 CHI 6.4
25 TB 5.9
26 OAK 5.2
27 STL 5.1
28 JAC 4.4
29 DEN 4.4
30 BUF 4.0
31 CAR 3.3
32 ARI 2.7

For OFFDVOA and DEFVOA, I’ve converted to pts/gm using the following formulas:
PTS/GM=(380.83OFFDVOA + 341.05)/16 (r = 0.822)

RNK TM PTS/GM
01 HOU 28.93
02 NE 28.22
03 IND 27.41
04 PHI 27.00
05 KC 25.50
06 GB 25.31
07 SD 25.03
08 ATL 24.89
09 NYJ 24.84
10 DAL 24.62
11 NYG 24.22
12 MIA 23.22
13 NO 23.10
14 BAL 23.03
15 PIT 22.98
16 CIN 22.32
17 DEN 21.79
18 TEN 21.48
19 WAS 20.93
20 BUF 20.91
21 DET 20.46
22 CLE 20.24
23 JAC 19.98
24 TB 19.89
25 SEA 18.98
26 OAK 18.77
27 SF 18.63
28 MIN 18.22
29 STL 17.96
30 CHI 14.44
31 ARI 12.82
32 CAR 11.89

PTS/GM=(406.5*DEFDVOA + 343.09)/16 (r = 0.776)
For ST DVOA, I’ve converted to point differential per game using the following formula:

RNK TM PTS/GM
01 NYG 15.63
02 TEN 16.13
03 PIT 16.92
04 SD 17.48
05 CHI 18.93
06 CAR 19.51
07 MIN 19.66
08 SF 20.12
09 KC 20.33
10 PHI 20.45
11 SEA 20.94
12 GB 21.16
13 MIA 21.16
14 NO 21.52
15 NYJ 21.57
16 BAL 22.03
17 ARI 22.08
18 ATL 22.21
19 STL 22.26
20 CLE 22.28
21 WAS 22.41
22 CIN 23.15
23 IND 23.42
24 DET 23.53
25 TB 23.60
26 OAK 23.68
27 NE 23.96
28 DAL 24.42
29 DEN 27.54
30 HOU 28.20
31 JAC 28.46
32 BUF 29.40

PTDIFgm = 49.306*STDVOa - 0.0412 (r = 0.278)

RNK TM y = 49.306x - 0.0412
01 SEA 4.84
02 CHI 3.80
03 JAC 3.71
04 DET 3.56
05 NE 3.46
06 NYJ 3.07
07 BUF 2.23
08 TEN 2.08
09 KC 1.98
10 CLE 1.73
11 BAL 1.68
12 PIT 1.44
13 MIN 1.24
14 ARI 0.25
15 ATL 0.21
16 DAL -0.09
17 OAK -0.14
18 TB -0.29
19 SF -0.53
20 CAR -0.73
21 CIN -0.78
22 HOU -0.88
23 WAS -0.98
24 PHI -1.37
25 NO -1.37
26 MIA -1.67
27 STL -1.77
28 DEN -1.96
29 IND -2.80
30 GB -3.20
31 NYG -3.54
32 SD -8.62

Final, I’ve combined as follows: OFFPT/GM – DEFPT/GM +PTDIFgm = PTDIF

RNK TM OFF-DEF+Stdif
01 NE 7.72
02 PIT 7.50
03 TEN 7.43
04 KC 7.16
05 NYJ 6.33
06 PHI 5.18
07 NYG 5.05
08 SEA 2.89
09 ATL 2.89
10 BAL 2.69
11 IND 1.18
12 GB 0.95
13 DET 0.49
14 MIA 0.39
15 NO 0.21
16 DAL 0.12
17 HOU -0.15
18 MIN -0.20
19 CLE -0.30
20 CHI -0.69
21 SD -1.07
22 CIN -1.61
23 SF -2.03
24 WAS -2.45
25 TB -4.00
26 JAC -4.77
27 OAK -5.05
28 STL -6.06
29 BUF -6.26
30 DEN -7.71
31 CAR -8.35
32 ARI -9.01

That’s my two cents worth for now.

88 Re: Week 7 DVOA Ratings

Cool. I think I understand everything, except for how you came up with the ST equation. Oh, and what data you used for the rest (this year, last years, several years, etc).

95 Re: Week 7 DVOA Ratings

Sorry.

