Writers of Pro Football Prospectus 2008

Most Recent FO Features

SimBla15.jpg

» OFI: Don't Make Saban Angry

Notre Dame and Baylor entered the one-loss group in what is shaping up to be an extremely tight race for playoff consideration.

15 Sep 2009

Week 1 DVOA Ratings

by Aaron Schatz

You love them when your team is high! You hate them when your team is low! DVOA ratings are back in effect for 2009!

This year's DVOA ratings feature the return of DAVE, our rating that combines the preseason projection with the results of early games to give us a better prediction of how each team will rank at the end of the year. For those who don't know the story, this metric is called DAVE as a reaction to criticism that our stats are too much alphabet soup. I mean, who can argue with a guy named Dave? (Technically, it stands for DVOA Adjusted for Variation Early.) In this week's DAVE ratings, the preseason projection counts for 90 percent, and the current VOA counts for 10 percent. Also, a handful of projections have been tweaked to reflect Week 1 injuries (in particular, Chicago defense and Philadelphia offense).

With all the crazy close finishes this week, it's no surprise that the team with the lower VOA rating won three different games in Week 1. What's interesting is that the game where VOA disagrees the most with actual results is not one of the prime time near-upsets, but rather the Colts' 14-12 win over the Jaguars. Jacksonville, 11th, is the highest-ranked 0-1 team.

The first week of pages always takes time to set up -- MNF goes super late, all the Excel sheets must be run for the first time, technical bugs crop up -- but we'll have all the stats pages updated with 2009 information as soon as we can. (Same goes for an updated playoff odds page.) Offensive line, defensive line, and FO Premium will not be updated until after Week 2.

UPDATE WEDNESDAY 11:30am: 2009 stats pages are now updated, along with playoff odds.

* * * * *

These are the Football Outsiders team efficiency ratings through one week of 2009, 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 VOA are adjusted to consider all fumbles, kept or lost, as equal value. SPECIAL TEAMS VOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season.

There are no opponent adjustments in VOA until the fourth week of the season, which is why it is VOA right now rather than DVOA. 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 VOA to get a more accurate forecast of how a team will play the rest of the season. Right now, the preseason projection makes up 90 percent of DAVE.

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
TOTAL
DAVE
RANK W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
1 PHI 102.3% 14.1% 11 1-0 -5.9% 23 -94.3% 1 13.9% 3
2 NO 79.0% 0.6% 14 1-0 56.4% 2 -53.4% 2 -30.9% 32
3 NYJ 59.5% -12.2% 23 1-0 20.8% 8 -33.1% 5 5.5% 6
4 DAL 59.4% 4.9% 13 1-0 92.8% 1 33.6% 30 0.3% 14
5 GB 53.9% -0.7% 15 1-0 5.3% 11 -49.6% 3 -1.0% 18
6 MIN 41.2% 19.8% 5 1-0 25.3% 7 -17.0% 8 -1.1% 19
7 SEA 36.5% 15.4% 8 1-0 17.7% 9 -15.1% 9 3.7% 9
8 ATL 33.0% -12.1% 22 1-0 15.2% 10 -27.3% 6 -9.5% 26
9 BAL 32.9% 16.5% 7 1-0 40.5% 4 -6.5% 12 -14.0% 29
10 PIT 23.4% 21.0% 4 1-0 1.3% 18 -5.8% 13 16.3% 2
11 JAC 10.6% 16.8% 6 0-1 4.5% 14 -3.3% 16 2.8% 11
12 CIN 5.1% -4.6% 17 0-1 3.1% 15 -5.5% 14 -3.5% 22
13 SD 5.0% 33.6% 1 1-0 4.7% 12 3.1% 21 3.4% 10
14 NYG 4.7% 12.4% 12 1-0 -5.4% 22 -9.5% 10 0.6% 12
15 WAS 4.3% -6.3% 19 0-1 1.3% 17 -3.6% 15 -0.6% 17
16 DEN 3.6% -22.0% 31 1-0 -3.5% 21 -7.4% 11 -0.3% 16
TEAM TOTAL
DVOA
TOTAL
DAVE
RANK W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
17 NE 2.6% 23.3% 3 1-0 27.7% 6 31.3% 28 6.2% 5
18 ARI 1.3% -18.1% 28 0-1 -36.9% 28 -33.8% 4 4.5% 7
19 BUF 0.0% -25.8% 32 0-1 31.3% 5 21.2% 26 -10.1% 27
20 IND -0.2% 29.7% 2 1-0 2.6% 16 -2.2% 19 -5.0% 25
21 SF -1.5% -20.5% 30 1-0 -33.4% 27 -25.2% 7 6.7% 4
22 OAK -5.7% -11.8% 20 0-1 4.6% 13 5.6% 22 -4.6% 24
23 TEN -16.1% 14.6% 10 0-1 0.6% 19 -3.1% 17 -19.8% 31
24 KC -33.7% -5.7% 18 0-1 -1.3% 20 32.3% 29 -0.1% 15
25 CHI -40.8% 14.9% 9 0-1 -47.0% 30 -2.5% 18 3.7% 8
26 CLE -41.1% -13.7% 24 0-1 -22.0% 24 17.3% 25 -1.7% 21
27 TB -44.6% -15.2% 27 0-1 43.0% 3 75.7% 32 -11.9% 28
28 MIA -46.4% -14.4% 25 0-1 -26.5% 25 15.7% 23 -4.1% 23
29 HOU -55.3% -14.7% 26 0-1 -29.6% 26 24.3% 27 -1.5% 20
30 STL -60.5% -2.2% 16 0-1 -43.6% 29 17.3% 24 0.3% 13
31 DET -100.2% -18.9% 29 0-1 -60.3% 31 62.7% 31 22.7% 1
32 CAR -116.4% -12.1% 21 0-1 -96.7% 32 1.6% 20 -18.1% 30

Posted by: Aaron Schatz on 15 Sep 2009

159 comments, Last at 20 Sep 2009, 8:07pm by Jetspete

Comments

1
by Will Allen (not verified) :: Tue, 09/15/2009 - 5:03pm

Minnesota is clearly ranked too high because the Allen-o-meter has detected that they will yield 30 more punt returns for touchdowns this season. Allen-o-meter rulz, DVOA sux.

52
by billsfan :: Tue, 09/15/2009 - 10:18pm

Well, the Bills set the bar pretty low for ST DVOA.

That McKelvin fumble must have had a huge effect--Gostkowski honked an easy FG, and the Pats are still 5th in special teams.

(I also like the Eagles)

57
by mm (not verified) :: Tue, 09/15/2009 - 10:48pm

I must point out that, for 1 week at least, the Saints set the standard for poor special teams play.

2nd in offense VOA, 2nd in defense VOA, 32nd in special teams play!

89
by Theo :: Wed, 09/16/2009 - 9:59am

That Bush guy is one explosive player, isn't he.

5
by Temo :: Tue, 09/15/2009 - 5:13pm

Apparantly DVOA thought the Giants-Skins game was much closer than everyone else thought it was. Also, Philly offense is surprisingly pedestrian. Wonder what their VOA at the time of McNabb's injury was?

6
by BigDerf :: Tue, 09/15/2009 - 5:14pm

Agreed. The way Washington played I figured their DVOA would come out really poor but it didn't. Heck even Jason "Checkdown" Campbells offense had higher DVOA than the Giants offense.

Also... Giants had better Special Teams DVOA than the Redskins? Is that fake field goal getting applied to offense?

8
by C (not verified) :: Tue, 09/15/2009 - 5:24pm

Be careful, you can't call Jason Campbell "checkdown". The FO guys like him, it's real hard and takes a lot of brain power to complete check downs, I mean, he could go on forever like that...

Drew Brees threw 6 TD's in one game and Jason Campbell threw 13 TDs in a 16 game season last year, but why should we talk about the upside of a QB? Let's just talk about how he didn't make mistakes last year! It's not his fault, he has no RB, no Line, no WR's, really, he is good, it's just everybody around him that prevents him from success. Where have I ever heard that before?

Let's also not count his sack/fumble that got picked up for a touchdown either. I mean, who wouda thunk you are supposed to step up in the pocket on a freaking 9 step drop when you have 5 freaking yards in front of you and two strong pass rushers on the outside.

I hope the Redskins fans DON'T believe me, and are very happy with their quarterback. If they actually listened to me they'd dump him and be better off. Oh well.

17
by Ashley Tate (not verified) :: Tue, 09/15/2009 - 6:06pm

@c: Please pay attention, I'm only going to post this once!

There were definitely extended periods when Washington appeared to be dominated by NY. But in the end there were two very similar plays with very different outcomes that could have completely flipped the game result: When Eli Manning was sacked and fumbled in the 2nd quarter the ball was buried under a pile of players and Washington recovered for no gain. When Campbell was sacked and fumbled just 4 plays later the ball squirted into the open field where it was picked up and returned for a touchdown.

Somehow all of the Washington Post reporters neglected to mention this salient fact in their columns yesterday and in their "news" write-up today. The only football site I've seen that mentioned Manning's fumble so far was this one--in the extra points column today.

26
by BigDerf :: Tue, 09/15/2009 - 6:55pm

It has to do with the plays on which the fumbles happened. One the Eli play the whole Giants o-line was collapsing around him and he was smushed between multiple players and fumbled with the ball pulled down.

Campbell had plenty of room to step up into the pocket but had the ball knocked out of his hand by the Defensive End coming around from his blindside. Quarterbacks can't let that happen when there is room to step up. Campbell didn't feel the pressure and that and his generally Checkdowniness are why people are so down on him today.

Edit - This is in response to the lack of press about the Eli fumble vs. the Campbell fumble. I'm fully aware that DVOA watched for things like fumble luck.

104
by Mark S. (not verified) :: Wed, 09/16/2009 - 12:43pm

Yeah, I can definitely see why DVOA thought the game was close (and IMO it probably was in fact closer than it felt), and if Osi doesn't recover that fumble and run it in the outcome may have indeed been very different. The Giants did "seem" to play better, as most observers and the Audibles guys noted, but if not for that play we might be talking about all the missed Giants opportunities on short yardage.

