Writers of Pro Football Prospectus 2008

Most Recent FO Features

CoxMar15.jpg

» Varsity Numbers: Honing in

Bill Connelly again looks at which college football teams the F/+ ratings are sure about, and which teams remain a mystery (led by Appalachian State).

31 Oct 2006

Week 9 DVOA Ratings

by Aaron Schatz

Here's a look at this week's DVOA ratings, with commentary now available on FOXSports.com. As of this week, the DAVE formula that includes preseason projections is gone, and we start using WEIGHTED DVOA, although the difference between that and regular DVOA is really small -- basically, just enough to have the Giants above the Eagles for the purposes of FOX.

I'm sure that FO readers, as opposed to the general public, will understand why Philadelphia and Pittsburgh are still fairly high -- you guys are used to the fact that we consider every game, so one game won't have a huge impact on the ratings unless it is a blowout. The commentary will talk a lot about Philadelphia, and Any Given Sunday covers the situation with Pittsburgh.

Last week, I actually responded to a few comments in the discussion thread. It's really going to be hard to do that in the future. These comment threads are getting very long, and I have so many other things I need to do, and it is tough to sort through all the bad suggestions (some mean-spirited, some well-meaning) to get to the few actual good suggestions.

If you really want to help improve DVOA -- and despite what some readers seem to believe, we do not think it is perfect by any means -- here are some good guidelines to follow:

1) I've said it before, and I'm going to say it again: If you want to really get my attention, send e-mail.
2) Any e-mail that suggests that we favor a certain team, or hate another team, gets tossed immediately.
3) Make a specific suggestion. Don't send e-mail that says "Team X has a high DVOA and they suck so DVOA sucks." An e-mail that says "I understand what you say about luck but if Pittsburgh is 2-5 they must blow goats" is just pointless.
4) Please read the description of how DVOA works before you make suggestions. If you own a copy of Pro Football Prospectus 2006, go back and read the essay that details all the work I did this past off-season improving the system, tweaking some things and trying others. People keep suggesting things that are already in the system, things I've written about over and over again, like "you have to consider turnovers more."
5) Please don't suggest something that isn't in the play-by-play. I love the game charting project, but we just can't use that information in DVOA because we don't turn around the data fast enough.

And now, to completely ignore what I just said about not responding to comments, I would like to respond to a comment that was made earlier today in the Any Given Sunday thread:

Rich Conley: This site increasingly disappoints me. Not the site, but the commentators. Originally, DVOA was based on questioning everything, now it seems like everyone just accepts it as being correct, even when it obviously has trouble with things. I still think there are plenty of aspects where we need to stop assuming things are luck, when theyre really not.

If fumbles by a RB are recovered 70% of the time by the offence, and a team is only doing it 10%, yeah, it could be luck. It could also be that the team has stupid lineman, who arent keeping their heads up. Problem? Theres no way to really tell, unless the trend stays that way.

As to Int return yardage being luck, thats a joke. If the ball gets picked off while a team is in either red zone, its run back for a TD a high percentage of the time. If teams aren't getting yards on those returns, maybe its not luck ... maybe their corners arent that fast.

I'll take the second part first.

When I first posted this, I misread Rich's comment. I thought he asked about end-zone interceptions. I actually address this in the DVOA commentary. Can you guess how many end-zone interceptions have been returned for touchdowns in the past six regular seasons?

Six.

That's it. One per year.

But it turns out he's talking about red-zone interceptions. So let's look at those too. Rich could mean two different things, so I will look at both.

From 2000-2005, there were 1,065 interceptions caught by the defense between their own 20-yard line and the end zone. 35 were returned for touchdowns, or three percent.

From 2000-2005, there were 394 interceptions where the offensive line of scrimmage was in the red zone, i.e. the 20-yard line or closer. 28 of these were returned for touchdowns, or seven percent.

OK, what about when a team is backed into its own end?

From 2000-2005, there were 333 interceptions where the offensive line of scrimmage was on or behind the offensive 20-yard line. 57 of these interceptions were returned for touchdowns, or 17 percent.

As I note in the commentary -- go down to New Orleans -- there is a difference between throwing an interception where you throw the pass from your opponent's red zone, and throwing an interception where the defender catches the ball in the red zone.

OK, but what if certain defenders have a better record of returning interceptions for touchdowns, something you can see stay constant from year to year? Well, I discuss this in PFP 2006, in the article on predicting fantasy defense. The correlation of defensive touchdowns from year-to-year is .01. And if, for some reason, that result is mistaken, and there really are certain defenders who have a special ability to return interceptions for touchdowns, do you really think Nnamdi Asomugha is one of them?

Now we'll take the first part of the question. That's a legitimate argument, that some defenders or linemen are just better than others at recovering fumbles. So I went and looked at recovering fumbles in the same season. Instead of comparing one year to the next, I compared the first half of one season to the second half of the same season, so each team basically is working with the same roster.

In 2005, the correlation between fumble recovery percentage in the first half of the season and fumble recovery percentage in the second half of the season was -.15 on offense, and -.09 on defense.

OK, maybe that was a one year fluke. Let's look at 2004.