PTDIFgm = 49.306*STDVOa - 0.0412 is the regression equation for the correlation between STDVOA and per game pointdifferential.

This correlation is slightly than the correlation between STDVOA and wins. Since STDVOA contains both Offensive and Defensive elements of the game, it made sense to me to use this metric rather than offensive points or defensive points and it allows for a combination of all three DVOA components with a different metric (OFFPT/GM – DEFPT/GM +PTDIFgm = PTDIF). I realize it is not a particularly robust correlation, but it is statistically significant. It adds interesting, although possibly pointless opportunity for debate...i.e. TOTDVOA v. this statistic as a marker of team quality.

The correlations and their equations are based on data from 1996 to 2008.

110 Re: Week 7 DVOA Ratings

If you are simply regressing STDVOA onto point differential, won't that end up giving STDVOA more credit than it should have? Maybe that would only happen if STDVOA is correlated with OFFDVOA or DEFDVOA... What about fitting the following:

PointDif/G = b1*OFFDVOA + b2*DEFDVOA + b3*STDVOA + b0

Then using the following to convert STDVOA to a point value:

STptval/G = b3*STDVOA

ACTUALLY, why not just use FO's point values for special teams directly? They calculate point values for ST, then convert to DVOA for display here. Simply using the point values listed on the ST pages seems like a good way to go.

112 Re: Week 7 DVOA Ratings

"won't that end up giving STDVOA more credit than it should have?"

I don't know. To be honest, I'm starting quick and dirty and trying to find a way to make DVOA make a little more intuitive sense and haven't thought through. You're correct regarding overvaluing STDVOA. Maybe I should have multiplied it by the R-Square of the correlation coefficient.

I guess I could (should) do the same with the other points; however, it's worth noting that when you enter a 0.00% DVOA into these equations, the output comes within a decimal place or two of the league average over the years, so I don't know.

Other statheads have thoughts?

While I'm commenting here: Regarding the WINS value. My interpretation is that TEAM X's DVOA of Y is consistent with a team the wins Z games in year (Substitute points as appropriate). It should be noted that in general, and I don't have the specific numbers on hand b/c this is the first time this season I've pulled this out, that the regression equation is fairly effective for "predicting" within +/- 2 games. Unfortunately, a 4 game swing is huge around the mean. Within +/- 1 game, it hits on about 50% if I recall. I'll look into this more.

175 Re: Week 7 DVOA Ratings

"I don't know. To be honest, I'm starting quick and dirty and trying to find a way to make DVOA make a little more intuitive sense and haven't thought through."

To be honest, if you want a quick and dirty way, Aaron did it before:

http://footballoutsiders.com/dvoa-ratings/2005/2005-dvoa-projections

You just add 1, and multiply by 21.5 (special teams you just multiply by 21.5).

"You're correct regarding overvaluing STDVOA. Maybe I should have multiplied it by the R-Square of the correlation coefficient."

You're actually probably overvaluing them all. You really want to do a multiparameter regression straight off to point differential for a combined measure. I would do it to point difference/*drive* rather than ptdiff/*game* since points/drive is pace-independent. You can get those numbers from the drive stats game (if you want to go back to points/game for basic reference, just multiply by average drives/game). The statistical properties might look a little worse, but it's at least more fundamentally sound.

The problem is that, if you think about it, imagine an ideal world, with football having offensive performance O, defensive performance D, and special teams performance S, even if offensive DVOA is perfectly correlated to O, points scored/drive (PS) isn't sum over all drives(PS(O)). It's sum over all drives(f(PS(O),starting field position for drive), and starting field position, on average, is correlated to D and S. So really, it's sum over all drives(f(PS(O),D,S)).

Ditto with points against. So when you do a 1-D regression on each, you're just averaging the influence on each, but, of course, there is a natural correlation between O, D, and S: the team. So you overestimate the influence of each, and then add the overestimated measure together, because you have few "good offense, bad defense" teams in the sample.

(Adding DVOA directly, the way it's done here, is actually okay because they're built upon independent measures: the plays themselves. To first order, excluding score effects which are corrected for, teams don't play offense differently because of another team's offense.)

But to be honest, I think if you do all that, you're... basically recorrelating DVOA back to the set it was tuned on, and I'd be surprised if you got something much different than the simple scaling.