That said, the reason it didn't feel as close as DVOA says it was is probably because the Redskins' TDs came on a fake FG and in almost-garbage time down by 2 scores - and after the first 20 minutes or so they never had the ball with the opportunity to take the lead. After the opening Portis run for 34, the 'Skins rushing attack did nothing, while the Giants actually ran the ball effectively aside from the "* and 1" situations. Not saying those plays should be discounted - just that that's why the Giants seemed to be more in control of the game.

28
by BucNasty :: Tue, 09/15/2009 - 7:03pm

I hope the Redskins fans DON'T believe me, and are very happy with their quarterback. If they actually listened to me they'd dump him and be better off. Oh well.

I know. I can't believe some front offices are so thick and arrogant that they ignore C's advice. Can we give this man a team, already? He clearly knows best.

10
by C (not verified) :: Tue, 09/15/2009 - 5:29pm

Be careful, you can't call Jason Campbell "checkdown". The FO guys like him, it's real hard and takes a lot of brain power to complete check downs, I mean, he could go on forever like that...

Drew Brees threw 6 TD's in one game and Jason Campbell threw 13 TDs in a 16 game season last year, but why should we talk about the upside of a QB? Let's just talk about how he didn't make mistakes last year! It's not his fault, he has no RB, no Line, no WR's, really, he is good, it's just everybody around him that prevents him from success. Where have I ever heard that before?

Let's also not count his sack/fumble that got picked up for a touchdown either. I mean, who wouda thunk you are supposed to step up in the pocket on a freaking 9 step drop when you have 5 freaking yards in front of you and two strong pass rushers on the outside.

I hope the Redskins fans DON'T believe me, and are very happy with their quarterback. If they actually listened to me they'd dump him and be better off. Oh well.

137
by Wanker79 :: Fri, 09/18/2009 - 12:37pm

Chris, you may be able to change your name but your irrational hatred for Jason Campbell, the childish way you go about defending your opinion, and above all else the cockiness of your posts, will give you away every single time. So why bother trying to hide? It'd be like RaiderJoe changing his name.

33
by Aaron Schatz :: Tue, 09/15/2009 - 7:20pm

Right now, fake FGs and punts are applied to offense, yes. I know that's something I should probably go change in the future, although it requires a lot of manual work. I also wonder if it is worth not counting those plays, as they are fairly "non-predictive."

9
by VarlosZ :: Tue, 09/15/2009 - 5:26pm

I'm surprised that DVOA sees this game as being played to a draw, as the Giants moved the ball much more effectively throughout the game. Aside from Washington's almost garbage-time 4th Quarter TD drive to pull to within 6, six of the game's longest seven drives belonged to New York. Interceptions were even, and Washington fumbled three times to New York's one. 3rd/4th Down conversion rates were almost identical. Etc.

What's up with that?

http://espn.go.com/nfl/drivechart?gameId=290913019

14
by JasonK :: Tue, 09/15/2009 - 5:47pm

The Giants had 2 fumbles-- besides the one Eli lost (which was caused by an un-flagged facemask), he fumbled a snap on their first drive that Seubert recovered.

I expect that the Giants offensive VOA is held back by their very poor effectiveness in power situations (of 7 runs where they needed 1 or 2 yards to get a first down, they converted only twice), and by the fact that many of their early drives were characterized by ineffective plays on 1st & 2nd followed by a nice 3rd down pass.

15
by Temo :: Tue, 09/15/2009 - 5:52pm

If you look at it from the viewpoint of conventional stats, Cambell had an identical QB rating to Eli (making up for the small loss of yards per attempt with a higher completion %), and Redskins RBs averaged 4.0 yds/carry to the Giants' 3.3 yards per carry. So there's the offensive difference.

I think it may have been the Giants' overall lack of run game that held down their offensive VOA (31 attempts at a 3.3 yards per carry clip is pretty bad), plus the Redskins' drive in (just about) garbage time apparently wasn't' considered garbage time.

70
by navin :: Wed, 09/16/2009 - 1:22am

Well it's not a garbage time drive because if the Giants fumble the kickoff and the Redskins recover and score a touchdown, the Redskins win the game. This is just like the Bills-Pats game.

Aaron, are timeouts accounted for when determining the importance of a score? Scoring with one minute left and three timeouts is a lot better than scoring with no timeouts left. I wonder if there's any work done on that.

75
by Jerry :: Wed, 09/16/2009 - 2:03am

By looking at timeouts remaining, I think that you're looking for much finer distinctions in assigning value than the DVOA model provides.

2
by TBW (not verified) :: Tue, 09/15/2009 - 5:07pm

Wow, HUGE game this weekend #1 vs #2 ! Has there ever been a game with a combined DVOA > 180% like this week's NO-Philly game ?

4
by Temo :: Tue, 09/15/2009 - 5:13pm

VOA, not DVOA. Makes a big difference. I think the record for highest combined DVOA was the NE vs. Indy game in 2007.

65
by Bobman :: Wed, 09/16/2009 - 12:06am

Ah yes, the game of the day/week/month/season/decade/century/millennium/geologic epoch. That is, until the next one....

3
by Jeremy Billones :: Tue, 09/15/2009 - 5:08pm

To the naked eye, Special teams are being given full weight in VOA rather than 1/3.

129
by TGT2 (not verified) :: Thu, 09/17/2009 - 4:12pm

I believe the 1/3 is factored in before the numbers get to us. By the end of the season, you'll rarely see a ST with DVOA in the double digits (positive or negative), but there will normally be offenses and defenses at +/-25%

132
by tuluse :: Thu, 09/17/2009 - 7:03pm

I'm pretty sure special teams are 1/7 of total DVOA, not 1/3.

133
by Dan :: Thu, 09/17/2009 - 7:23pm

1/7 ST, 3/7 Off, 3/7 Def means that ST is worth 1/3 as much as Off (or Def)

134
by Bowl Game Anomaly :: Thu, 09/17/2009 - 9:06pm

Dan and TGT2 are correct. Aaron didn't design DVOA to make ST worth 1/7 of team value, that was something he discovered from DVOA. The formula adds O, D, and ST equally. It's just that the best and worst ST tends to be 1/3 as much as the best and worst O or D.

7
by C (not verified) :: Tue, 09/15/2009 - 5:17pm

Redskins have a higher offensive DVOA than the Giants, huh? I guess converting 3rd downs isn't as good as scoring a garbage touchdown at the end of a game by a QB who didn't complete a real pass until the 2nd quarter. Maybe the Redskins should play Hunter Smith at QB since I'm sure he has a high DVOA.

16
by dbostedo :: Tue, 09/15/2009 - 6:00pm

Wow...bitter? Perhaps looking into how DVOA is derived and understanding what it means would help. (If you have, I apologize - your comment makes it seem like you haven't.) And yes, Hunter Smith would have a good DVOA (1 fourth down conversion of a few yards for a TD). But he wouldn't have much in the way of DYAR.

98
by DaveRichters (not verified) :: Wed, 09/16/2009 - 11:38am

Great idea! I'd love to look into how DVOA is derived. Do you have a suggestion as to how this might be done? DVOA is a model that is unpublished, and I can find very little information about it. How many free parameters does the model have? Is the model theoretically motivated, or is it simply Aaron drawing a line through a bunch of points? Honestly, I think it is weird that so many people here love DVOA so much when, unless I'm mistaken, we don't know anything about it. If DVOA were open and available then we could evaluate it, but when I see comments about how it is tweaked and adjusted I have to wonder exactly what this model is. In an abstract sense it is obviously good, but I doubt it would survive a peer review process in the current form. Maybe it would, but until it is published I don't see any point in thinking about it too much.

106
by dbostedo :: Wed, 09/16/2009 - 12:55pm

Good points DaveRichters...I was referring to the basic ideas and concepts behind it, and what we DO know of what factors into it - which is quite a bit of info. Obviously you're talking about the intimate details of the formulas and their derivation.

I would like the think that could happen as well, but it would probably be very detrimental to this site as a business. If the formulas and methods are all "open-source" everyone could use them and run with them and FO would possibly lose out on some of the things it has going on with other sites and magazines. Call it FO IP I suppose...

108
by DaveRichters (not verified) :: Wed, 09/16/2009 - 2:44pm

Right on. I guess it is a tricky situation. In my job, my goal is to learn things and spread the information to as many people as I can, so I am benefitted financially by being open. In fact, there is no way to publish without doing so. But really, they don't need to publish their model for me to take it seriously, but they do have to answer some simple questions that I've asked (and got no response, but whatever). So I find it hard to think that DVOA is in any way useful as a tool, especially given the methods they HAVE published, like the "Curse of 370", which is theoretically grounded but statistically is a violent assault upon reason. Also, FO really needs to stop saying "this data" instead of "these data" if they ever want to be taken seriously by the scientific community, which I admit is probably a low priority for them.

112
by Pat (filler) (not verified) :: Wed, 09/16/2009 - 11:49pm

"Also, FO really needs to stop saying "this data" instead of "these data" if they ever want to be taken seriously by the scientific community"

If the scientific community doesn't take someone seriously for that, said members of the scientific community seriously need to get off of their freaking high horse.

From the American Heritage Dictionary:

USAGE NOTE The word data is the plural of Latin datum, “something given,” but it is not always treated as a plural noun in English. The plural usage is still common, as this headline from the New York Times attests: “Data Are Elusive on the Homeless.” Sometimes scientists think of data as plural, as in These data do not support the conclusions. But more often scientists and researchers think of data as a singular mass entity like information, and most people now follow this in general usage.

Hint: Scientists should be scientists, not linguists or English majors. If your biggest concern in an argument is whether or not a verb matches the tense of a word, you just might be missing the larger picture. English especially is a low entropy-per-word language; extract the meaning and ignore the details.