In 2004, the correlation between fumble recovery percentage in the first half of the season and fumble recovery percentage in the second half of the season was -.16 on offense, and -.17 on defense.

So, if these numbers are to be believed, a team that recovers a high percentage of fumbles in Weeks 1-9 is actually slightly more likely to recover a low percentage of fumbles in Weeks 10-17. That's sort of silly, of course. More likely, there's just no connection. In other words, it is random.

Remember that you can always use the keyword "DVOA" to access the latest DVOA commentary at FOXSports.com.

* * * * *

These are the Football Outsiders team efficiency ratings through Week 8 of 2006, measured by our proprietary Defense-adjusted Value Over Average (DVOA) system that breaks down every single play and compares a team's performance to the league average based on situation in order to determine value over average. (Explained further here.)

OFFENSE and DEFENSE DVOA are adjusted based on strength of opponent as well as to consider all fumbles, kept or lost, as equal value. Opponent adjustments are currently set at 80% and will increase each week until they are full strength after Week 10. SPECIAL TEAMS DVOA is adjusted for type of stadium (warm, cold, dome, Denver/Mexico City) and week of season.

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

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


TEAM TOTAL
DVOA
LAST
WEEK
WEI.
DVOA
WEI.
RANK
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
1 CHI 47.4% 1 47.3% 1 7-0 5.6% 9 -28.1% 2 13.7% 1
2 PHI 36.3% 2 36.0% 3 4-4 23.7% 2 -13.5% 7 -0.9% 20
3 NYG 36.1% 4 36.7% 2 5-2 16.3% 4 -17.8% 5 1.9% 12
4 SD 34.1% 3 33.8% 4 5-2 21.7% 3 -6.2% 12 6.2% 2
5 BAL 23.9% 5 23.5% 5 5-2 -12.0% 24 -32.8% 1 3.1% 6
6 NE 22.2% 7 22.7% 6 6-1 16.2% 5 -3.2% 14 2.8% 8
7 JAC 17.8% 8 18.0% 7 4-3 -3.2% 20 -21.1% 4 -0.1% 16
8 DAL 17.3% 11 17.1% 8 4-3 3.8% 13 -12.7% 8 0.8% 14
9 IND 16.0% 9 15.9% 9 7-0 34.8% 1 16.7% 28 -2.0% 25
10 KC 13.7% 13 13.2% 11 4-3 1.5% 15 -9.8% 9 2.4% 9
11 PIT 13.6% 6 14.0% 10 2-5 -1.9% 18 -21.2% 3 -5.7% 31
12 DEN 10.7% 15 10.9% 12 5-2 4.0% 12 -7.5% 11 -0.8% 19
13 CIN 7.2% 16 7.0% 13 4-3 6.2% 8 3.0% 20 4.0% 4
14 STL 7.1% 10 6.9% 14 4-3 13.1% 6 5.6% 25 -0.4% 18
15 NO 5.6% 12 5.3% 15 5-2 4.9% 10 3.6% 21 4.3% 3
16 ATL 0.9% 18 0.8% 16 5-2 0.8% 16 -2.7% 15 -2.6% 27
TEAM TOTAL
DVOA
LAST
WEEK
WEI.
DVOA
WEI.
RANK
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
17 CAR -1.3% 17 -1.3% 17 4-4 4.2% 11 1.7% 19 -3.8% 30
18 MIN -1.9% 14 -1.7% 18 4-3 -15.6% 26 -13.8% 6 0.0% 15
19 GB -3.7% 22 -3.1% 19 3-4 2.2% 14 4.1% 22 -1.8% 24
20 WAS -4.5% 19 -5.1% 20 2-5 10.5% 7 17.1% 29 2.1% 11
21 CLE -14.7% 27 -14.6% 21 2-5 -18.3% 27 0.1% 18 3.7% 5
22 SEA -14.8% 20 -15.1% 22 4-3 -8.9% 23 5.6% 24 -0.2% 17
23 BUF -16.8% 23 -17.6% 24 2-5 -14.0% 25 5.3% 23 2.4% 10
24 TB -17.5% 21 -17.2% 23 2-5 -19.5% 29 -4.2% 13 -2.2% 26
25 NYJ -19.0% 24 -19.5% 25 4-4 -3.3% 21 18.7% 30 2.9% 7
26 MIA -19.8% 26 -19.8% 26 1-6 -18.7% 28 -1.6% 16 -2.7% 29
27 DET -20.8% 25 -20.9% 27 1-6 -2.6% 19 15.7% 27 -2.6% 28
28 HOU -29.1% 30 -28.7% 28 2-5 -0.9% 17 27.1% 32 -1.0% 21
29 OAK -30.4% 31 -30.0% 29 2-5 -38.0% 32 -8.8% 10 -1.3% 22
30 SF -33.1% 29 -33.1% 30 2-5 -6.6% 22 24.9% 31 -1.6% 23
31 ARI -34.9% 28 -34.5% 32 1-7 -26.7% 31 -1.0% 17 -9.2% 32
32 TEN -35.3% 32 -34.4% 31 2-5 -22.2% 30 15.0% 26 1.8% 13

