DVOA Analysis
Football Outsiders' revolutionary metrics that break down every single play of the NFL season

Week 11 DVOA Ratings

by Aaron Schatz

After 11 weeks, we're starting to see a little bit of separation in the Football Outsiders DVOA ratings. Last week, Philadelphia was a definitive number one, with a lot of other good teams packed behind them. This week, the Steelers jump up to join the Eagles in a Pennsylvania twosome, thanks to their big win over Oakland. The Steelers had 116.0% DVOA for the game, not only the top single game of the 2010 season but the only game currently listed with a DVOA over 100%.

Although the two Pennsylvania teams are currently on top of our ratings, the all-Pennsylvania Super Bowl is far from guaranteed. This week's playoff odds report gives the Keystone State Bowl a slightly lower chance than the Super Bowl XXXIX rematch because, with only six games left, that one-win lead that New England and New York have over Pittsburgh and Baltimore is very important. Even if the Steelers have the highest DVOA in the AFC, the Jets and Patriots have a better chance of getting home-field advantage throughout the playoffs. The Patriots are currently listed with a nearly 50 percent chance of the top seed; although the Jets have the current tiebreaker advantage, the Patriots have a higher DVOA and home-field advantage in the Week 13 rematch. The Jets get the top seed in one of four simulations, which is still more often than Pittsburgh, and the two AFC East teams are currently the two most likely teams to make it to the postseason.

Despite Pittsburgh's big week, we're still seeing more parity in the DVOA ratings than we have since the early years of the decade. However, this parity is only strange if we compare it to the recent past. The current standard deviation of DVOA is 19.9%. If we compare that to the standard deviation of DVOA after Week 11 in every season going back to 1993, that ranks only eighth out of 18 seasons. Here's a graph showing the standard deviation of DVOA in each year:

Parity usually brings us lots of teams that overperform or underperform their total DVOA, and this year is no exception. We've got three 5-5 teams in the top dozen.

First comes Tennessee, which dropped from second to fifth this week. Vince Verhei really gets into the Titans in today's Any Given Sunday, which you can read if you are an ESPN subscriber. If you aren't, I'll summarize: Tennessee still has excellent defense and special teams despite the mediocre offense. The Titans have also faced a tough schedule this year. Still, they aren't the fifth-best team in the league as currently constituted, not with Rusty Venture at quarterback.

Next comes San Diego, which actually dropped a spot to eighth despite beating Denver. Everybody knows why the Chargers are just 5-5: awful luck and the worst special teams in the last two decades, perhaps ever. The Chargers look pretty hot right now and are probably going to end up with a winning record.

Finally, we've got Miami. I got a tweet from @mike_toback last night, and he wanted to know why the Dolphins still rank highly in DVOA despite their 5-5 record. This week, they dropped one spot from 11th to 12th, but they still rank ahead of three different 7-3 teams, including a New Orleans team that many people feel has a good chance to repeat as Super Bowl champs. Miami's high rating heavily tied to its schedule, which has been the second-hardest in the league by average DVOA of opponent. (Cleveland is the team with the toughest schedule so far, and the Browns rank right behind Miami even though they're 3-7.) Miami's schedule has been particularly tough since Week 3. In their last eight games, the Dolphins have played seven teams in the DVOA top 10, plus Cincinnati (18th) and Chicago (19th). Overall, they come out as average or slightly above average in pretty much everything. The Dolphins are 13th in defense, 17th in special teams. They rank ninth overall in offense, even though they rank lower than that in both passing (11th) and running (a very surprising 18th). That's thanks to the other elements added into offensive DVOA a couple years ago; the Dolphins have only 11 combined false starts and delay of game penalties, tied for 26th in the NFL. The other element that makes them look worse than they really are: terrible luck with opposing kickers. Their opponents have been above-average on field goals, kickoff distance, and punt distance, which puts Miami 29th in the "hidden" special teams value.

The other part of parity has been the horrifying stench emanating from the NFC West. Every time one of those four teams looks like it might be climbing its way to respectability, it goes out and gets spanked. The latest victim was San Francisco, which drops from 17th to 21st this week. The 49ers are still the top team in the division by DVOA, with St. Louis, Seattle, and Arizona all sitting at 28th or worse. So starting this week, we're going to use the playoff odds simulation to figure out the odds that we're going to get the first-ever losing playoff team in NFL history. After Week 11, here are your odds:

  • Odds that the NFC West champion is 8-8 or worse: 70.8 percent
  • Odds that the NFC West champion is 7-8-1 or worse: 27.4 percent