114
by Bill Barnwell :: Thu, 09/17/2009 - 2:55am

- This isn't your job. There's plenty of explanations on the site of how DVOA works on the macro level. There's plenty of evidence, provided both here and elsewhere, that it's a useful tool. It's certainly not perfect or anything close.

- I monitor the FO e-mail account. Neither that mailbox nor my personal inbox has any e-mails from a Dave Richter or anything similar.

- If you read an article that totally distorts what we've said into a strawman in order to prove a point, then yes, you might consider the "Curse of 370" to be a violent assault upon reason.

- If saying "this data" instead of "these data" is the difference between us being taken seriously by the scientific community (are you their emissary now?), I'll do without the scientific community.

117
by DaveRichters (not verified) :: Thu, 09/17/2009 - 8:14am

Hi Bill, thanks for taking the time to respond, although I think you shouldn't be so defensive.

First, you are begging the question (and I'm using this term appropriately to mean that you are making an assumption that you shouldn't be making). My conclusion that the curse of 370 is statistically poor was based on the original article, not anything else. I also read, years later, a piece on advancedNFLstats.com that attempted to debunk the curse, and that piece actually made the curse argument better than you did.

Second, responding to Bill and the earlier poster, it is true that grammar does not need to be applied correctly for the meaning to be extracted, but what's your point? Somehow I think that an essay written by Raiderjoe (no offense Raiderjoe) would benefit from editing even if it is understandable in its original form. Also, you shouldn't assume that a line that I added in jest was the focal point of my post. But "this data" sounds as bad to me as "this chairs". The point is not that you should bend over backwards to appease some grammar snob, the point is that you should try to be correct and sound intelligent. When John Clayton says "this data" it fits; he's an idiot and he sounds like one. I'm just trying to say that you should strive to sound more intelligent than John Clayton because you are.

Third, I appreciate your taking the time to look and see if you have ever received an email from me, that is nice and more than I expected. I'm not sure what email address I used, I suppose I could figure it out if you care a lot. But perhaps you could simply answer the question here. How many free parameters does the model have?

120
by billsfan :: Thu, 09/17/2009 - 9:56am

Real Scientists routinely commit grammatical atrocities on a grand scale--it's enough to make William Safire spin in his grave, and if he's not dead yet, then put him there first. Use of the passive voice is highly encouraged. New words combining Latin and Greek roots and affixes interchangeably are routinely invented. Prepositional phrases are strung together without regard to whether they're going anywhere. And colleagues whom I regard more highly than anonymous commenters on a football website routinely misuse data and datum as well as other archaic Latin words. I don't dismiss their work because of grammatical errors, and I take my pedantry very seriously.

This scientist likes DVOA, and not just because it likes the Eagles (the team that was written off last season, although not by DVOA, and somehow went to the NFC championship game). If you told us exactly how it worked, anyone could do it. Many long-time readers know enough about what goes into it that we trust it. While certainly not perfect, it is more illuminating than other readily-available stats.

(I also like the Eagles)

123
by DaveRichters (not verified) :: Thu, 09/17/2009 - 10:58am

Billsfan, you are one crazy dude.

I don't think the fact that scientists make grammatical errors is in any way relevant to whether or not FO should use "data" correctly. I'm honestly not seeing the connection there. I pointed this out in jest, but it is a pet peeve of mine. I would never dismiss an argument for poor use of the English language, most of my colleagues speak English as a second language, and quite poorly. However, if I edit a document, I don't say "Well, I'll leave this error in because other scientists make errors and I can understand what it means even though it is incorrect."

I agree with you that DVOA is more illuminating than other stats, which is why I read FO. And you admit it is not perfect, so we agree on that as well. What do we disagree on?

Are you really a scientist?

128
by Pat (filler) (not verified) :: Thu, 09/17/2009 - 4:04pm

However, if I edit a document, I don't say "Well, I'll leave this error in because other scientists make errors and I can understand what it means even though it is incorrect."

Well, duh, if you notice an error you fix it. The point is that

1) why should scientists be spending so much time looking for minor grammatical errors,

and

2) why did you insist that a minor grammatical error (which is even now considered acceptable usage) would result in someone not being taken seriously?

Really, though, point 2 is the big one. Fantastic scientists can be absolutely awful writers (and several of them are, God, don't I know that), and they still should be taken seriously. They're scientists, not writers. If they can convey their point, that's the entire idea.

148
by billsfan :: Sat, 09/19/2009 - 1:27pm

Yeah, that post was typed pre-coffee.

I'm a chemist, and judging by the subsequent posts, not the only football-nerd scientist around here.

(I also like the Eagles)

152
by Rick (not verified) :: Sat, 09/19/2009 - 5:22pm

I work in media, and part of my job is statistical analysis.

I've got a relatively good command of the English language. Grammar and spelling are generally not my strong suits, but I know from experience I'm better at both than 95% of the people in the US and 99% of the people with whom I work (LOL - imagine that, in media!)!!

I agree that when I see poor spelling or grammar, it drives me up a wall, particularly if it's my own. But I don't use it as a sledgehammer to reduce the value of what I'm reading. Even if the error is repeated frequently. Even if attention is drawn to the error and it's repeated. Simply put - we all make mistakes, and we all get comfortable doing what we do. So why take something trivial so seriously?

I've been on FO for about 5 to 7 years now - can't really remember how long, that's how long it's been. But I can say this: since I have a strong background in statistics (based on my Economics Master's Degree), I've found little to quibble with here. About the WORST argument I've had against anything here are the two times Aaron has analyzed the Super Bowl and said he expected blowouts based on DVOA data (both New England games, one against Philly and one against the Giants). He's been fair, though, noting that "anything can happen", and certainly the data indicated blowouts should have been in the cards. As it stands, New England has never blown out anyone in a Super Bowl and the Giants SB was a monster upset (though not if you believe, as I do, that Defense counts for more at playoff time, in which case my advice to friends who like the Giants was to NOT take the points and bet the Giants to win since it was more likely that if they covered, they'd win - this was purely based on a judgement call on my part).

Still, if that's the worst disagreement I've had, it's not a bad one and I have to admit the work here is terrific. It is one of the few sites that recognizes the difference between correlation and causation (I can't tell you how many people mess that up in incredibly disastrous ways - just talk to a politician and you'll die from listening to their ignorance on that topic).

Anyway - enjoy the site for what it is. It's an attempt to bring some logic and understanding to the statistical minefields of football. In most ways, it accomplishes this task effectively. I haven't seen too many years when their top teams haven't made the playoffs, or a top 5 team hasn't won it all. The fact is, good teams will have consistently good statistical performance and this system will account for that in many ways that are more meaningful that other standard measurement resources.

153
by DaveRichters (not verified) :: Sat, 09/19/2009 - 6:43pm

Hey Rick,

What sort of quibbles do you look for? I only ask because I don't think having a good statistics background is relevant for the issue I raised earlier. My point was simply that if a model has a lot of free parameters it can fit any data, but if a model is highly constrained by theory then a good fit is meaningful.

And the issue with "data", which I don't honestly care that much about, is that the scientific community universally uses "datum" as singular and "data" as plural. I don't think I've ever seen "data" as singular in a scientific journal, though I concede it is commonplace in general usage. My regard for this site is high enough in an intellectual sense that I feel it is appropriate for the writers here to adopt the scientific standard.

--Dave

155
by BucNasty :: Sun, 09/20/2009 - 3:12am

I think plenty of people realize that data is plural, it's just that it never gets treated as such because it just doesn't feel right. Saying things like "the data is..." or "the data says..." is wrong but natural, while "the data are..." or "the data say..." is correct but grating.

157
by DaveRichters (not verified) :: Sun, 09/20/2009 - 8:25am

Hi BucNasty,

It seems in YOUR lexicon "data" is singular, but for me "data are" sounds correct and "data is" sounds grating. How correct the usage sounds to you (and native speakers as a whole) is really the best test of whether something is grammatical, but that's not my point. I was simply suggesting that FO use the parlance of science in this case rather than what is generally acceptable. The thought appealed to me because it is harmonious in an artistic sense with the content here, which often strays from what most football fans believe is true about the game, like "run to win" and such.

--Dave

156
by BucNasty :: Sun, 09/20/2009 - 3:17am

About the WORST argument I've had against anything here are the two times Aaron has analyzed the Super Bowl and said he expected blowouts based on DVOA data (both New England games, one against Philly and one against the Giants). He's been fair, though, noting that "anything can happen", and certainly the data indicated blowouts should have been in the cards. As it stands, New England has never blown out anyone in a Super Bowl...

Maybe not, but stating that just because the Patriots have never blown out a Super Bowl opponent, obviously they will never blow someone out is a fallacy. As one of my favorite writers likes to say, history repeats itself until it doesn't. If you believe that DVOA is reliable and predictive, there's nothing wrong with believing the numbers when the system projects a blowout. Of course, there is also nothing wrong with disbelieving the numbers, either. As the disclaimer says, your judgement should be informed by the numbers, not dictated by them. But even without the aid of DVOA, my naked eyes told me that if ever there was to be a Patriots Super Bowl blowout, surely 2007 was the year. The offense bore no resemblance to the previous Pats teams, and the week before they began their playoff run that Giants defense that we now regard as prolific surrendered something like 37 points to them. You didn't have to be seeing with homer-vision to think things were about to get ugly.

113
by Pat (filler) (not verified) :: Thu, 09/17/2009 - 12:27am

unless I'm mistaken, we don't know anything about it.

You have to dig through the archives. The basics are under Methods, under About.

DVOA 1.0-4.0: here

DVOA 6.0: here

The tweaks are primarily changing success points and relative weights of plays in given situations, and then adding things that were simply ignored (due to lack of data) before.

but I doubt it would survive a peer review process in the current form

Completely made up theories submitted as a joke made it through peer review. This isn't a high bar.