  • NON-ADJ VOA shows what the rating looks like without adjustments for strength of schedule, luck recovering fumbles, or weather and altitude on special teams.
  • ESTIMATED WINS uses a statistic known as "Forest Index" that emphasizes consistency as well as DVOA in the most important specific situations: red zone defense, first quarter offense, and performance in the second half when the score is close.  It then projects a number of wins adjusted to a league-average schedule and a league-average rate of recovering fumbles.  Teams that have had their bye week are projected as if they had played one game per week.
  • PAST SCHEDULE lists average DVOA of opponents played this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • FUTURE SCHEDULE lists average DVOA of opponents still left to play this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • VARIANCE measures the statistical variance of the team's weekly DVOA performance.  Teams are ranked from least consistent (#1, highest variance) to most consistent (#32, smallest variance).


TEAM TOTAL
DVOA
W-L NON-ADJ
TOT VOA
ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VARIANCE RANK
1 CHI 47.4% 7-0 66.1% 6.5 3 -18.0% 32 -1.9% 22 30.2% 3
2 PHI 36.3% 4-4 27.9% 6.7 2 -0.9% 22 2.8% 9 4.2% 30
3 NYG 36.1% 5-2 31.6% 6.7 1 4.8% 5 6.0% 7 9.4% 21
4 SD 34.1% 5-2 41.1% 5.8 4 -5.1% 27 -8.7% 30 7.1% 26
5 BAL 23.9% 5-2 35.5% 5.4 6 -1.9% 24 -1.2% 19 11.5% 18
6 NE 22.2% 6-1 34.3% 5.7 5 -8.1% 30 -5.2% 26 10.0% 20
7 JAC 17.8% 4-3 18.1% 4.6 11 7.6% 3 -5.4% 27 38.6% 1
8 DAL 17.3% 4-3 17.8% 4.7 9 2.5% 14 2.2% 14 13.7% 12
9 IND 16.0% 7-0 20.3% 5.4 7 -3.3% 26 0.0% 15 7.8% 25
10 KC 13.7% 4-3 8.5% 4.8 8 -2.5% 25 -0.2% 17 36.2% 2
11 PIT 13.6% 2-5 7.5% 4.4 13 3.4% 9 2.6% 12 15.6% 10
12 DEN 10.7% 5-2 7.3% 4.5 12 4.7% 6 -1.3% 20 9.2% 22
13 CIN 7.2% 4-3 5.7% 4.1 17 2.1% 15 10.3% 3 6.7% 27
14 STL 7.1% 4-3 19.0% 4.6 10 -7.8% 29 -7.5% 28 10.4% 19
15 NO 5.6% 5-2 5.8% 4.3 15 3.0% 11 2.3% 13 11.6% 17
16 ATL 0.9% 5-2 8.3% 4.1 16 1.3% 17 2.7% 10 30.0% 4
TEAM TOTAL
DVOA
W-L NON-ADJ
TOT VOA
ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VARIANCE RANK
17 CAR -1.3% 4-4 -5.0% 3.8 20 3.0% 10 8.6% 4 8.1% 24
18 MIN -1.9% 4-3 -0.9% 4.4 14 1.6% 16 -8.9% 31 12.7% 15
19 GB -3.7% 3-4 -4.1% 3.8 19 3.0% 13 -4.3% 25 13.2% 14
20 WAS -4.5% 2-5 -5.4% 3.8 18 3.0% 12 13.4% 2 13.2% 13
21 CLE -14.7% 2-5 -17.4% 2.7 25 -0.5% 20 6.7% 5 3.9% 32
22 SEA -14.8% 4-3 -18.4% 3.0 22 6.7% 4 -11.2% 32 8.8% 23
23 BUF -16.8% 2-5 -20.5% 3.0 21 3.8% 8 -1.9% 21 16.1% 9
24 TB -17.5% 2-5 -30.5% 2.9 23 10.4% 2 5.5% 8 4.3% 29
25 NYJ -19.0% 4-4 -13.1% 2.8 24 -7.3% 28 -3.6% 24 18.2% 8
26 MIA -19.8% 1-6 -9.3% 2.3 29 -9.7% 31 6.5% 6 4.2% 31
27 DET -20.8% 1-6 -19.3% 2.4 27 -0.2% 19 -0.6% 18 5.1% 28
28 HOU -29.1% 2-5 -30.6% 2.3 30 4.0% 7 -2.7% 23 14.0% 11
29 OAK -30.4% 2-5 -36.3% 2.2 31 -0.1% 18 2.6% 11 12.6% 16
30 SF -33.1% 2-5 -43.0% 2.6 26 10.5% 1 -7.5% 29 29.0% 5
31 ARI -34.9% 1-7 -27.0% 2.4 28 -1.8% 23 -0.2% 16 26.5% 6
32 TEN -35.3% 2-5 -26.4% 2.0 32 -0.6% 21 15.5% 1 19.7% 7

Posted by: Aaron Schatz on 31 Oct 2006

240 comments, Last at 04 Nov 2006, 1:24am by Pat

Comments

1
by mannie fresh (not verified) :: Tue, 10/31/2006 - 5:53pm

me first

2
by michael (not verified) :: Tue, 10/31/2006 - 5:54pm

warning warning:

As to Int return yardage being luck, thats a joke. If the ball gets picked off while a team is in either red zone, its run back for a TD a high percentage of the time.