One thing that's interesting about this year's parity is that it only exists in terms of total team performance. There are extremes at the top and bottom of the ratings for offense, defense, and special teams. (Well, the bottom of the ratings for defense, anyway. Not the top.) Let's go back to the Patriots for a moment. Do you wonder how the Patriots have gotten to 8-2 despite a young, undisciplined and outright poor defense? Well, no matter how good you think the Patriots offense has been, they've been better. Based on our numbers, this is the second-most efficient offense the Patriots have put on the field, trailing only the record-breaking 2007 team. In fact, this could be the best offense we've ever measured other than that team. Yes, the Patriots are only seventh in the league with 5.6 net yards per play, but they've been outstanding at extending drives and extremely good at avoiding turnovers. The Patriots are first in DVOA on third and fourth downs, and their 78.8% rating is far ahead of any other team. (Atlanta is second at 50.6%.) Meanwhile, only Kansas City (seven) has fewer turnovers than New England's nine, and one of those Patriots turnovers was a Hail Mary pass at the end of regulation before their overtime victory over Baltimore. The Patriots aren't keeping the turnovers down through fumble recovery luck -- they have only one fumble that they themselves have recovered to go with the three that their opponent have recovered. As a result, the Patriots are tied with the 2004 Colts as the second-best offense we've ever tracked through Week 11.

On the other side, you've got teams like Houston and Carolina. The Texans, of course, have a great offense and a defense that couldn't cover a fly with a circus tent. The Texans' defensive DVOA improves since they kept the Jets somewhat in check for 58 minutes, but they still end up as the worst defense we've ever tracked through 11 weeks. (I'm sorry I didn't notice before this week and write about this earlier.) The Panthers are actually worse off overall, as their impotent offense is only balanced by average defense and special teams. That makes the Panthers the worst team in the NFL this year -- and it means the time has come to bring back those tables you've seen in the DVOA commentary in past years, checking out how current teams compare to the best and worst teams of the DVOA Era (back to 1993). We'll toss in the San Diego special teams in there while we are at it.

Best and Worst DVOA Ever Watch

2007 NE 53.6% x 2005 SF -46.7% x 2010 HOU 29.7% x 2010 SD -14.7%
2004 IND 40.7% x 2007 SF -43.0% x 2008 DET 29.2% x 2008 MIN -11.4%
2010 NE 40.7% x 1993 TB -42.6% x 2008 STL 28.3% x 1997 PHI -10.9%
2002 KC 37.2% x 2010 CAR -42.5% x 2004 STL 28.0% x 1995 PHI -10.2%
2005 SD 34.8% x 2002 HOU -39.8% x 2005 HOU 27.1% x 1997 STL -9.6%
1998 DEN 34.6% x 1997 NO -39.8% x 2004 SF 25.1% x 2006 ARI -9.0%
1995 DAL 33.2% x 1999 ARI -36.5% x 2009 DET 25.0% x 2007 CAR -8.9%
2000 STL 33.1% x 1996 STL -36.4% x 2000 SF 24.8% x 2007 IND -8.9%
2004 KC 33.1% x 2004 MIA -35.7% x 2008 HOU 24.6% x 2008 MIA -8.9%
2007 DAL 32.1% x 2006 OAK -35.5% x 2010 DEN 23.1% x 1996 NYJ -8.9%

By the way, to answer a common question: No, right now the Mike Scifres fake punt does not count in the San Diego special teams rating. I count fakes as runs or passes. This is probably an error -- Jim Schwartz, of all people, has specifically mentioned that I should change this -- and when I upgrade the special teams formulas in the offseason I'm going to look into doing that. But for now, the Chargers special teams don't get the credit for last night's excellent fake punt. Even with that credit, they would be the worst special teams we've ever measured by a very healthy margin

* * * * *

These are the Football Outsiders team efficiency ratings through 11 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. WEIGHTED DVOA represents an attempt to figure out how a team is playing right now, as opposed to over the season as a whole, by making recent games more important than earlier games. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.