118
by DaveRichters (not verified) :: Thu, 09/17/2009 - 8:35am

Hi Pat,

When you mention a made-up theory making it through peer review, I assume you are talking about an incident a while back involving a Theoretical Physics Journal. Yes, the bar is not high there, I agree. That's really an exception, though, because none of the articles there use actual data or anything. But in general the peer review process is much better.

The impression I get from the links you posted (thanks!) is that the model looks at many statistics, which may or may not be chosen arbitrarily, and uses the correlations to assign weights. That's a great way to draw lines through points, but it is hardly a theoretical model. If I add enough variables, some will be highly correlated with success just by happenstance. These variables make my fit to previous data nice but have absolutely zero predictive power. Again, I admit the peer review process has let things like this pass. The emerging field of social neuroscience is built around such "Vodoo Correlations" as Vul called them. This is why the variables in the model need to be theoretically motivated. The more the model is constrained by theory, and the fewer free parameters it has, the better (even though the fit will be worse). I don't think DVOA passes this test, but I'm not sure.

Don't get me wrong, I don't hate DVOA. I like seeing adjustments for opponent defense and situation, that's fantastic. I just think it could be a lot better and I'm bothered that the model isn't clearly defined.

119
by nat :: Thu, 09/17/2009 - 9:38am

The general theoretical basis of team DVOA is not in dispute:

(1) Good teams make good plays.
(2) Yards are good.
(3) First downs are good.
(4) Turnovers are bad.
(5) TDs are good.

There are a few points that have been disputed, but which seem to hold true:

(1) Fumble recoveries don't tell you much about how good a team is.
(2) Hail-mary plays don't tell you much about how good a team is.
(3) Some penalties tell you something about a team's ability to win in the future; others don't.
(4) Interception returns don't tell you much about how good a team is.

The theory is sound, and chooses its factors from a small set that is not at all arbitrary. But it doesn't tell us anything about how the factors should be weighted. And since we're trying to adjust two measures (VOA correlation with wins and DVOA correlation with future DVOA), the weighting is going to be something of a matter of taste.

This isn't an academic setting. The FO team tells us that they are not fudging the data, and we have to take that on faith. Peer review is the wrong model for what FO is doing.

122
by DaveRichters (not verified) :: Thu, 09/17/2009 - 10:48am

Great points Nat! But the devil is in the details. I'm not suggesting that FO is fudging anything, I'm saying that I don't regard DVOA highly because I don't know what statistical or logical errors may be there. I'm not even saying there ARE errors, I'm just saying that because the model is not published it is impossible for you or I to know if errors are there.

I don't know what you mean about peer review being the wrong model. I'm saying I would trust it a lot more if it were peer-reviewed. Wouldn't you?

127
by Arkaein :: Thu, 09/17/2009 - 3:39pm

Actually Dave, if you are mainly concerned with reliability of the method, you ought to be able to check it using black box testing methods.

Team DVOA has been published weekly in some form I believe for as long as I've been following this site, several years anyways. You could go back through the archives and dig up the old data and check how good a job it did at predicting future performance, since that's really the driving criteria in developing DVOA.

There are also copies of Pro Football Prospectus as well as this year's FO Almanac.

So while FO doesn't publish the exact details of their methods, they do publish a lot of data that they claim has strong predictive power, and those predictions can be tested, either by evaluating past projections against historical data, or by taking more current projections and waiting for the season to unfold.

I, like many other who frequent this site, have followed FO's projections long enough that I'm confident in it's predictive powers compared to other methods or prognosticators. However an independent, in-depth analysis of the projections would be possible and interesting (but a lot of work).

130
by Pat (filler) (not verified) :: Thu, 09/17/2009 - 4:32pm

When you mention a made-up theory making it through peer review, I assume you are talking about an incident a while back involving a Theoretical Physics Journal. Yes, the bar is not high there, I agree. That's really an exception, though, because none of the articles there use actual data or anything. But in general the peer review process is much better.

1) Saying theoretical physics doesn't use data borders on delusional. Theoretical physics is basically math, eventually bounded by data. Deep enough in, it's essentially all math. Saying that math doesn't use data is completely nuts - of course it uses data. If you use an equation or theorem from someone else, that's data. If you Monte Carlo a model, that's data.

(Note: saying that to a physicist may have been a bad idea.)

2) There's no logic in the statement that theoretical physics would be the place where peer review fails. A single peer reviewer can verify the arguments made in a purely theoretical paper. A single peer reviewer absolutely can not verify the data from a huge multi-collaboration experiment. Heck, the peer reviewer couldn't verify the data from a simple sociological survey.

The real reason peer review does anything is the belief that it works. People don't submit fake papers; reviewers assume that the person involved does due diligence. In general, reviewers won't catch a logical mistake; there have been tons of papers that were later retracted, or disproven, due to the entire foundations of the paper being flawed.

Honestly, peer review isn't intended to reject an idea. It's intended to make sure that the idea is fleshed out enough that others can determine whether or not to reject it. But sometimes there's no interest in an idea, and so a bad idea sticks around without anyone saying whether it's good or bad. That's what happened in the physics paper, and it's happened countless times historically. I can give you a list two pages long of complete garbage papers I've found that never got refuted because no one cared.

This is why the variables in the model need to be theoretically motivated.

No, they don't. They just need to correlate, and be predictive. Theory can come before or after. Neuroscience has problems because they can't prove predictivity. That's not the case here. In any case, any variable that's based on the result of a play in a football game can be theoretically motivated by the fact that it happened in the game. So there's absolutely no bar there.

If I add enough variables, some will be highly correlated with success

But they won't correlate across time if the correlation isn't causative. Did you miss that part? That's predictivity, and a stat that's correlative to success and predictive across time is all you need.

Also, the idea that football - a game involving hundreds of human beings and nearly a hundred separate events - could ever be overspecified by a model is, uh, a bit crazy.

I just think it could be a lot better and I'm bothered that the model isn't clearly defined.

What parts aren't? You've got success points, and you've got weights for plays in various portions of the field and time. Calculate baselines over a time period, divide, average. Penalties are tossed, except for the few mentioned in the 6.0 rollout. The actual "play->success point" values aren't always specified, but those are tunables.

135
by DaveRichters (not verified) :: Fri, 09/18/2009 - 11:18am

Hi Pat, let me respond.

You say:

1) Saying theoretical physics doesn't use data borders on delusional.

I say:

Perhaps our fields define data differently. In experimental psychology we clearly differentiate between simulation and experimental data. When I see a paper that contains only a model and simulation it is no where near as compelling as ACTUAL data that cam from an experimental manipulation, or even observation. I'm quite sure that you will agree with that, it is in no way controversial. As an aside, I'm perfectly happy to debate the points we have been discussing but let's refrain from calling each other delusional and nuts.

You say:

2) There's no logic in the statement that theoretical physics would be the place where peer review fails. A single peer reviewer can verify the arguments made in a purely theoretical paper. A single peer reviewer absolutely can not verify the data from a huge multi-collaboration experiment. Heck, the peer reviewer couldn't verify the data from a simple sociological survey.

I say:

You're not making much sense here. I can't figure why you talk about "verifying data", which doesn't really have anything to do with peer review. Perhaps we should introduce ourselves so we can see why we have such different conceptions of peer review, even though we are both scientists (which I assume you are, but it is hard to tell). I'm an experimental psychologist studying sensation and perception. My background is in color vision (which is as close to physics as psychology gets). Right now I'm studying active perception, specifically eye movements. I work at a major university as a postdoc. What do you study? What journals have you published in? I'm not trying to be fancy or anything, but the physicists I know seem to have the same ideas about peer review as I do, so I'm a bit curious as to why we are so different.

You say:

The variables in the models don't need to be theoretically motivated, they only need to correlate.

I say:

I disagree. So if one variable, just by chance because there are so many variables, correlates well you would put it in the model even if it were the length of the LT's shoe strings. See, I wouldn't and that's where we are different.

You say:

But they won't correlate across time if the correlation isn't causative. Did you miss that part? That's predictivity, and a stat that's correlative to success and predictive across time is all you need.

I say:

You are just wrong. They WILL correlate across time if a third variable cause both, for example. But aside from that, I did not miss that point, and I'm seeing now that we kind of agree, so your hostile demeanor to me is just you being a peacock or something. My point is that if the variables are not theoretically motivated they will correlate well to past data but be useless in predicting. That's really the point of what I'm saying, so no, I certainly didn't miss that.

You also say that I'm wrong, and the model and methods ARE specified? Ok, great! This is great news! Could you send me a link or something? I can't wait to see it!

--dave

136
by Wanker79 :: Fri, 09/18/2009 - 12:35pm

What part of "proprietary" is so difficult for you to understand? If FO published the specific details of their calculations they would cease to be the unique source for this/these (which ever will appease the grammar Nazis) data. So do you just intrinsically distrust anything which has a proprietary component?

138
by DaveRichters (not verified) :: Fri, 09/18/2009 - 12:53pm

Hey Wanker79,

I understand proprietary, and I've never suggested that FO should publish their model. I'm saying, and again this is in no way controversial, that IF it were published THEN we would all be able to evaluate it better and trust it either more or less depending on the valuation. FO has weighed two factors and determined that the financial benefit of secrecy is worth more than the intellectual benefit of openness. It's simple, and I find it hard to believe that anyone could honestly disagree with that statement. I may disagree with FO on that choice, and in my life I have turned down better paying non-academic jobs because I prefer openness, but I understand why FO made their choice and I don't hold it against them, but we would all be donkeys if we act as if the model HAS been published and vetted when it hasn't.

I don't mistrust DVOA intrinsically, in fact I trust it intrinsically a great deal. Circumstantially, I mistrust everything that is proprietary and put forth for financial gain. Maybe that's just me, but I think it makes a great deal of sense.

--Dave

140
by TGT2 (not verified) :: Fri, 09/18/2009 - 2:34pm

Point taken. I prefer to use group open source software over proprietary software because it has been more thoroughly vetted for security. That doesn't mean I don't trust Microsoft Office to create Word and PowerPoint documents. I don't think anyone would require Microsoft Office open their doors to make sure they are correctly creating documents either.