176 Re: Week 7 DVOA Ratings

In reply to by Pat (filler) (not verified)

Oh, just to expound on the reason for the natural O/D/S correlation - when a team's bad, they tend to be bad everywhere, and if they're good somewhere, they're likely not terrible elsewhere. A simple correlation of offensive/defensive DVOA will show that - it's a fairly sizable correlation.

178 Re: Week 7 DVOA Ratings

"A simple correlation of offensive/defensive DVOA will show that - it's a fairly sizable correlation."

Uh, no r = 0.007, or zero.

181 Re: Week 7 DVOA Ratings

Yeah, welcome to statistical weirdness. I should've explained better.

Correlation, O/D, year by year (p, rsq, slope)
2009: -0.357, 0.128, -0.55
2008: -0.173, 0.03, -0.18
2007: -0.32, 0.102, -0.58
2006: 0.0594, 0.004, 0.08
2005: -0.341, 0.116, -0.56
2004: 0.0913, 0.008, 0.12
2003: 0.0708, 0.005, 0.08
2002: 0.0499, 0.002, 0.07
2001: -0.07, 0.005, -0.08
2000: 0.0575, 0.003, 0.07

Note that there are no strong positive correlations: in a truly random sample the correlation coefficient would be randomly distributed. So what does that mean?

It means that if you imagine a 4-quadrant space with "-/+ O/D", "+/+ O/D", "-/- O/D", and "+/- O/D", that 3 of those 4 quadrants are well-populated: -/+, -/-, and +/-.

+/+ (good offense, bad defense - e.g. the 2004 Minnesota Vikings quadrant) is really poorly represented (well, spatially). A graphic helps: ( http://img138.imageshack.us/img138/4269/32917315.png ) - here X is offense, Y is defense.

So yeah, I should've just said "just plot the two" and it would've been obvious they're not random variables.

177 Re: Week 7 DVOA Ratings

Thanks, that's way more work than I want to commit to this, as I said, I'm looking for something quick and dirty, makes sense. I'm not really trying to predict as much as simplify things a bit.

"don't play offense differently because of another team's offense"

Are you sure? The most obivous example is that the conventional wisdom that teams look for ways to keep the Colts Offense off the field. It seems reasonable that there could be other examples to a lesser degree.

179 Re: Week 7 DVOA Ratings

I think he was referring to a far more simple effect. If your opponent scores a lot of points, you have to try and score a lot points. That leads to riskier/more aggressive play calling.

182 Re: Week 7 DVOA Ratings

"Are you sure? The most obivous example is that the conventional wisdom that teams look for ways to keep the Colts Offense off the field. It seems reasonable that there could be other examples to a lesser degree."

Yeah, that's second order. The "keep the Colts offense off the field" is quite possibly one of my most hated things announcers say. They're not trying to keep the Colts offense off the field, they're trying to shorten the game, because the more possessions a game has, the more an offense has an advantage. The Colts offense is going to come on the field after them no matter what.

121 Re: Week 7 DVOA Ratings

The concept behind DVOA is actually pretty simple. It's the difference between a team and the NFL average on a per play basis, adjusted for opponent strength.

This makes more sense to me than some kind of arbitrary "this is how many points a team should score" stat.

122 Re: Week 7 DVOA Ratings

I think for an overall metric, something on the scale of points or wins might be good, which is why I'm interested in the above regressions. However, for all the breakdowns that can be done using DVOA (throwing to a TE, 3rd & long plays, plays where the team is down by a FG in the 4th quarter, etc), the % above/below average is much more intuitive. Because those need to be per play, and saying a team is "0.002 points per play above average on 3rd and long plays" ... just doesn't work at all.

124 Re: Week 7 DVOA Ratings

I'm glad it does for you. However, I have no idea in the real world what the difference between 2 team's DVOA's can be translated to. For example, that the difference between HOU's ant PIT's offense is basically a touchdown rather than 25 DVOA percentage points.

I'm not saying this information is better than DVOA, and in fact it is in some ways worse b/c there is an extra margin of error in these numbers.

But if I sit down in a bar to have a discussion with a member of the unwashed masses, saying to them that "Well, HOU's Offense DVOA of 32.0% means they're better than PIT's Offense DVOA of 7.0%" because of theri difference between their play and the NFL average on a per play basis, adjusted for opponent strength,"
is not going to go over too well. But if I can say that in general I expect HOU's offense to be able to produce one more touchdown per game than PIT's, then I have a little more credibility...that is until I have to explain that this number is derived from the regression equation of the correlation coefficient between DVOA and Wins in a year.