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>


1 PHI 35.1% 1 35.4% 1 7-3 30.2% 2 -6.9% 7 -2.0% 25
2 PIT 34.2% 3 33.1% 2 7-3 14.0% 8 -17.5% 1 2.7% 10
3 NE 26.9% 4 27.7% 3 8-2 40.7% 1 16.8% 27 3.0% 9
4 GB 25.9% 5 25.4% 4 7-3 18.3% 6 -10.9% 4 -3.4% 28
5 TEN 24.2% 2 25.1% 5 5-5 7.7% 16 -10.1% 5 6.4% 2
6 NYG 19.0% 6 21.8% 6 6-4 10.2% 12 -14.3% 2 -5.5% 31
7 BAL 18.4% 8 20.4% 7 7-3 12.5% 10 -2.5% 9 3.3% 7
8 SD 15.8% 7 18.0% 8 5-5 19.3% 4 -11.2% 3 -14.7% 32
9 ATL 14.6% 10 13.0% 9 8-2 18.1% 7 4.9% 19 1.4% 13
10 NYJ 13.6% 12 10.9% 12 8-2 6.2% 17 -3.3% 8 4.1% 5
11 IND 12.0% 9 11.7% 11 6-4 18.6% 5 2.3% 14 -4.3% 30
12 MIA 11.6% 11 12.3% 10 5-5 12.5% 9 0.9% 13 0.0% 17
13 CLE 8.6% 14 10.1% 13 3-7 4.2% 18 -1.3% 10 3.2% 8
14 NO 8.2% 13 8.6% 14 7-3 9.2% 14 -1.3% 11 -2.3% 26
15 KC 3.3% 15 -0.4% 16 6-4 9.6% 13 6.0% 21 -0.4% 18
16 TB -0.3% 20 0.0% 15 7-3 7.8% 15 7.1% 23 -1.0% 22
17 HOU -1.5% 16 -0.4% 17 4-6 29.0% 3 29.7% 32 -0.8% 20
18 CIN -7.1% 18 -7.8% 20 2-8 -0.5% 23 3.6% 18 -2.9% 27
19 CHI -8.1% 24 -9.2% 22 7-3 -23.8% 30 -9.3% 6 6.4% 3
20 DET -8.6% 21 -7.7% 19 2-8 -6.3% 24 3.2% 16 0.9% 14
21 SF -8.7% 17 -5.0% 18 3-7 -9.1% 25 -0.1% 12 0.3% 16
22 JAC -10.4% 22 -8.4% 21 6-4 4.1% 19 19.5% 29 5.1% 4
23 WAS -10.8% 25 -9.4% 23 5-5 -0.4% 22 9.4% 24 -1.0% 21
24 DEN -15.0% 23 -14.4% 25 3-7 11.6% 11 23.1% 31 -3.5% 29
25 DAL -15.1% 27 -17.9% 29 3-7 0.5% 20 17.0% 28 1.4% 12
26 MIN -17.2% 26 -17.8% 28 3-7 -9.4% 26 6.1% 22 -1.7% 24
27 OAK -17.8% 19 -14.4% 24 5-5 -14.9% 29 2.4% 15 -0.6% 19
28 STL -18.0% 28 -16.7% 26 4-6 -11.4% 27 5.5% 20 -1.1% 23
29 BUF -18.4% 30 -17.6% 27 2-8 0.3% 21 20.2% 30 1.5% 11
30 SEA -22.5% 29 -26.6% 30 5-5 -13.8% 28 15.6% 26 7.0% 1
31 ARI -35.9% 31 -36.1% 31 3-7 -28.6% 31 10.7% 25 3.4% 6
32 CAR -45.4% 32 -45.5% 32 1-9 -42.5% 32 3.2% 17 0.4% 15
  • NON-ADJUSTED TOTAL DVOA does not include the adjustments for opponent strength or the adjustments for weather and altitude in special teams, and only penalizes offenses for lost fumbles rather than all fumbles.
  • 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).

1 PHI 35.1% 7-3 32.5% 8.2 1 4.6% 10 -6.3% 26 8.1% 7
2 PIT 34.2% 7-3 29.5% 7.2 3 8.7% 3 -5.0% 23 16.3% 20
3 NE 26.9% 8-2 18.9% 7.3 2 7.2% 5 2.7% 13 13.8% 16
4 GB 25.9% 7-3 28.1% 7.1 4 -3.5% 22 5.9% 9 16.2% 19
5 TEN 24.2% 5-5 17.7% 6.8 7 4.7% 9 2.3% 14 7.4% 5
6 NYG 19.0% 6-4 21.4% 6.9 6 -4.5% 23 2.0% 16 26.3% 28
7 BAL 18.4% 7-3 13.4% 6.9 5 2.4% 17 7.0% 6 5.8% 2
8 SD 15.8% 5-5 18.6% 6.3 10 -6.7% 26 -5.6% 25 16.5% 21
9 ATL 14.6% 8-2 12.0% 6.4 8 3.4% 13 -13.2% 31 6.3% 3
10 NYJ 13.6% 8-2 19.4% 6.0 11 3.1% 15 6.5% 7 5.1% 1
11 IND 12.0% 6-4 12.1% 5.9 12 3.8% 12 3.5% 11 9.7% 12
12 MIA 11.6% 5-5 9.6% 6.4 9 9.2% 2 0.7% 18 18.8% 24
13 CLE 8.6% 3-7 5.2% 5.6 14 10.1% 1 -1.1% 21 11.9% 15
14 NO 8.2% 7-3 14.5% 5.8 13 -11.8% 31 -1.3% 22 8.6% 9
15 KC 3.3% 6-4 11.5% 5.4 16 -7.1% 27 -5.6% 24 25.3% 27
16 TB -0.3% 7-3 6.4% 5.5 15 -9.5% 29 -0.1% 19 15.4% 18
17 HOU -1.5% 4-6 -2.4% 4.8 17 2.1% 18 12.8% 3 14.8% 17
18 CIN -7.1% 2-8 -10.5% 3.9 22 6.2% 7 16.5% 1 8.9% 10
19 CHI -8.1% 7-3 -0.3% 3.9 23 -8.1% 28 12.6% 4 22.3% 25
20 DET -8.6% 2-8 -7.8% 4.2 20 0.6% 20 6.5% 8 6.9% 4
21 SF -8.7% 3-7 -6.9% 4.5 18 -5.8% 25 -11.8% 30 31.1% 30
22 JAC -10.4% 6-4 -14.3% 4.3 19 4.9% 8 4.2% 10 18.8% 23
23 WAS -10.8% 5-5 -12.7% 4.1 21 8.1% 4 -0.8% 20 11.4% 13
24 DEN -15.0% 3-7 -16.7% 3.7 28 2.8% 16 -9.0% 27 22.9% 26
25 DAL -15.1% 3-7 -18.0% 3.9 24 3.1% 14 7.3% 5 28.8% 29
26 MIN -17.2% 3-7 -20.9% 3.9 25 4.4% 11 1.4% 17 9.3% 11
27 OAK -17.8% 5-5 -8.2% 3.3 30 -2.4% 21 2.9% 12 43.3% 32
28 STL -18.0% 4-6 -8.7% 3.7 27 -12.0% 32 -11.8% 29 11.6% 14
29 BUF -18.4% 2-8 -17.3% 3.5 29 6.6% 6 13.0% 2 7.9% 6
30 SEA -22.5% 5-5 -16.5% 3.8 26 -9.7% 30 -9.1% 28 31.3% 31
31 ARI -35.9% 3-7 -30.6% 2.6 31 -5.6% 24 -18.5% 32 16.6% 22
32 CAR -45.4% 1-9 -45.8% 1.4 32 1.1% 19 2.3% 15 8.2% 8