You're not going to get FO to turn over their DVOA formula. Instead of complaining about it, if you're really worried that it hasn't been vetted, test its predictive power yourself.

One thing I would like to see is what the earlier versions of DVOA would be for different teams. That would be useful in determining efficacy of predictions and showing that DVOA actually is improving.

139
by TGT2 (not verified) :: Fri, 09/18/2009 - 2:25pm

You say:

But they won't correlate across time if the correlation isn't causative. Did you miss that part? That's predictivity, and a stat that's correlative to success and predictive across time is all you need.

I say:

You are just wrong. They WILL correlate across time if a third variable cause both, for example. But aside from that, I did not miss that point, and I'm seeing now that we kind of agree, so your hostile demeanor to me is just you being a peacock or something. My point is that if the variables are not theoretically motivated they will correlate well to past data but be useless in predicting. That's really the point of what I'm saying, so no, I certainly didn't miss that.

I believe the point you are responding to used some ambiguous and possibly erroneous language, but the idea behind it is correct, and you completely missed it.

For the purposes of determining if 1 team is better than another and predicting their future success, it really doesn't matter if we have the root causes. You don't seem to understand that. You assume you need causation to have prediction. That's what regression analysis is for. FO didn't just grab 5 seasons of data and pull out the factors that correlated with winning. They also tested their results against other non-baselined games and seasons to see what the predictive power was. They are constantly performing this regression and updating their baseline accordingly. They have explained this process. It's valid.

Clearly you are not a practicing statistician. neither am I, but I did take multiple advanced theoretical statistics classes in college, and though I would love to see their exact numbers, I don't think it's necessary. As has been mentioned multiple times in this thread, FO has described their general methodology for developing and improving DVOA. It fits with accepted statistical practices. You keep asking for this information, despite having been pointed to it. Maybe what you mean to ask for is the intermediate results of their analysis. You don't need the results to evaluate the methodologies.

There is one good point you do make. Without publication, we do not know how rigorously they practice their methodologies, or whether they occasionally tweak numbers on their own. Sure, they could be making everything up or cheating, but why should they bother? They have all the data to perform the statistical analysis. They have the requisite training to perform the analysis. They also have a track record of providing insight outside the standard box score (ALY, Stop Rate, Defeats, etc...). They have even documented when they've made subjective decisions (hail mary interceptions), when they've clearly made mistakes (Schedule adjustments), and which features they've found to be predictive (gaining % of yards on each down) and non-predictive (most penalties). For a good example of their candor, you should read about how the Colts backloaded Schedule one year broke DVOA. Technically, it could all be a croc, but I find that extremely unlikely.

So long as their results are predictive, they're fine. If you have a problem with FO, show their results aren't predictive. That's what a real scientist would do.

141
by DaveRichters (not verified) :: Fri, 09/18/2009 - 2:50pm

TGT2,

First, thanks for the advice about what a real scientist would do. I'll write that down because that will certainly help in my career as a real scientist. First, I assure you that I understood the argument perfectly well. I'm less sure about your understanding. Tell the Thanksgiving turkey that he doesn't need to understand root causes so he can be perfectly satisfied that he will continue to eat well and be happy for years to come because that's what his correlations tell him. You and others have really aggravated me with your insults which are off base and silly. Really, I don't understand the tone of your response, but if you want to result to attacking me, that's your decision, but I'm starting to think that you are either a witling or a cockatrice.

I know what I mean to ask for. I understand models. My lab, among other things, creates models of neural responses from the retina to the cortex, and the one thing I've asked is how many free parameters the model has. I've read the material that has been posted on FO about DVOA and I've not seen anything that I did not understand.

Thanks for giving me credit for the idea that FO may be "fudging" things, but I don't think they are and I haven't suggested that.

Listen, sorry if I have come off offensive, but you people need to deal with criticism better.

--Dave

142
by Wanker79 :: Fri, 09/18/2009 - 3:37pm

The only thing I take offense to is you lumping psychology in with real sciences. ;-p

143
by DaveRichters (not verified) :: Fri, 09/18/2009 - 3:40pm

Ha! I secretly take offense at that also!

149
by Kurt Klein (not verified) :: Sat, 09/19/2009 - 2:57pm

Hi Dave,

I've read most of your posts in this thread, and I sympathize with your frustration. It is pretty obvious that you are trying to have a discussion about some nuance or aspect of DVOA (which would be a really interesting discussion) and that no one really understands your point, leading them to blindly defend DVOA with guns blazing even though you are largely in agreement. I usually get that kind of reaction when discussing certain religious or political topics, not football!

The fact is, there is a lot about DVOA that seems completely arbitrary. For instance, teams are ranked according to Adjusted Line Yards:
* Losses: 120% value
* 0-4 Yards: 100% value
* 5-10 Yards: 50% value
* 11+ Yards: 0% value
Both the ranges and the weightings here are certainly arbitrary (as is any classification system). Why not have 0 yards = 100%, 1 yard = 90 %, 2 yards = 80%, etc.? There might be a very compelling reason to do it the way it is done, and it seems that you just want to know why. Good question, because there is no obvious theoretical underpinning here. I'm guessing it is to get lines to fit points better, as you suggested. (If anyone reading this starts to defend DVOA or Adjusted Line Yards, you've completely missed the point.)

Maybe it would be better not to expect people to understand subtleties involving 'correlation', 'causation', 'theoretical', etc. Nat's post exemplifies this problem because he addressed you with qualitative, self-evident statements ("yards are good") which do not translate into any kind of inevitable, quantitative, analytical, predictive model. Also, whoever thinks experimental psychology is not real science should go back to school!

150
by tuluse :: Sat, 09/19/2009 - 3:02pm

Adjusted Line Yards are separate from DVOA, although they may factor into projected DVOA.

151
by MC2 :: Sat, 09/19/2009 - 3:54pm

I never fail to get a kick out of debates about what constitutes "real science". I get the feeling that if they had lived in an earlier time, many of today's scientists would have spent their time condemning and punishing heresy, just as they do today.

154
by DaveRichters (not verified) :: Sat, 09/19/2009 - 7:05pm

Hi Kurt,

Thanks for your kind words!

You say:

Nat's post exemplifies this problem because he addressed you with qualitative, self-evident statements ("yards are good") which do not translate into any kind of inevitable, quantitative, analytical, predictive model.

I say:

"Yards are good" is a fine theory, but if you don't have a theory driven constraint about exactly HOW good yards are and you let a correlation determine the weighting, AND you do that for many many variables, then the model is describing a combination of the things that really cause success AND random noise, which in this case is probably of great magnitude, AND confounding variables. I think it is debatable if a model built in such a way could ever hone in on the "truth" just by adding more data, but such models generally do not get published.

--Dave

144
by Dan :: Fri, 09/18/2009 - 3:58pm

close tag? Did that work?

145
by Wanker79 :: Fri, 09/18/2009 - 3:59pm

TGT2 broke the internet!

146
by tuluse :: Fri, 09/18/2009 - 4:04pm

Attempting to end the italics.

147
by Tom Gower :: Fri, 09/18/2009 - 5:02pm

Closing italics, I hope.

11
by vesini :: Tue, 09/15/2009 - 5:31pm

Re: PHI vs. CAR - After one game, the DVOA difference between the two is 218.7%??? Is this the biggest DVOA difference in a single game? If not, what are the biggest?

Question for the crowd - What is the best DVOA matchup in history?

vesini, who did not use the proper stats, and is dead.

12
by BD (not verified) :: Tue, 09/15/2009 - 5:40pm

I'm amazed to see Cincinnati's offense rated so high, given what I felt was very sloppy play on all but one drive. Also, I wonder what the defensive DVOA would look like without Stokely's TD. I would assume that one play wouldn't affect the overall rating that much, but still, I wonder.

13
by Rich Conley (not verified) :: Tue, 09/15/2009 - 5:47pm

I still haven't watched the whole Pats/Bills game, but I'm surprised VOA says its so close. The Pats held the ball for 40 minutes. They held the Bills to 270 total yards, and had 11 more first downs than the Bills.

20
by Will Allen (not verified) :: Tue, 09/15/2009 - 6:35pm

This is as good a place as any to note how truly awful the 2nd (I was in the car for the first) roughing the passer penalty was. All bad cliche/jokes aside, if that is how that infraction is going to be called, the game can hardly be played.

44
by BucNasty :: Tue, 09/15/2009 - 8:04pm

If the second roughing call was on Thomas, I thought there was somewhat of a case for it. I haven't seen any highlight shows yet, but he did it several seconds after the play was blown dead with Edwards in the grasp. It probably wouldn't have been called a few years ago, but there's at least a case for it. The Wilfork call was the one that bothered me. He made contact a split second after the ball was thrown, but he got flagged because he chose to wrap the waist instead of hitting him high.

45
by Agamemnon :: Tue, 09/15/2009 - 8:43pm

I thought it was more an issue of Wilfork being so fat and slow that by the time his dive reached Edwards, the QB had scooted backwards enough that Wilfork landed on his lower half.

42
by PatsFan :: Tue, 09/15/2009 - 7:55pm

Really? I thought just the opposite. The call on Wilfork was totally bogus, but the one Thomas was fine.

64
by Anonymous Dude (not verified) :: Tue, 09/15/2009 - 11:49pm

I thought the call on Wilfork was a direct result of the Pats complaining about the Bernard Pollard play being dirty. They created a rule and it was enforced, a bad rule maybe but called correctly. If that was a bad call then so was the tuck rule call.

126
by TGT2 (not verified) :: Thu, 09/17/2009 - 12:47pm

Called correctly? The rule is about diving at a QB's knees. Wilfork hit him in the beltbuckle and wrapped up around the thighs. By your logic, if a defender grabs a players jersey around the shoulder pads, he should be called for a facemask.