Can you guess how many end-zone interceptions have been returned for touchdowns in the past six regular seasons?

I'm not disputing the facts, of course, since I don't have them. I just need to point out one of you is talking about RED zone INTs, and one of you is telling him he's wrong about END zone INTs.

3
by Pat (not verified) :: Tue, 10/31/2006 - 5:58pm

Pittsburgh really isn't the surprise this year. We know what's wrong with them - estimated wins tells us. Too high variance. Too sucky special teams. Tough schedule.

It's the exact same thing with Jacksonville. The team that, surprise, surprise, did the same thing as Pittsburgh, and went out and got whacked by Houston, after waxing the hell out of the Jets. So here's the lesson: whatever you were thinking about Jacksonville last week (they must suck after all of their injuries, they lost to Houston), you're thinking about Pittsburgh right now.

Both of them are out of the top 10 in Estimated Wins.

Philly's the real "what the hell?" team this year. They're doing everything right that you historically have needed to win games, and still losing.

4
by zip (not verified) :: Tue, 10/31/2006 - 6:00pm

Rich said "red zone interception," you said "end zone interception." Sounds like you aren't talking about the same things, to me.

Also, the rareness of end zone runbacks on INTs doesn't really prove that INT yardage is random (although it does refute Rich). I would like to know if there are teams that get INT return TDs at well above the league average, but I think sample sizes are so small it would be hard to ever prove anything. But you could make the argument that a guy like Ronde Barber is really good at jumping short out routes and taking them back the other way, so the average Ronde Barber interception produces more points than a replacement-level CB get a pick in a similar field location.

Anyway, I'm not saying I believe what I just wrote, but it seems possible that some cornerbacks, or defensive scheme, or coaching could be better at jumping short routes and thus would have more valuable interceptions than the league average.

On the other hand, the fumble analysis seems shockingly conclusive. If anything, maybe teams that get a lot of lucky fumble recoveries slack off as the season goes on because they are just so used to getting it :)

5
by Fnor (not verified) :: Tue, 10/31/2006 - 6:07pm

A lot is being made about Stillers fans being happy about them being ranked so high. I'm not. It just makes me watch or hear the game and think "What the hell," we're better than this. It'd be much easier to just say we're terrible and write it all off. I'd like to be able to do that. But based on what I've seen, I don't think I can, and it makes it that much more frustrating, because I know we're on a train to nowheresville this year.

6
by Kal (not verified) :: Tue, 10/31/2006 - 6:08pm

#4: Ronde had 5 TDs off of INTs before that last game. I think that's true - it might've been that he had that many afterwards. In any case, he's not any better than anyone else.

Endzone INTs aren't the same as redzone INTs, no matter how you spell it. Endzone INTS count some redzones, some hail marys, some long routes. Just doesn't seem like a reasonable comparison. Regardless, common sense should be able to refute Rich Conley; how many interceptions are there? Now, how many are returned for TDs? Now, of all of those INTs, how many INTs are thrown inside the 20? Even if you assume all INTs for TDs are from redzone INTs, and you assume that 1 in 5 INTs are in the redzone, I'd imagine it'd still be grossly out of favor for that proposition.

As to DVOA - oh, sure, your DVOA and statistical analysis and science are so smart. But tell me, Aaron, can your science predict the rain? Hmm?

7
by Goathead (not verified) :: Tue, 10/31/2006 - 6:09pm

The team that sticks out the most for me in terms of record not matching ranking is Ind. Can't believe they haven't lost yet....

8
by Riceloft (not verified) :: Tue, 10/31/2006 - 6:11pm

Cleveland at this point has the least variance and the 5th hardest remaining schedule. That isn't a good thing for a Cleveland fan. Ouch.

9
by NewsToTom (not verified) :: Tue, 10/31/2006 - 6:11pm

Things I don't like to see:
32 TEN -35.3%
Future Sched 15.5% 1

Oh, and I clearly recognize I'm being a homer when it comes to believing the Titans are better than their rankings indicate. They're still a bad team, but (i) they've been blown out twice and (ii) their much better form since switching to VY at QB hasn't shown up the statistics, particularly the passing metrics. The latter is, of course, totally subjective and utterly uncapturable by DVOA, but by garsh et golly, I can hang my hat on it.

10
by admin :: Tue, 10/31/2006 - 6:11pm

Erg. Sorry about mis-reading "end zone" and "red zone." I've now fixed the answer to the question to reflect all possibile definitions of "red zone touchdown."

11
by B (not verified) :: Tue, 10/31/2006 - 6:12pm

3: I still think it's penalties, and also clock-management. Normally penalties don't correlate well with performance, because those include penalties like holding which aren't consistently called, but Philly this year seems to be committing way too many false start, illegal formation and 12 men on the field penalties. I, of course, don't have any data to back this up, but it's what I noticed when I saw them on Sunday. Also, it's possible they keep playing highly variant teams who play better than their average against Philadelphia.