150 comments, Last at 28 Nov 2010, 2:21pm

1 Re: Week 11 DVOA Ratings

I like how 3 of the 4 worst Special Teams rankings are on teams in the top half of the league. Oh, and San Deigo, I'm really pulling for you to wind up with the worst ST ever.

2 Re: Week 11 DVOA Ratings

Yay! The Pats' D is back up to 27th, where it belongs! That one week at 28th was discouraging!!

81 Re: Week 11 DVOA Ratings

Yeah...though the last one struck me as the type of play that would happen about once in twenty in that situation. Pats kinda lucked out.

3 Re: Week 11 DVOA Ratings

I am curious about how you guys deal with injuries. I would think neither the Titans nor the Dolphins can be considered as dangerous with their QB injuries.

45 Re: Week 11 DVOA Ratings

The short answer is that DVOA doesn't consider injuries.

The slightly longer answers are that (1) while it's easy to separate which QB was in for each play, it's harder/impossible at other positions, and many of those injuries have a major effect as well. (2) The DVOA number reflects every play the team has run in the season, regardless of who the quarterback, fourth receiver, middle linebacker, or nickel back is.

Going forward, it's entirely reasonable to say that we don't expect the Titans to be as good with Rusty Smith as they were with Young or even Collins, but their DVOA is still going to include what Young and Collins did.

56 Re: Week 11 DVOA Ratings

That all being said, I suspect there are ways that DVOA could be improved simply by noting who started the season and deducting value based on the position removed. While it's hard to separate out the value of certain players in a system, even reducing it by 1/11th to replacement value for each injury would likely result in a more accurate predictive quality.

71 Re: Week 11 DVOA Ratings

DVOA is, fundamentally, intended to be descriptive, not predictive. Any descriptive system worth its salt will also be somewhat predictive, but tweaking the first to improve the second goes against what DVOA is measuring.

77 Re: Week 11 DVOA Ratings

Then what's the use of having cumulative values for DVOA? Heck, what's the use of having it rate your performance based on the value of other teams? If another team loses their star QB and your defense plays excellently against...Jim Sorgi, how is that a descriptive value? According to DVOA, that performance would be rated way too high; we all know Sorgi isn't great compared to whoever he replaced, but DVOA looks at it as 'the team that's been awesome for 8 weeks getting crushed, therefore'.

Even as a descriptive value DVOA needs some value of prediction or even descriptive adjustments. One of the most common comments we see is "we know that team X is bad but DVOA hasn't caught up to that yet' because of injury. Factoring injury would be much more useful from a descriptive and predictive value.

79 Re: Week 11 DVOA Ratings

This is very difficult to do objectively.

I mean in 2001 didn't we *know* that a sophomore 6th round pick QB would be worst than Drew Bledsoe? And in 2006 didn't we *know* that a journeyman QB on his 4th team would be worse than McNabb?

84 Re: Week 11 DVOA Ratings

Just this season: Kolb gets hurt, Vick comes in. The team overall plays better offensively, and production from the QB position improves, and the Vick Special with RB performance appears.

Why would Philly be considered weaker when the team begins playing better?

94 Re: Week 11 DVOA Ratings

Just because there are outliers to the rule doesn't mean the rule isn't otherwise reasonable. That's what, two examples in 10 years? And both are easily correctable after you see that QB play.

Mostly, I'm talking about opponent adjustments not based on the historical data of the opponent but when you played that opponent. In a very few cases this would result in being incorrect (adjusting downwards for injury) but most of the time this would be entirely correct. DVOA as far as the value of how that team played wouldn't change - or rather, VOA wouldn't change. DVOA might based on the other team.

But this allows for some descriptive value that is more than 'they're boosted by beating team X which at the time sucked but got players A and B back' or 'they played team Y closely but since then Y has sucked, so their rating has gone down because of it'.