71
by Rich Conley (not verified) :: Wed, 09/16/2009 - 1:35am

If they're going to call roughing on those sorts of play, they need to blow them dead earlier. THomas really had no choice but do what he did: put his leg between Edwards' and roll him backward. Its a pretty standard wrestling takedown, and looks much more violent than it actually is. There was really no other way for him to pull Edwards down with the way they were locked up.

The whistle blew when Thomas had already put his leg in and started to shift his weight.

74
by BucNasty :: Wed, 09/16/2009 - 2:01am

That's not how I remember it. I remember hearing a whistle at least once, maybe more, before Thomas even began to lift him up. Maybe he didn't hear them in the middle of all those bodies, but it was late.

94
by Rich Conley (not verified) :: Wed, 09/16/2009 - 10:53am

Thats not what I said though. He definitely lifted him after the whistle.

BUT, he had already planted the leg and started to shift weight before the whistle was blown. They were both going down, whether it was him continuing the throw like he did, or him just falling down with Edwards on top of him.

99
by Todd S. :: Wed, 09/16/2009 - 11:46am

I agree with Rich that it looked worse than it actually was. And if Thomas stops trying to tackle him and Edwards breaks free, well, it's bad for the defense. I don't think you throw the flag there.

102
by Mathias Kiwanuka (not verified) :: Wed, 09/16/2009 - 12:19pm

I completely disagree. He has to let him go.

56
by billsfan :: Tue, 09/15/2009 - 10:33pm

The Pats held the ball for 40 minutes because three drives led to nothing--two failed conversions on 4th-and-short and a missed field goal. The Bills had a goal-line stand leading into halftime, forcing the Pats to settle for a field goal from the two-yard line. There were also the two failed two-point-conversion attempts, for whatever they're worth.

(I also like the Eagles)

18
by young curmudgeon (not verified) :: Tue, 09/15/2009 - 6:31pm

These data would be better if they were saranated. Let's hear it for SVOA!!!

39
by NY expat :: Tue, 09/15/2009 - 7:38pm

I suddenly have an inexplicable desire for BBQ pork

49
by drobviousso :: Tue, 09/15/2009 - 9:48pm

Is a comment thread a venue that can be filled with delicious, porky BBQ smoke? Lets find out.

93
by Christmas Ape (not verified) :: Wed, 09/16/2009 - 10:45am

We found that 42% prefer SVOA over the other VOAs.

121
by MCS :: Thu, 09/17/2009 - 10:36am

mmmm, saranated.

19
by jimm (not verified) :: Tue, 09/15/2009 - 6:31pm

Before the season started I would have thought it was very unlikely the Vikings give up as many punt return td's as last year - but leave it to them to be ahead of a record setting pace.

I'm thinking/hoping Harvin and Reynaud will be very effective in their punt and kick return roles which might help offset the awful punt coverage.

21
by Will Allen (not verified) :: Tue, 09/15/2009 - 6:42pm

Jimm, I'm about 75% serious in saying that if the Vikings aren't any better than this in covering punts, they oughta' declare most of the field 4 down territory, and to hell with punting.

22
by Red Hedgehog :: Tue, 09/15/2009 - 6:42pm

I'm surprised that Philly's special teams come out so high. I know they returned a punt for a touchdown, but they also had a kick blocked and had one kickoff out of bounds.

23
by jimm (not verified) :: Tue, 09/15/2009 - 6:43pm

Last year when Minnesota played GB at the Metrodome DVOA gave the advantage to GB. Watching the game I thought Minnesota completely over powered them but GB stayed in the game because of punt and interception returns for scores. Watching the game I felt Minnesota was the far superior team.

I had the same feeling about Chicago in the game on Sunday. GB eked out a win but I think Chicago is a better team.

DVOA obviously finds correlation to winning in things that seem more good fortune to me.

25
by Will Allen (not verified) :: Tue, 09/15/2009 - 6:52pm

jimm, the Vikings absolutely dominated the line of scrimmage in that game, but were completely dominated in special teams and quarterback play. DVOA penalizes ints, of course, but really treats pick sixes as random. There really isn't much randomness in letting the opponent consistently catch punts and run with them 60 yards.

59
by Aerogopher (not verified) :: Tue, 09/15/2009 - 11:13pm

I was able to catch the Vikes preseason games and the opener. Overall, it looked to me like the coverage teams are superior to last years. The return in the Dallas game was a fluke and easily fixable. I also believe with all the changes to the special teams, they will get better with more games.

51
by MountainTiger (not verified) :: Tue, 09/15/2009 - 10:14pm

As a Packer homer, I have to disagree about GB-CHI. The CHI offense was horrendous in the first half, making one big play and promptly turning the ball over. DVOA likes consistency, and completing less than 50%, throwing four picks, and getting sacked twice all ruin CHI for that game. I think this was born out in the game; if Collins plays the TD right, the refs don't call minimal contact AT five yards on the go-ahead FG drive for the Bears, and Crosby hits a makeable FG early, that game would have looked far better for GB. To me, CHI required some luck and a bad offensive day from GB to stay in it and should be glad that a four turnover (five counting the fake punt, maybe the most damaging) day didn't turn out worse.

60
by Dan :: Tue, 09/15/2009 - 11:24pm

There were lots of plays that could've gone the other way. If Knox doesn't step on the sideline when Woodson dives at his ankles, or if Clark keeps moving along the back of the end zone, then the Bears take a 7-3 lead and Jolly doesn't get his INT (although the Bears don't get their safety). If Mannelly snaps the ball quicker, or just plays it safe, the Packers don't take the lead with that FG. And the bogus illegal contact penalty helped the Bears but wasn't essential - without it the Packers would have gotten the ball deep in their own territory with over 8 minutes left on the clock, and the Bears probably get the ball back with plenty of time left and a chance to take the lead (and their offense had been pretty effective in the 2nd half).

I expected GB to have the better VOA, but I was surprised to see their offense score positively. They really only had 2 drives all game, one for the TD at the end and one for a FG earlier (their other TD & 2 FG attempts came on a short field). The lack of TOs helps a lot, but other than the INTs their numbers were worse than the Bears - fewer yards, first downs, yards per play, etc.

63
by Flounder :: Tue, 09/15/2009 - 11:46pm

It irks me to see so many people kind of discount Jolly's TD as lucky. He was actually responsible for pass coverage on Forte in the defense called if he came up the middle. He was there because he was supposed to be there, and made the play.

67
by Dan :: Wed, 09/16/2009 - 12:29am

I didn't say that Jolly's pick was lucky, just that the Bears nearly had a touchdown on two of the preceding plays, which would have prevented that play from ever happening. Not that those 'nearly's mean much - the comment that I was responding to pointed out a few 'nearly's that could have gone GB's way.

That play by Jolly was probably the best defensive play by anyone all game, and it was impressive whether Jolly was there because he read the screen (as I initially thought) or because the defense was designed for him to be there (as you say). The one bit of solace that I can take as a Bears fan is that it's probably not very predictive of future Cutler interceptions (especially if they stop calling those foolish middle screens).

79
by Ambientdonkey :: Wed, 09/16/2009 - 6:44am

I guess 350 pound men frequently make diving 1 handed interceptions in your world. I wish I got to watch those games.

124
by MCS :: Thu, 09/17/2009 - 11:03am

It actually happens more often than you'd think.

http://www.pro-football-reference.com/play-index/psl_finder.cgi?request=...

90
by MountainTiger (not verified) :: Wed, 09/16/2009 - 10:07am

I don't mean to say that only GB suffered from missed opportunities, but the illegal contact and missed FG had nothing to do with the Bears play and little to do with the Packers and both could have greatly altered the course of the game. The TD is qualitatively different and (I can admit) really just a gripe; it was a great play by Hester and Cutler that most teams don't have the speed and arm strength combination to slip in and probably should have been stopped even against the Bears.

As for the positive offensive VOA, I am surprised but not shocked. GB benefits in VOA terms for only making one big play and sucking by constantly making 4 yard gains instead of a few 15 yard gains and a lot of minimal gains. That the sacks don't hurt more does surprise me, though. I would have expected a slight negative, not a slight positive, before opponent adjustments.

24
by shake n bake :: Tue, 09/15/2009 - 6:50pm

So I'm guessing it's the 2 turnovers to 0 that makes the Jags come out higher than the Colts (VOA doesn't know that Addai fumbles once per 200 touches rather than once a game and doesn't realize that Peyton Manning doesn't get picked off in the endzone regularly).

37
by inglorious boston (not verified) :: Tue, 09/15/2009 - 7:34pm

It's week 1, so what do you expect? It's not like they can use last season's data at all, since rules have changed, personnel have changed, and team chemistry for all teams is inarguably different. And while it might be comforting to think that Manning will rarely throw an interception in the end zone this season, the current data don't allow for that conclusion yet. The same can be said for Addai. We just won't know how often those scenarios will occur until the sample data is large enough, which will take considerable time, considering how many games it will take to see Addai's "200th touch" (if he gets thirty handoffs a game, DVOA won't see a "proper" adjustment until after week ). And even then, adjustment won't occur unless it turns put he actually only fumbles once every 200 touches. With all the off-season turmoil, you should be happy that they have a positive DVOA value. Expect it to rise over the season, but don't complain that it is low now; they beat a mediocre team by two points, and the teams combined for less than 30 points. DVOA isn't particularly helpful until week 8 or so, and even then it's only useful if there are no serious injuries on the team. It's too early to be touchy.

38
by inglorious boston (not verified) :: Tue, 09/15/2009 - 7:37pm

Oops. Parenthetical aside should end with "9."

40
by bmerryman :: Tue, 09/15/2009 - 7:53pm

"they beat a mediocre team by two points"

DAVE has the Jags as #6. Hardly mediocre.

43
by shake n bake :: Tue, 09/15/2009 - 8:01pm

Maybe I should have been more clear in my tone since there's a lot of negativity in the thread. I'm not bashing. I understand that it's the way any objective system is going to work in a small sample. I was just throwing out my best guess for why VOA says what it does about the Colts and wondering if anyone had a different or additional explanation.

plus (as partially pointed out already) by the preseason projections the Colts are neither fortunate to be above average nor played a mediocre team on Sunday.