12
by dryheat (not verified) :: Tue, 10/31/2006 - 6:13pm

It seems anecdotally to me that if you have a cornerback with amazing speed and agility, and outstanding awareness and instincts.....a guy like Deion Sanders, perhaps.....then maybe long INT returns aren't chance?

Sounds logical to me, but I'm not sure how relevant it is to our purposes.

13
by B (not verified) :: Tue, 10/31/2006 - 6:15pm

From my experience, the closer an int is to the sideline, the more likely it is to be returned for big yardage. Also it depends on who makes the interception. free safeties and corners are more capable of getting a long return than linemen and linebackers.

14
by NewsToTom (not verified) :: Tue, 10/31/2006 - 6:15pm

Really not DVOA, but over 1/4 of the NFL (9 of 32) teams currently has a record of 2-5. Yet there's only one 3-4 team and three 1-6 teams (counting 1-7 ARI). Just me, or is this kinda bizarre?

15
by ChrisFromNJ (not verified) :: Tue, 10/31/2006 - 6:16pm

#7:

I'm not actually surprised about Indy- they just barely squeaked by NYJ and TEN (who DVOA has as dead last, as odd as that is to my eyes), and they were outplayed in Week I against the Giants, who only lost on the strength of the same penalties, miscues, and unlucky bounces that have been dooming Philly all season.

16
by Pat (not verified) :: Tue, 10/31/2006 - 6:17pm

I would like to know if there are teams that get INT return TDs at well above the league average

You don't really want to know that - what you want to know is if there are teams which consistently get interception returns with high yardage above the league average. There are always going to be teams above and below the league average, even significantly above. The question is whether or not that trend continues.

Even still, as I've said before, it so just doesn't matter. Any game that can be completely changed by replacing one play with the league average play was a coinflip, simply because there are so many things that could go on during a play. Maybe an official sees holding. Maybe an official sees pass interference.

17
by jackg (not verified) :: Tue, 10/31/2006 - 6:19pm

The AFC vs. NFC groupings within this weeks DVOA are starting to look interesting to me. It looks like there are 4 groups emerging:

1) NFC Elite Teams (#1-3)

2) AFC Playoff Contenders (#4-13 excepting Dallas)

3) NFC Playoff Contenders (#14-22 excepting CLE)

4) Cannon Fodder (#22-32 pulling equally from both conferences)

I know this is probably mostly a fluke, and I realize that there are many intra-conference games so this isn't like an AL/NL split in baseball, but I wonder if there is something related to the relative strengths of the two leagues that is driving this tiering. Could it be that while the top tier NFC teams (CHI-PHI-NYG- maybe DAL) are indeed strong, and maybe just as strong as the best teams in the AFC the relative lack of strength in the second tier NFC teams (ATL, CAR, etc.) actually makes these top NFC teams look artificially better?

As I say that it seems kind of perverse because weighted DVOA should reqard victories over good teams, but maybe the play-by-play aspect of it makes dominance over medicre teams look much more valuable than close wins (or tough losses) against very good teams (like Indys recent win over Denver).

Just a thought - but the conference stratification this week is pretty noticable.

18
by undethebus (not verified) :: Tue, 10/31/2006 - 6:21pm

In FUTURE SCHEDULE, it says, "It is not adjusted for which games are home or road."

Anyone know what the adjusted change is for home field? Or how much home field advantage is actually worth?

19
by Ryan (not verified) :: Tue, 10/31/2006 - 6:23pm

Well I know the adjustment for Chicago should be about +100% DVOA in their favor, at least.

20
by B (not verified) :: Tue, 10/31/2006 - 6:24pm

18: My understanding is it's about 17%.

21
by Pat (not verified) :: Tue, 10/31/2006 - 6:28pm

Also it depends on who makes the interception. free safeties and corners are more capable of getting a long return than linemen and linebackers.

Except corners and free safeties are also more likely to intercept the pass downfield, which means the return yardage will usually just reduce the amount of return yardage lost due to the long pass.

That'd be something to look at though, but you'd have to do something like compare the return yardage for DBs vs linebackers based on the distance of the interception from the line of scrimmage. So if a LB intercepts the ball 10 yards downfield, what's his average return yardage, compared to a DB.

22
by undethebus (not verified) :: Tue, 10/31/2006 - 6:29pm

#20 thanks, any back-up? I'm just curious cause I've been doing a 'pick 'em' using DVOA and I'm getting hammered. 71/256 is my record as of now. I wasn't using DAVE (though in retrospect I should have). I'm assuming that I didn't make home field advantage adjustments.

23
by undethebus (not verified) :: Tue, 10/31/2006 - 6:30pm

oops I mean my record is 71/128.

24
by zip (not verified) :: Tue, 10/31/2006 - 6:37pm

#6

I was just using Ronde as an example :) But he has 30 career picks and 5 TDs off them -- so that's a 16.7% TD rate, which is as good as the league average on picks thrown in your own red zone (according to Aaron's recently-changed commentary).