Like I said, this is the sort of thing that you can adjust on the fly as well. If the negative position X doesn't work because the backup is legitimately better, just ignore that adjustment. Problem solved.

107 Re: Week 11 DVOA Ratings

"And both are easily correctable after you see that QB play."

These are subjective elements. You're advocating number fudging. DVOA is an objective statistical metric. If the gambler in you is looking to get predictive using DVOA I suggest treating TEN and MIA as stay-aways: we probably do not have enough information to make an accurate assessment.

118 Re: Week 11 DVOA Ratings

No, I'm really not.

DVOA is objective to a point; while it treats the numbers as objective once they've been computed it certainly subjectively views them beforehand (as they're done by Aaron et al). What I'm suggesting is to add another conversion here which can be modified by hand (as, for example, many of the interceptions as hail-marys are) depending on what the actual results are. This really is not different.

If I were programming it, I would take the modification to the team as follows:
If a new player is replacing an old player and we have not seen them play before, assume they are at replacement level and reduce the team's rating (for purposes of other teams) by that position's value. (which for the first attempt would be equivalent).
If after that game DVOA indicates that that player played at the same level of play as the prior player, make no adjustments to that team's DVOA against opponents.
Otherwise continue with the negation of the overall value of DVOA for purposes of giving other teams value.

Again, the descriptive nature of the system fails for injuries because it cannot distinguish between a team getting their starter back and a team not getting their starter back. As an example, how much worse would Pittsburgh's opponents look if DVOA knew that they had played against Ben's backups and lost in weeks 1-4? How much better does it look when NE plays against Pitt now and the weight of weeks 1-4 aren't dragging down Pitt's offensive value? We view NE beating Pitt as a huge win for the NE defense, but DVOA didn't care as much because Pitt doesn't look that strong offensively - and part of the reason is because of weeks 1-4.

This really has nothing to do with betting and everything to do with giving teams the descriptive value they should have right now.

122 Re: Week 11 DVOA Ratings

OK - fine, but how do you account for something like Kolb/Vick? Kolb played above replacement level (15.7% DVOA) and, until Vick came back from injury and opponent adjustments were at full strength, DVOA saw them as having performed at a similar level. Now Vick is playing at much, much higher level (according to DVOA) than Kolb ever did and some of Kolb's opponents have declined, so his opponent adjustments are weaker.

What should Aaron have done in the week leading up to the Indy game? How about the next week? Before the Indy game, Kolb looked essentially the same as Vick as far as DVOA was concerned. After the Indy game, Vick had surpassed him significantly, but only on the strength of one game (and remember that even at that point, many people, including FO writers weren't sold on Vick as "actually" being an upgrade over Kolb.) What should Aaron have done about it, in your estimation? Because I don't want him starting to get his opinions all over what the Qb changes mean for the Eagles.

And don't give me a "well, that's a tough rare case" because you have similar situations in Cleveland with Delhomme, Wallace (who DVOA likes) and McCoy (who DVOA is not crazy about) and Carolina. That's 3 teams this year, at least, that have totally muddled QB situations where the "good QB" and the "replacement level" QB are not clear and it took weeks for the situation to even begin to sort itself out.

137 Re: Week 11 DVOA Ratings

With Kolb/Vick and Delhomme/whoever, you do what I mentioned - you play them at the DVOA at the time, no modifications. That still isn't perfectly accurate, but it's likely closer to accurate than not. Retroactively changing opponent adjustments isn't hard based on future information, and is done already to some nth degree (which Aaron hasn't mentioned).

Specifically in the case of the Cleveland situation, chances are all three QBs are around the same level anyway and are all close to 'replacement level', so regressing to replacement level isn't going to be much of a change.

It would, however, require a bit more tweaking to do all 22 starting players. I'd probably start with doing a couple easy ones like QB and RB and go from there. Maybe not even RB; something like LT. Then see if it makes the valuations more or less predictive or accurate and proceed accordingly. I may be wrong and it might be the case that this makes things less predictive, but anecdotally there should be a difference in performance that is detectable when a starting player comes in over a backup and vice versa.

138 Re: Week 11 DVOA Ratings

The 3 QB's in Cleveland have wildly different DVOA's and the one you probably think has the worst has the best and the one you think has the best (maybe) has the worst - that's actually why I mentioned: DVOA think Seneca Wallace played well (17% DVOA) and Delhomme stinks. It thinks McCoy is significantly worse than Wallace, but significantly better than Delhomme. Now, I forget: who gets the baseline "replacement level" designation and for what reason and to what end? That's what makes it such a potent example: it means Cleveland's DVOA would have jumped around really randomly throughout the season.

Same thing for Vick/Kolb: the Eagles' DVOA would have randomly jumped around based on some completely artificial "replacement level" until the numbers settled down enough to be accurate. And then add in that their opponents DVOA will be randomly flung around based on a completely fantasy-based notion of "replacement level" and the whole creates a mess for almost no reason. I think made up numbers are far more worthless than just having DVOA describe what happened on the field and let the chips fall where they may: there's no satisfying way to speculate on what the Eagles' DVOA should be working from the assumption that Vick would have been better than Kolb in 4 games in which Vick did not play. It's even harder to make those speculations earlier in the season when the sample size is tiny (back when DVOA thought they were playing at the same level) - like say, after 4 games when the Steelers QB hasn't played yet.