27
by Anonymous1 (not verified) :: Tue, 09/15/2009 - 7:01pm

I'm surprised Buffalo graded out so low. I had that pegged for the weekly "team somehow outplays the other by a mile according to DVOA but still loses" game.

72
by Rich Conley (not verified) :: Wed, 09/16/2009 - 1:39am

The Patriots pretty much moved the ball at will all game...until they got in the red zone, or got to fourth down.

I just finished finally watching the game, and NE was definitely the better team. Couple fluky plays gave Buffalo the lead(the missed Gost FG, a fumble on a kickoff being advanced 20 yards, Thomas being called for roughing on third and long, etc).

81
by steelberger (not verified) :: Wed, 09/16/2009 - 6:54am

Yeah, that and the Pats defense being bad.

91
by billsfan :: Wed, 09/16/2009 - 10:14am

The McKelvin fumble, of course, was not a fluky play, just a manifestation of the Patriot's indefatigable nature. Fred Jackson moving the ball consistently in 8-15 yard chunks was also clearly a repeated fluke.

If I were you, I'd be more concerned about Wilfork's inability to get consistent penetration on three linemen starting their first NFL game, Mayo's injury, and Brady's inability to throw the ball more than 20 yards.

(I also like the Eagles)

97
by Rich Conley (not verified) :: Wed, 09/16/2009 - 11:28am

McKelvin fumbled earlier in the game too, so I'd have to think that they went into that play trying to hold him up if he ran the ball out.

As to Brady not throwing more than 20 yards, have you ever seen the patriots play football? They've been playing a short passing offense since 2001. The question here should be, knowing that, why the hell did Buffalo have their safeties 20 yards from the LOS all game?

109
by Purds :: Wed, 09/16/2009 - 4:37pm

The Bills had the safeties 20 yards deep because without the kick off fumble (ie: stupidity), the Bills likely win that game.

110
by billsfan :: Wed, 09/16/2009 - 6:28pm

I live in Boston; they're on every week.

Did you watch a single Patriots game in 2007, the last time that their current starting QB played for them? Every other play was max-protect with two tight ends and deep balls to Moss behind the safeties. Even in Brady's first pre-season appearance, he went deep twice--once for a completion, once for a 50-some-yard pass interference call. The fact that he's not throwing deep may mean that his pre-season shoulder injury is more serious than they're letting on.

(I also like the Eagles)

29
by bubqr :: Tue, 09/15/2009 - 7:05pm

Eagles Def DVOA of 94,3 % Wow. Delhomme is what, 75 % of those 94,3 % ?

30
by RaiderRefugee :: Tue, 09/15/2009 - 7:12pm

The Haley's comet rarity of a positive Raider offensive DVOA rating is indeed a crumb of respectability for the great unwashed masses otherwise known as the black hole. JaMarcus Russell appears to have pinpoint accuracy between the hash marks and the eye sight of Mr. Magoo outside those same hash marks.

31
by Key19 :: Tue, 09/15/2009 - 7:12pm

I thought Dallas' defense would be low, but 30? Ouch. That REALLY needs to be improved. Offense looked pretty good for the most part but 1 is a bit over-the-top. I know that it's early in the season so the kinks are still being worked out, but still. I really thought that Special Teams would be in the top 10, but they missed that by a small margin. That said, they're SO much better than they were last season. I think we had 4 touchbacks this week. That's 4 more than all of last season! Really excited about that.

Too bad that if the defense continues to be terrible, we'll never get anywhere.

55
by MilkmanDanimal :: Tue, 09/15/2009 - 10:25pm

I'm admittedly playing a bit of the homer card, but I thought the Bucs really controlled that game. The secondary was utterly awful for three plays that resulted in long TD passes, but they ran extremely well, and Leftwich never got sacked. Leftwich!

I'd be interested to see what the DVOA splits would look like without those three TDs passes. Yes, yes, I know, they're part of the game and most definitely something to take into account, but this really struck me as a game Tampa probably should have won except their heads were shoved up their @#$!!!! on three occasions.

62
by The Powers That Be :: Tue, 09/15/2009 - 11:32pm

The Cowboys got those three long TDs because they used formations to set up mismatches based on what they'd seen earlier in the game. Yeah, the Bucs could've done a better job on them, but they weren't just random blown coverages: they were deliberately set up blown coverages. See:

http://www.bloggingtheboys.com/2009/9/14/1029728/up-from-the-depths-aqua...

for a very good dicussion of this, focusing on the role of Martellus Bennett in creating mismatches.

32
by Yinka Double Dare :: Tue, 09/15/2009 - 7:19pm

I'm rather surprised that Green Bay actually has a positive number for their offense. They didn't do a damn thing until Lovie Smith, in his infinite genius, left the corpse of Nathan Vasher on an island against Greg Jennings on 3rd and 1. About the only positive thing you could say about the Green Bay offense on Sunday was that they didn't turn it over; otherwise they flat-out sucked. They had 27 yards in the second half until that last possession where most of the yardage came from the Jennings bomb.

34
by jimm (not verified) :: Tue, 09/15/2009 - 7:23pm

Will - There really isn't much randomness in letting the opponent consistently catch punts and run with them 60 yards.

Apparently not for the Vikings.

But I would have thought otherwise. Doesn't special teams DVOA fluctuate far more greatly than defence and offence?

Either way - I'll take the team that dominates the line of scrimmage over the team winning on returns in the long run.

41
by Bowl Game Anomaly :: Tue, 09/15/2009 - 7:54pm

The return team consistently getting big returns is lucky (thus non-repeatable). The coverage team allowing big returns is (lack of) skill based (thus repeatable). It's easy to be bad, hard to be good.

It's no different than field goals. Opponents making all their FGs against you is unlucky. You missing all your FGs means your kicker is bad.

35
by manginijm :: Tue, 09/15/2009 - 7:23pm

Any chance of posting the DAVE/DVOA ratings for the Offense and Defense breaking down the Pass and Rush?

36
by jimm (not verified) :: Tue, 09/15/2009 - 7:29pm

Will the NFC catch up to the AFC this year?

Just week one but 7 of the top 8 teams in week one are NFC teams including two that played AFC opponents.

Sticking with my Oakland will surprise pick I made late last year. I think they will challenge SD for the West. And no, I'm not just trying to make Raiderjoe happy.

68
by RickD :: Wed, 09/16/2009 - 12:38am

We have two interconference games so far - Minnesota flattening Cleveland and Atlanta beating Miami at home.

I think we'll have to wait a bit until we see a few more interconference games. Using intraconference games to gauge the relative strengths of the conferences is a bit silly.

More generally, Week 1 DVOA/DAVE numbers are inherently silly for a variety of reasons. You cannot credibly have opponent adjustments when you've only had one opponent.

46
by speedegg :: Tue, 09/15/2009 - 8:46pm

Can I complain if my team's DVOA is too high? I mean how can the Chargers (who are FOA favorites to go to the Superbowl) play so inexplicably bad? Aren't championship teams suppose to CRUSH the cupcakes on the schedule? I know it's Oakland and not Detroit, but I have my doubts about their chances. What happened to the O-Line? The defense? The offensive play calling? The...oh yeah, they got Norv'd.

Is there a y-combinator that adjusts the DVOA algorithm to account for (bad) head coaching?

Can San Diego get Riveria or Cameron to be the head coach next year?

47
by Raiderjoe :: Tue, 09/15/2009 - 9:05pm

Raiders is not cupcake tema. Raiders on upswing but not yet fully figure out hwo to win. By 2nd meeting between them Raiders going to tkake Chargers to school. Beat them behind woodshop. Raiders 27, Chargers 13 in San diego.
Last night Raiders played not to win. Hopefully learn from this. Will too. charger oline was real inexperience , raiders should have brought house. There was no way Chargers would kept rivers uprighted

Raiders too low at 22 DVIA. Going to takme care of that next week, egt to 1-1 and get DVOa up in 12-15 area. Then wehn 2-1 and 3-1 cna get into top 10.

48
by Will Allen (not verified) :: Tue, 09/15/2009 - 9:24pm

Balance has been brought to The Force.

61
by Doug Farrar :: Tue, 09/15/2009 - 11:28pm

Nobody saranates a thread like Raiderjoe.

73
by Jerry :: Wed, 09/16/2009 - 1:56am

+1.

85
by dbostedo :: Wed, 09/16/2009 - 9:07am

+2 (because my fantasy team name is Emmitt Speaks Good)

58
by speedegg :: Tue, 09/15/2009 - 10:52pm

Whoa, easy there Pale Rider! I've been to a Raiders practice and Cal's practice is better! That's tough for me to say because I'm not a Coach Tedford fan. Russell needs to work on his footwork and mechanics. For example, Carson Palmer will throw a deep pass and the football will arc with tip pointing down, hitting his receiver in-stride. Russell will throw a bomb and he's chucking a duck. Is that what $60 million and two years will get you?

I agree, Rivers won't be upright in the next matchup. Mouthing off to the D-Line won't help his chances, but if costs the Chargers a win and a playoff spot, Norv will lose his job, and this year won't be a total loss.

Is it bad to wish Schottenheimer was around for an extra year or two?

76
by dangerdonkey (not verified) :: Wed, 09/16/2009 - 2:52am

I have a feeling that the Chargers are this years 2008 Seahawks. Meaning I think this team is likely to have a run on injuries that will keep them from reaching the potential. If they loose anymore linemen, loose players to suspension, or god forbid Rivers goes down, this is a 6 win team and we get to see another 8-8 team ala the Raiders take a lousy AFC West.

Did I just say the Raiders? Too much Sierra Nevada...

87
by MilkmanDanimal :: Wed, 09/16/2009 - 9:56am

Honestly? That would be utterly awesome. Al Davis would take it as a validation of the way he runs the team, Raider Nation would start crowing about how the glory days had come again, Cable would get a 14 year, $245,000,000 contract extension, and next year the Raiders would win five games again. It'd be ensuring the comedic value of the Raiders for the next decade and a half.