Hmm, lets look at a few more well-known CBs who have several years of experience, just for the hell of it:
----------------
Champ Bailey, 3/32, 9.3%
Nate Clements 4/20, 20%
Deltha O'Neil 3/30, 10%
Samari Rolle, 1/25 4%
Chris McAlister, 4/19, 21%
Ty Law, 7/48, 14.6%
Sam Madison, 2/33, 6%
Dre Bly, 5/30, 16.7%
Shawn Springs, 2/26, 7.7%
Aaron Glenn, 5/40, 12.5%
Patrick Surtain, 2/34, 5.9%

And of course..
Mike Brown, 4/14, 28.6% (7/20 if you add in fumble recs!)

Anyone want to figure out what the league average is for TD return interceptions made by a CB? I'm sure that's the position with the highest average.

25
by Becephalus (not verified) :: Tue, 10/31/2006 - 6:39pm

I think what people like Rich don't understand is that the reason things are getting more conservative is because things are gettting better and more comprehensive. There is just factually less improvment to be made. In the past people came up with actual novel ideas often. Now 99% of the time when I or someone else asks Aaron "what about X". X has already been looked at, or isn't able to be measured easily.

26
by Rich Conley (not verified) :: Tue, 10/31/2006 - 6:41pm

Aaron, I was talking plays that start in the redzone.. not talking balls that are thrown into the red zone, but I think you understood that.

These two situations:
"
From 2000-2005, there were 394 interceptions where the offensive line of scrimmage was in the red zone, i.e. the 20-yard line or closer. 28 of these were returned for touchdowns, or seven percent.

OK, what about when a team is backed into its own end?

From 2000-2005, there were 333 interceptions where the offensive line of scrimmage was on or behind the offensive 20-yard line. 57 of these interceptions were returned for touchdowns, or 17 percent.
"

Whats the total ratio of Picks vs Pick6s in the league?

My point is, if only 7 percent of INT's thrown in the redzone are run back, and a team is having 20% of theirs run back, it may not be luck, and may be more of an issue with running risky plays.

27
by Pat (not verified) :: Tue, 10/31/2006 - 6:41pm

zip: Also consider the quantization (rounding) error there - one divided by the number of samples. That is, if a guy has 4 TDs out of 14 picks, that's 28.6%... but he could only have had 14%, 21%, 28%, etc.

28
by Pat (not verified) :: Tue, 10/31/2006 - 6:44pm

My point is, if only 7 percent of INT’s thrown in the redzone are run back, and a team is having 20% of theirs run back

20% of what? Of 5? That's 1. You couldn't even get 7% with 5 interceptions. What about 10? You can't distinguish between 10% and 20% with only 10 interceptions.

No way you can distinguish anything with that small sample size. It's all identical.

29
by Pat (not verified) :: Tue, 10/31/2006 - 6:44pm

can't distinguish between = can't distinguish anything between.

30
by Pacifist Viking (not verified) :: Tue, 10/31/2006 - 6:49pm

Question on Estimated Wins (and sorry if this is explained somewhere on the website):

If the majority of teams have estimated wins higher than their actual wins (even if it's less than one win over), does that mean the system for determining estimated wins is a little inflated?

I'm just asking--obviously the system is pretty accurate.

31
by zip (not verified) :: Tue, 10/31/2006 - 6:50pm

Pat --

I know Mike Brown has the smallest sample size on that list, he's just a favorite around here so I threw him on there :)

Without knowing league average, it's pretty hard to draw any conclusions from that list (like I said, I was just screwing around). I think the most interesting number is Samari Rolle's 1/25, which is considerably lower that everyone else on that list... maybe he was too busy lateralling to Ed Reed :)

32
by Josh (not verified) :: Tue, 10/31/2006 - 6:51pm

I do think Rich has something with risky plays in the red zone. This is something TMQ has written about a lot recently and I agree with him on. The square-out inside the opponents ten seems to be very risky, I think that's one play in particular which is likely if intercepted to be returned for a TD, becuase often you have a CB on the sideline with a lot of green in front of him.
I don't know if this is anything DVOA can incorporate, and if it can't that doesn't take away much from the system. But I think Rich is correct that in this regard, returning the INT for a TD is not simply random luck.

33
by Pat (not verified) :: Tue, 10/31/2006 - 6:51pm

I think what people like Rich don’t understand is that the reason things are getting more conservative is because things are gettting better and more comprehensive. There is just factually less improvment to be made. In the past people came up with actual novel ideas often.

Yeah. My one suggestion so far that has ever been implemented was flipping the coefficient for variance in estimated wins when a team gets bad enough.

I think the best thing I've heard so far in the past few weeks, though, would be looking at the player type who intercepted the ball - so making an interception thrown to a CB worse than throwing an interception to an LB. That definitely could have an effect. If it's true, that is. I'm not so sure it is. There's a ton of bias in the fact that LBs don't play 30 yards deep.