More importantly though, the fact that PIT's and HOU's offenses are roughly only a touchdown apart may tell me something about the relative importance of a fortunate bounce of the ball here or there.

125 Re: Week 7 DVOA Ratings

You can't just say "the Houston offense is about 25% better than the Pittsburgh offense?" It's pretty easy to understand what 25% better means.

According to your numbers, the difference is 6 points, not a TD. Which makes thing even more confusing. Does that mean the offense is going to drive down the field two extra times to score 2 FGs? Or does it mean the offense is going to score an extra TD on a single drive? There's a big difference between those two outcomes.

If you want to keep making these feel free, I just wanted to voice a contrary opinion. The theory behind DVOA is actually quite simple. The implementation is the tricky part, and if you believe DVOA is actually measuring what it claims to be.

127 Re: Week 7 DVOA Ratings

According to your numbers, the difference is 6 points, not a TD. Which makes thing even more confusing. Does that mean the offense is going to drive down the field two extra times to score 2 FGs? Or does it mean the offense is going to score an extra TD on a single drive? There's a big difference between those two outcomes.

Really? He said it was a per game average, it's not that complicated. Your comment above is akin to him asking you "Houston is 25% OF WHAT better? Are they going to score 8.75 points on a drive where Pittsburgh scores 7?

Though I would agree that expressing the offenses in terms of points per game is misleading. Some of the value of the offense could be translated into worse opponent field position, which is more like preventing points than scoring points. If we want to express the OFF/DEF/ST in terms of points, a point differential for all of them might work better than points scored/allowed. So, for example, Houston's offense is +7.6 pts/g, while Pittsburgh's is +1.7. Meaning that an average team, paired with Houston's offense, facing another average team, would expect to win by 7.6.

Again, this would be more intuitive to me, when comparing overall team strength. But once you do any kind of breakdown (even into passing DVOA and rushing DVOA), the approach doesn't really work so well any more.

128 Re: Week 7 DVOA Ratings

I can say they're 25 percentage points better, but I don't know what that means as I'm watching a game.

Which is intuitively more meaningful, HOU is 25 DVOA percentage points better on offense or HOU's offense can be expected to be, on average, 6 pts/game better? (BTW: I'm aware that pts has the inherent prolem that not all points are scored by the offense - if I had some way to look at only points scored by the offense, the measure would be more accurate).

"There's a big difference between those two outcomes."
How, they both result in 6 points? There is a difference in the process.

"If you want to keep making these feel free."
Thanks. I guess at the end of the day, I'm trying to figure out what DVOA means in the real world.

143 Re: Week 7 DVOA Ratings

DW, what about stepping back a step or two instead of diving in deeper? Is there a way to just cluster the offenses in rough terms of performance/capability...then live with the fact that it's probably impossible to pin things down to the nth degree beyond those clusters.

Something like:

Elite offenses:
About a field goal worse than elite:
About 6-7 points worse than elite:

Etc...same thing on defense. I've talked in other places about thinking in terms of clusters rather than totem poles. To me, we're at a point where it's becoming a matter of applying super-math to areas where the inputs aren't that super-accurate to begin with. It's not "garbage in-garbage out" so to speak. But, "stuff you can't really know to the nth degree in---variations of unlikely to be pristine totem poles out." Everybody has their own totem pole...but reality isn't a totem pole (similar to what we're seeing with the various rankings over in the college stuff with S+P, FEI, F+ and all the varying elements in the BCS process). Reality is stuff clustering.

In terms of speaking English with everyone, and having discussions in bars...this may be a good way to go...and everyone would be on the same page in the discussions. Back up a few steps and see if there are reasonable ways to cluster the teams in a way that is universally agreed on. Might be harder than it seems. I'm not going to have Kansas City as a top five offense, for example, given that they couldn't drive the field for points vs. SD, Cle, SF, and Indy (7-6-7-0 points on drives of 60+ yards). I know they're capable of exploding vs. soft defenses like Jax and Houston. If we can agree on clusters universally...then move forward. If not, then doing regressions and the like is just carving up mud...

My take anyway. Sometimes zooming the microscope lens in too far messes up the image...