96
by Rich Conley (not verified) :: Wed, 09/16/2009 - 11:25am

Plus, they'd get to play the other division winners the year after, which, combined with a healthy SD, a better KC and Denver, would lead to another <4 win season, and the Patriots getting the first overall pick.

101
by bravehoptoad :: Wed, 09/16/2009 - 11:59am

The one thing I hear most often about the San Diego Charges is that they're the most talented team in the league.

I don't know why everyone comes to that conclusion; I think it's something someone said and now everyone parrots. A. J. Smith is supposed to be a great GM, right? So the Chargers must have the most talented team in the league. They have L. T. and Antonio Gates and Shawne Merriman and Phillip Rivers, right?

The thing is, the most talented team in the league would be the one least susceptible to injuries. Their keen talent evaluation would mean they had the best depth, allowing them to plug in under-rated low round picks without losing too much.

Well, it's clear that they may have a few super stars, but their O-line is terrible, their D-line is mediocre, and their secondary is bad. How are they the most talented team in the league, again?

105
by Bowl Game Anomaly :: Wed, 09/16/2009 - 12:44pm

Most talented =/= best talent evaluators. Usually when people say "the most talented team" they are referring to the starters, not the depth.

That said, the Chargers may very well not have the best collection of starters in the league. Off the top of my head I can't name another team with better starters, though.

125
by bravehoptoad :: Thu, 09/17/2009 - 11:33am

Unless you're on FO, in which case they mean depth.

It's one of their fundamental principles, that a team of stars and scrubs is not as good as a deep team.

If most people think that having a few stars makes you the most talented team in the league, then this is just another wrong idea that they need to be disabused of.

131
by Mr Shush :: Thu, 09/17/2009 - 4:48pm

Really? Fundamental principles? I'd have said it was more a loose general conclusion, with extensive caveats. The DVOA best team ever, the 2007 Patriots, strikes me as a team that was reliant on a few irreplaceable stars and featured a considerable amount of dross. The DVOAlicious Greatest Show on Turf was good primarily thanks to its three certain and two probable Hall-of-Famers, not because of its wonderful depth. And it's not like the Chargers don't have other good players besides their elite quarterback, pass rusher and receiving threat: Jackson is a pro-bowl calibre wide receiver, Phillips an excellent all-around OLB, Jamal Williams is still a high class space eater when healthy, and Castillo's a pretty effective 3-4 DE. I'm a big fan of Eric Weddle. I really don't think it's fair or accurate to characterise the Chargers as a few stars and a lot of dreck. If Merriman isn't able to return to pre-injury form, they're not the most talented team in the league. They're still very talented. Rivers-Jackson-Gates and a bunch of scrubs would walk the AFC West, and the Chargers are not Rivers-Jackson-Gates and a bunch of scrubs.

50
by starzero :: Tue, 09/15/2009 - 10:00pm

now i want a pork cupcake tema.

53
by DGL :: Tue, 09/15/2009 - 10:18pm

The Steelers' ST rank is higher than either their offense or defense ranks.

In other news, Satan is reporting two inches of ice cover, and pigs are circling over Greater Pittsburgh International Airport.

115
by bengt (not verified) :: Thu, 09/17/2009 - 4:04am

Have those pigs been saranated?

Special teams has Sepulverizer back, a good returner in Stefan Logan, and their coverage is one of Tomlin's hobby horses. The result you describe is not so surprising to me.

54
by PatsFan :: Tue, 09/15/2009 - 10:23pm

Pats DVOA is too high because Mayo has a sprained MCL and may be out for as many as six weeks.

66
by Telamon :: Wed, 09/16/2009 - 12:13am

When was the last time the Lions were number one in anything positive?

69
by Kalyan :: Wed, 09/16/2009 - 12:57am

I am mighty upset that the Pats got placed as high as 17! Their performance didn't deserve any better than early 20s at best!

And this from a Pats fan ..

77
by ammek :: Wed, 09/16/2009 - 4:06am

What strikes me is how predictable the week one results were. The top half of the VOA rankings is filled with teams with high-ish DAVE scores. In fact, only three of week one's losers are ranked higher in DAVE than the opponent who beat them: Cincinnati (a fluke loss), San Francisco (barely ranked lower than the Cards in either VOA or DAVE) and Green Bay.

And that corresponds with my sentiment about week one. There seem to be a lot of teams that everyone expects to be horrible this season, and they almost all lost. The bumbling Bs — Bengals, Bills, Browns, Bucs — all found a way to lose. The Broncos barely escaped with a win. Perennial patsies the Lions, Raiders, Chiefs and Rams got stuffed. The Texans flattered to deceive. That's ten teams that almost everyone expected to be bad, and they went 1-9. I suppose you could add the Niners to make it 2-9, although I've seen projections that have them winning their division.

Is this the end of parity? Or am I forgetting this is Jump to Conclusions Week?

78
by Dan :: Wed, 09/16/2009 - 4:50am

The wins & losses were mostly predictable, but the VOAs were not. Look at the middle of the rankings - VOA from -2% to 5%. The top 3 teams in DAVE are all there in the middle, as are 4 of the bottom 5. The teams with the best preseason projections (SD, NE, IND, CHI) all looked average or worse (5.0% max), and the teams with the worst projections (BUF, DEN, SF, ARI, NYJ, ATL, OAK) all looked average or better (-5.7% min). A few of those can be attributed to good teams playing other good teams (IND-JAX) or bad teams playing other bad teams (SF-ARI, ATL-MIA), but most of them can't.

80
by Ben :: Wed, 09/16/2009 - 6:45am

Anyone have any thoughts on why the Colts D is ranked 19th? They held to Jags to 40% 3rd down conversions, and a under 4 yards per passing attempt, though they did give up over 4 yards a rush. They weren't world beaters, but that rank seems a little low.

82
by Dan :: Wed, 09/16/2009 - 7:05am

They had -2.2%, which is a little bit better than average. The ranking doesn't mean much - they were as close to 11th as they were to 21st.

83
by ammek :: Wed, 09/16/2009 - 7:46am

And they forced no turnovers.

84
by crack (not verified) :: Wed, 09/16/2009 - 8:22am

Does Defensive DVOA give credit for forcing fumbles whether or not they are recovered by the D?

88
by Dan :: Wed, 09/16/2009 - 9:56am

Yes.

86
by Anonymous12340987 (not verified) :: Wed, 09/16/2009 - 9:27am

The Giants are clearly ranked too low because you are stupid. My Dancing-Chicken-On-A-Hot-Plate system is way better than this. U suk cuz u dont getit. ;)

92
by Theo :: Wed, 09/16/2009 - 10:15am

I like to know the difference between "getting passes dropped, being out of sync" NE offense and the "let's be serious and beat the Bills" NE offense.

95
by Rich Conley (not verified) :: Wed, 09/16/2009 - 11:22am

Three quarters of actual football at actual game speed.

NE has a QB who played limited time in preseason, didn't play all of last year, and didn't play in preseason last year. The last time Brady took meaningful snaps was really the 2007 superbowl. These things are to be expected.

100
by Aaron Schatz :: Wed, 09/16/2009 - 11:47am

Just to let everyone know, the stats pages now feature 2009 stats instead of 2008 stats. Playoff odds are updated. In addition, older position stats pages now include links to each player's personal page. These links are up for QB and RB, they will be added for WR and TE later today. They aren't on 2009 stats pages yet, but we're working on code so that player page links get added automatically when we update the player pages each week.

103
by Telamon :: Wed, 09/16/2009 - 12:31pm

I'm digging the expanded "Special Superbowl Matchups" section. Ah, the possibilities of week one.

107
by The Powers That Be :: Wed, 09/16/2009 - 2:22pm

Yes, but where's the SB X, XIII, XXX rematch between the Cowboys and Steelers? That one seems to be at around 0.8% and would certainly be notable. Or is that matchup so common as to no longer be "special"?

111
by Nick Wells (not verified) :: Wed, 09/16/2009 - 8:01pm

St. Louis is ranked 16th according to Dave??!! I'm not a frequent enough visitor to this site to know what your preseason projections are based on, but to me, that's enough right there to discredit them. I think Green Bay (ranked one spot ahead of the Rams) will be a playoff team, while S.L. will be competing for the #1 overall draft pick. Granted, I'm being nit-picky, but I think the projections should account for a lesser percentage of the first week rankings.

116
by ammek :: Thu, 09/17/2009 - 4:42am

St Louis is certainly a controversial projection, but Aaron argues it pretty effectively in the book. It hasn't been discussed much on the site, surprisingly. The least convincing aspect of the argument, to my mind, is the "first-round rookie lineman effect", if only because for each example given I can think of a whole lot of alternative explanations for the improvement. But someone has to do well in the NFC Wet, and I suspect that health is going to make the difference.

Far less controversial is the argument that week one VOA doesn't tell us nearly as much as the preseason projection. This is, what, year six of FO stats, and every year tells the same story: because of the small sample size, week one is a much poorer indicator of future performance than DAVE. In any case, the statistical table is ordered according to VOA, so you can compare and contrast with your own subjective insights, and then return to this page in three months' time to see how wrong you were.

158
by Jetspete :: Sun, 09/20/2009 - 8:05pm

thanks a lot Aaron and Bill!!! i bought the premium package for ur picks!!! it took two weeks for my football season to end! I have invested in your site for four years, in the end it got me ridiculed for arguably the worst suicide pick in the history of football! i hope you enjoyed kris jenkins eating brady's flesh! take my 60 bucks and pretend u r cartman, cash it to quarters and swim in it so at least one of the two will enjoy it.

159
by Jetspete :: Sun, 09/20/2009 - 8:07pm

and another thing, what idiotic system would predict the jags over arizona? haha, great job DVOA. i dont know what's worse, the system that spewed out that pick or my dumbass for listening to it! four years of reading wasted on an incompetent pick.