I;m all for FO coming up with a better way to deal with injuries, but giving completely made up numerical designations to compensate is a terrible idea.

140 Re: Week 11 DVOA Ratings

Sorry, you're still not understanding what I'm suggesting.

I'm not saying that you lower the DVOA of the actual team. I'm suggesting that when you factor in opponent adjustments, consider changing the opponent adjustments down/up depending on injury to reflect what the actual state is.

This is already being done in an arbitrary way via weighted DVOA. It doesn't do it precisely, but it still does the same thing. If you have a problem with made up numerical designations, you should have a similar problem with weighted DVOA.

Again, the problem is that DVOA doesn't know about injuries or people being suspended. So in one example, Pittsburgh's offense for the first 4 weeks looks worse than it does for the rest of the season. Their DVOA is lower than it would be if (presumably) Ben had played the first four weeks. Thus, if a team does well against that offense they don't get as much credit as they would if Ben had played. If a team does poorly against that offense they look worse.

Similarly with injuries; if a team obliterates Seattle (say, the Giants) they look great against the mediocre Seattle offense. If DVOA knew that the starting QB went out with an injury, that makes the Giant D not look quite as earthshattering.

As to what the 'jumps' would entail - you're making this out to be far greater than I'm suggesting. I mentioned at the beginning that in lieu of actually figuring out the precise values for each position you value each the same (1/11th of the total DVOA of the team) and then replace the value for replacement player. If you like, you can have two orders; one is a 0 DVOA (which actually might be an improvement for some teams) and then a replacement value (which I think is like -15%).

So let's say a QB is lost from a 20% offense DVOA team - a pretty good offense. Using this system, they'd go to a 0% QB + 10 20% other guys, making their effective DVOA for opponent adjustments valued at 18.18%. This isn't a huge drop, and it wouldn't make DVOA jump around insanely. If instead we found out that they were actually pretty awesome as a QB, you could rate them back to that 20%. If you find out they sucked a lot, you could rate them a lot lower.

Right now DVOA does this anyway; if it turns out that you lost to a team and looked bad doing so but that team rocks hard for the next 10 weeks the value of that first game gets retroactively changed and jumps around anyway. This is just another jumping around point. Future knowledge changes the value of past results in DVOA all the time. This isn't anything new.

143 Re: Week 11 DVOA Ratings

Ok - I understand slightly better, but I still don't see the advantage of making up a number to represent the change, even if the made-up number only matters for a week or two. DVOA is going to take into accunt strength changes if they persist long enough anyway.

You're still working from a demonstratably false premise: it is easy to tell the value of a player that goes out with an injury. Go back to Cleveland: you have to arbitrarily assigne a DVOA to the QB changes and the actual value of Delhomme, Wallace and McCoy is extremely unclear, especially early on... And in your Giants/Seattle example, Whitehurst played terribly, but you can't say for certain that he looked bad because of the Giants or the Giants looked good because of him - the sample size is tiny. And I don't think it makes sense to penalize the Giants defensive rating one way or the based on the ASSUMPTION that Whitehurst stinks. Why not? Because the assumptions about Delhomme, Wallace and McCoy proved to be wrong. The assumptions about Vick vs. Kolb proved to be wrong (for weeks, many FO writers doubted Vick), in past seasons Collins vs. Young and Johnson vs. Flutie proved to be wildly wrong. There's no advantage to making immediate assumptions.

144 Re: Week 11 DVOA Ratings

One more reason this is not a good idea: what do you propose DVOA does about Kolb's numbers in the Atlanta game? Kolb has proven to be overall far inferior to Vick (15.7% DVOA vs. 35.2% DVOA)... but he absolutely annihilated Atlanta: 23 of 29, 326 yards and 3 td's. In your scenario, Atlanta should somehow be penalized because they didn't face Vick (and the Eagles rewarded.) But that assumption seems crazy because it is all but impossible to play than Kolb did that day. I think you have to let the DVOA stand as is. Anything else is artificial and speculative. And keep in mind, back in week 7, the idea that Vick was decisively better than Kolb was far from settled - so making something up is an even faultier idea...

145 Re: Week 11 DVOA Ratings

But you're still assuming that these things don't get changed in the future. I'm saying change it at the time, and then go back and change things as needed depending on future information. This is the basis on the D in DVOA, afterall - defense-adjusted, and not just for the week in time they played them.

So yes, Atlanta should be penalized some because they didn't face Vick. Why is that surprising given what we know about Vick? And again, that's one small example that is exceptional; very rarely do teams have a backup QB play significantly better than the starting QB.

DVOA is artificial and speculative. Weighted DVOA is the culmination of that. If you hhave a problem with being artificial, you have a problem with the concept of DVOA.