34
by NF (not verified) :: Tue, 10/31/2006 - 6:52pm

So here's the tiers of teams as DVOA sees it (listed in groups in no particular order):

Really Bad Teams: Oakland, Houston, San Francisco, Arizona, Tennessee

Good Bad Teams: Cleveland, Seattle, New York Jets, Miami, City of Tampa, Detroit, Buffalo

Somewhat Average Teams: Washington, Minnesota, Green Bay, Carolina, Atlanta

Good-Average Intermediate Teams: New Orleans, Cincinnati, St. Louis

Good Teams: Denver, Pittsburgh, Indianapolis, Dallas, Jacksonville, Baltimore, Kansas City, New England

Great Teams: New York Giants, San Diego, Chicago, Philadelphia

So the best division in the NFL right now is the NFC East, with two great teams, one good team, and one average team. The worst division is the NFC West, with two really bad team, one good bad team, and one team between average and good.

35
by Donald (not verified) :: Tue, 10/31/2006 - 6:53pm

gay, gay, gay. this is garbage.

36
by Pat (not verified) :: Tue, 10/31/2006 - 6:53pm

#30:

If the majority of teams have estimated wins higher than their actual wins (even if it’s less than one win over), does that mean the system for determining estimated wins is a little inflated?

Nope. It means most teams have had their bye. :)

37
by Tones (not verified) :: Tue, 10/31/2006 - 6:54pm

I'll ask the thread before I e-mail Aaron... is there a place to see the game by game DVOAs of each team? As a NE fan, I get the feeling the defense and offense have improved nearly every week, but it's hard to verify at this point.

38
by Will Allen (not verified) :: Tue, 10/31/2006 - 6:55pm

Woohoo! As a Vikings fan, I gotta love the future schedule strength stats most of all! Not that it will be easy for the horned ones, however.

I actually fear the road game in 49erland this Sunday, given how the Niners were embarrassed by the Bears, and the Vikings have a short week following what might be a dispiriting loss. We'll have a better handle on Childress as coach by observing how well the Vikings respond. A loss, or even a nailbiter, might be cause for panic.

The Vikings will play two close games with the Pack, because those games are always close, even when there is a large talent disparity, which there isn't. The Packers are better than many perceive. They will likely split the series. The Vikings will likely beat Detroit in Detroit, and AZ, and the NYJ, in Minny. Miami, who I did think was slightly overrated by many before the season began, seems to have really bitten the dust to my surprise, and we'll have to see is Saban can resurrect enthusiasm with what is a fairly old team.

Hopefully the Rams will be out of it before the last game, in the Metrodome, because Bulger is the one remaining qb outside Favre who can spread the field and execute anywhere close to how well Brady did last night. The Vikes have little chance in Soldier Field, of course, since they will have to play extremely well to score even 10 points, and Grossman is unlikely to play as poorly as he did in the first game.

Hey, all I hoped for before the season began was that the Vikings not be mathematically eliminated after week 16, and right now, it looks pretty darned good.

38
by Rich Conley (not verified) :: Tue, 10/31/2006 - 6:55pm

Josh, that what I'm saying, and Pittsburgh seems to be doing a lot of it.

If I were to look for correlations, I'd look for offenses that give up a lot of pick-6s, not for defences that have a lot of them.

40
by Pacifist Viking (not verified) :: Tue, 10/31/2006 - 6:55pm

Pat, forgive me for not being a math guy, but how does the bye affect the estimated wins?

41
by Rich Conley (not verified) :: Tue, 10/31/2006 - 6:56pm

37.

Tones, their D has improved every game since Denver, they were aroudn +10% after that game.

42
by NewsToTom (not verified) :: Tue, 10/31/2006 - 6:57pm

Re #37
Per-game O-DVOA and D-DVOA based on final-season statistics are available in PFP. PFP 2006, currently available at your favorite bookseller (because it can't be your favorite if it doesn't have it available), has per-game results for each team for the 2005 season.

Since DVOA is a dependent statistic, the DVOA for each week would vary based on the weekly opponent adjustments. So, what looks like "DVOA for Week 1" in Week 2 won't necessarily look particularly similar to "DVOA for Week 1" in Week 9 or Week 17.

43
by DavidH (not verified) :: Tue, 10/31/2006 - 6:59pm

Estimated wins treats every teams as if they have played 8 games. So teams who have had their bye are going to be off.

44
by Pat (not verified) :: Tue, 10/31/2006 - 6:59pm

Pat, forgive me for not being a math guy, but how does the bye affect the estimated wins?

I'll forgive you for not being a math guy, but not for being a non-reading guy. :)

Teams that have had their bye week are projected as if they had played one game per week.

Right in the definition of estimated wins. :)

45
by DavidH (not verified) :: Tue, 10/31/2006 - 7:00pm

by the way, somebody asks this question every week, maybe that part of the estimated wins definition should be in bold or something.

46
by Richard (not verified) :: Tue, 10/31/2006 - 7:00pm

I need to find a way to get to these threads before Pat. That way, I won't have to start posts with, "Pat's right..."

Anyway, Pat's right. The sample size is outrageously small. Just because a team's TD/INT rate differs from the average doesn't mean that it differs in a way that is statistically signficant.

47
by NewsToTom (not verified) :: Tue, 10/31/2006 - 7:00pm

Re #40
I believe the way they do it is based on maximum possible number of games until all teams have had their bye already. So, after Week 9, when all teams will have played 8 games, I would expect EST WINS to be based on 8 games.