146 Re: Week 11 DVOA Ratings

It's easier to assume the value of an injury as negative in certain positions. And while you might not be able to quantify a specific position's value, you can certainly assume (like much of DVOA does) that a backup will perform worse than a starter.

If Whitehurst had demonstrated that he could play as well as Hass that would be one thing, but it's been demonstrated that he doesn't. So instead, the Seahawks look slightly worse on offense than they normally do, teams that play them look slightly worse when they play Hass because of the Whitehurst effect, and then teams that play those teams get affected too.

Right now the Giants get the assumption that they're playiing against a Hass-led Seahawks. As far as DVOA is concerned, Hass played onne of the worst performances in the whole season by a QB, making the Giants look amazing. Is that more accurate than making an adjustment?

Again, penalize the Giants based on the notion that the Seahawks with Whitehurst are worse than they are with Hass. If it's not true based on other data, remove the penalty aand things are exactly the same. If he turns out to be actually better, improve the Giants. This isn't hard, and the demonstrably false issue works itself out through the season with new information - just like all the rest of DVOA works. These adjustments happen all the time - that's the basis of DVOA, FEI and S&P+. You're assuming that they don't and that you can't go back and change things; I'm not suggesting that at all. I'm suggesting making discrete changes to week-by-week adjustments to make that week more accurate.

Actually, that would likely be a better way to put it. Instead of applying the value of the opponent based on the current information only, base the value of the opponent not only on their current value but also their injury status at the time they played.

4 Re: Week 11 DVOA Ratings

I just had a quick question: when a player fumbles on an interception return (like Asante Sameul vs. NYG this week) does that get counted against the defensive DVOA? How does DVOA deal with that?

7 Re: Week 11 DVOA Ratings

In the non-adjusted numbers, recovering a fumble right after you've turned the ball over means that you no longer get penalized for the turnover. In the adjusted numbers, I don't even count it for either the offense or defense. Sort of just a random play.

5 Re: Week 11 DVOA Ratings

Tennessee is ranked really high in DVOA. Hard to believe that they remain so much higher than teams like the Falcons or Jets. I'll be the first to admit that the Falcons and Jets have gotten a little lucky and aren't the best teams (even with the best records). But regardless, when the cards are down, those teams seem to be able to rise to the occasion and persevere. I guess they just have to learn to apply that intensity and focus to the rest of the game and not let themselves get into do or die situations quite as much.

41 Re: Week 11 DVOA Ratings

I think what you meant to say was "Is computer drnuk?"

The answer is, maybe tipsy. It certainly cannot tell what the QB position will produce in the coming weeks. I certainly don't look at them and think "elite team."

87 Re: Week 11 DVOA Ratings

"...when the cards are down, those teams seem to be able to rise to the occasion and persevere."

The problem with that is that a team that doesn't let the cards get down in the first place is much better than one who has to battle back. The concept of "comebacks" usually seems to make people (especially sportscasters) think a team is better than it is.

This is especially true when applied specifically to QBs. Not that there's no value in the experience/cool/poise/moxie/etc to come back, just that it's better not to have to. Yet, sometimes you see more credit given to the team or QB that's able to come back from a deficit, than you do to one who doesn't have to.

The same thing applies to teams that "know how to win close games".

119 Re: Week 11 DVOA Ratings

You mean John Elway wasn't the bestest ever? But he had all those comebacks?

I never understood why a failure to score in the first half should be rewarded.

133 Re: Week 11 DVOA Ratings

Not necessarily a failure to score early.

Could also be your D and/or ST sucks unimaginably. When the other team never has to punt, or when they return two punts for TDs in a single game, well, hard to blame the QB.

8 Re: Week 11 DVOA Ratings

What I find interesting here is the diversity of teams on the "Best/Worst DVOA" lists with the exception of the worst defense list. I'm familiar with the Football Outsiders truism that offense is more consistent, year to year, than defense. That seems to fly in the face of what we're looking at on this list.

Maybe at the extremes, defenses are more likely to be consistently bad than offenses are? That is, maybe it's easier to scout offensive players, or to scheme around bad offensive players?

9 Re: Week 11 DVOA Ratings

Are the DVOA Ratings a good guide to use when betting on games? Specifically the POINT SPREAD and MONEYLINE? Talk to me people...

40 Re: Week 11 DVOA Ratings

sorry, this is meant as a reply to the below betting comment

it all depends on what model you use, and certainly over short periods of time you will find anomalies. this was a discussion in the dvoa comments a few weeks ago. Some have used it to varying degrees of success. I have used it in the past, and certain years have been positive and others, negative.

Overall, these projections have been most beneficial in the early season when FO properly pegs teams that are rising and or/fading. I'm a little more conservative than the average bettor, so i'm looking specifically for value against the spread. Does DVOA like a team siginificantly more or less than the general public. Over the last two weeks, DVOA has properly analyzed Philly according to my betting model, although their higher than Vegas rating on Tennessee has not worked.

10 Re: Week 11 DVOA Ratings

The Chargers are the only team with a top 5 offense and defense. If their special teams units can only be "bad" instead of "historically awful" the rest of the way, you gotta think they have a chance to be the best team going into the playoffs.

13 Re: Week 11 DVOA Ratings

yeah, that's been pointed out a lot and I think this game last night was a perfect example of how they should really just consider going for it on every 4th down. Even if that fake punt had fizzled, would that have mattered? It probably just would have been blocked, anyway...

48 Re: Week 11 DVOA Ratings

Well they would be Green Bay then.

San Diego Off: 19.3% (4th) Def: -11.2% (3rd) ST: -14.7% (32nd)
Green Bay Off: 18.3% (6th) Def: -10.9% (4th) ST: -3.4% (28th)

SD is all of 1.3% better than GB on Offense and Defense. GB has bad special teams play but not horrendous. Not a lot of difference there. So yeah, I do think both SD and GB have a chance to be the best teams going into the playoffs. :) I'll point out that GB has been improving in all 3 phases of the game the last few weeks as well.

Week 11: 18.3 (6th), -10.9 (4th), -3.4 (28th)
Week 10: 16.2 (7th), -10.9 (5th), -4.3 (30th)
Week 09: 14.5 (8th), -10.8 (4th), -4.3 (28th)
Week 08: 12.7 (8th), -7.0 (6th), -5.6 (30th)

That's a nice trend and one that if it continues could actually get this team up to great as opposed to just very good.

I've talked about the special teams before because well they have sucked for years for the Packers and yes I'm excited that if Masthay keeps punting the way he has the last 4 weeks that he'll actually end up around the 12th best punter in the league. Having a just above average punter is quite exciting for the Packers. They may also see a small improvement in kick returns since Sam Shields continually proves that he can correct his mistakes and just keep getting better so him on kick off returns could pay off, if nothing else it might get Jordy Nelson to stop being so hesitant like he has been since the Detroit game. But just getting the punting up to league average is a big help.

But really Philly, Pitt, GB, and SD are the teams that I currently like in the playoffs and it's for the same reason, they are the 4 teams with a top 10 offense and a top 10 defense. I really haven't looked to see if that predicts success in the playoffs, but it just feels like it gives you the best chance or at least in a year where we aren't looking at any historically great defenses and only NE and Philly seem to have great offenses (NE historically so)

149 Re: Week 11 DVOA Ratings

I've talked about the special teams before ...

So I want to start tracking some of the various special teams so I'm replying to myself. Besides I have to feed the narcissism. :-)

Packers Special Teams Breakdown
FG/XP........: 0.5 (12th) [ -12.4 to 8.8 ] -0.25
Kick...........: -1.7 (21st) [ -14.9 to 13.4 ] 1.8
Kick Return: -8.5 (27th) [-10.9 to 14.8 ] -1.2
Punt...........: -2.1 (21st) [-33.7 to 8.9 ] -1.2
Punt Return: -0.5 (23rd) [ -7.0 to 20.3 ] 1.05

So that category, Packers value (Packers Rank) [ League Range ] League Median.

I used Median (which in this case is actually doing the average of the 15th and 16th ordinal rankings) over average here as it was quick and dirty and did seem to represent where the middle of the pack was. I just wanted an bit of info for how much worse, or in one case better, than "average" they were.

If the punting continues on the pace it has been it's got a chance to just not be a worry and climb to mediocrity. The FG kicking is already at mediocrity. I thought the punt returns were closer to my gold standard of, yes mediocrity than they were but I can live with it, as long as the returners catch the ball.

So it's that wild card of kick returns. Shields didn't really have too many chances and that is good, the fewer kick off we return the better, but I'm not sold on him yet. But I'm excited that we might get 2 of the 5 phases of special teams up to mediocre. The life of a Packers fan. :)

51 Re: Week 11 DVOA Ratings

First have Raiders stadninhg n way in afc West. Raiders going to beat chagers again for the sweep. Raiders will be 11-5 or 10-6. Only question is game vs Cotls.
When Chargers lose to Raiders that will eb at least 6th chargers loss. Chargers not winning divisoon if 10-6 and Raiders 10-6 or 11-5.

Then have to dela with wild cards. Jets or Pates getting oen. Good channce Ravens or steelers get other one. So when relally look at things it looking like Charegsr on outside of playoff door

59 Re: Week 11 DVOA Ratings

Why do you continue to post and embarrass yourself? Learn to read and write, and then come back. Thanks.

66 Re: Week 11 DVOA Ratings

How many Oorang Indians players can you name? There's no-one on this site as football-educated as Raiderjoe. He just has a couple of blind spots: the Raiders, and typographical inconsistency. Learn to forgive them: you'll be rewarded if you do.

89 Re: Week 11 DVOA Ratings

There is. Haven't you heard? The next Sierra Nevada Pale Ale you crack open will have a dead mouse in it. Swear to God. It happened to a friend of a friend of mine. Well, not his first, I think it was like his 20th. All in the same night. And he's not sure if it was a mouse or if he was drunk off his ass.

Still, it's a curse and I wouldn't mess with it. Better safe than sorry.

oops, I mena better saef than srooy.