48
by CA (not verified) :: Tue, 10/31/2006 - 7:01pm

I have no empirical evidence to back this up, but I suspect that, at all levels of the game, teams and players that practice the "scoop and score" fumble recovery technique recover a lower percentage of fumbles but return a higher percentage for TDs than teams that practice the "fall on the ball" fumble recovery technique. If I am right, then fumble recoveries and defensive fumble recovery returns for TDs are necessarily not random, as they are at least partially determined by the strategy voluntarily employed by the team or player. That said, I would tend to guess that the use of one technique or another varies more by situation and position than by team or by players at the same position, so that the net effect of this choice of strategy is minor across the league.

49
by Jon (not verified) :: Tue, 10/31/2006 - 7:01pm

Does anyone know how you can translate these statistics, along with the homefield advantage into making informed pick'em picks? Is there any way to look at these numbers and recognize a poor spread?

50
by D (not verified) :: Tue, 10/31/2006 - 7:02pm

#17
I think the relative weakness of the second tier NFC teams should be relatively well accounted for because of the second order adjustment put into DVOA last year. If you recall in 2004 a far greater stratification existed with only one NFC team (the Eagles) to finishing in the top 11(!) of DVOA. It appears that the NFC has turned into a conference of have and have-nots with the handful of haves catching up to, and even surpasing, the AFC playoff contenders while the rest continue to lag.
Interesting note; of the top four NFC teams, only one currently has a starting QB who entered the season with more than a year and half as a starter.

51
by Will Allen (not verified) :: Tue, 10/31/2006 - 7:02pm

Yeah, you may be on to something there, Pat, but might another way of saying that be that some qbs toss interceptions to more dangerous areas of the field? Not that there is any hope of getting that info in the play by play, of course, so segregating by the position making the int, and seeing where that leads, makes sense.

52
by Pat (not verified) :: Tue, 10/31/2006 - 7:02pm

Jon: See here.

53
by Gerry (not verified) :: Tue, 10/31/2006 - 7:06pm

"maybe that part of the estimated wins definition should be in bold or something."

That is one idea. Another would be to do estimated wins as a function of a team's games played, so that it would be easier to eyeball-compare it to their actual wins.

54
by zip (not verified) :: Tue, 10/31/2006 - 7:07pm

Is it actually possible to extract the position of the intercepting player from the play-by-play? I don't think FO has a database with each player's name and position. You'd probably have to do it by hand using common sense.

That reminds me -- here's a question that I really should just email Aaron, but I'm too lazy -- does FO crunch all its numbers in Excel, or do you have any kind of SQL database involved? Makes me think there must be some terrifyingly complicated spreadsheet formulas involved :)

55
by Pat (not verified) :: Tue, 10/31/2006 - 7:07pm

I need to find a way to get to these threads before Pat. That way, I won’t have to start posts with, “Pat’s right…�

Avoid roughly 5:00. That's when I'm waiting for my wife to call me to tell me she's ready to leave. :)

I really do have to stress though. For anyone looking at percentages, you have to consider the sample size. Consider that any value at a minimum can freely be within 1/(sample size) above or below a number. If you want to be really safe, make it 3/(sample size).

56
by Jon (not verified) :: Tue, 10/31/2006 - 7:08pm

Thanks a lot Pat.

57
by zip (not verified) :: Tue, 10/31/2006 - 7:10pm

gay, gay, gay. this is garbage.

Man, you're not giving me much to work with here. Let's try it anyway:

DVOA is clearly ranked too high because it's garbage. Homophobic slurs are way better than this. ghey GAY g4Y!!

58
by Richard (not verified) :: Tue, 10/31/2006 - 7:11pm

Oh and quick ugly graph.

59
by Rich Conley (not verified) :: Tue, 10/31/2006 - 7:12pm

51 Will

Thats the thing, watching games, its clear that interception returns arent random. Where they happen, and who picks off the ball makes a big difference.

Can DVOA see this? I doubt it...that doesnt mean its not there, and doesnt mean its not repeatable.

It may be just selective memory, but I remember an awful lot of Drew Bledsoe Pick-6s on flat routes to the RB.

60
by Richard (not verified) :: Tue, 10/31/2006 - 7:14pm

55: Thanks for the tip.

61
by NF (not verified) :: Tue, 10/31/2006 - 7:14pm

#30: There is a simple answer to that. The estimated wins is projected as if the team had not had a bye week, and only 4 teams have not had one yet, so 28 teams are projected as if they played one more game. This goes away next week after all teams have played 8 games.

62
by NF (not verified) :: Tue, 10/31/2006 - 7:15pm

Note to self: refresh thread before posting.

63
by Fnor (not verified) :: Tue, 10/31/2006 - 7:17pm

#54: It's not in the PBP. It is in charting data, but, again, the turnaround is too slow. And yes, I realize that's in small part my fault ;P.

64
by Kal (not verified) :: Tue, 10/31/2006 - 7:21pm

That would be an interesting DVOA/DPAR metric on the QB - how many interceptions are thrown for TDs. I don't know if there is a correlation between a QB (or between an offense) and the number of INTs for TDs. I know that it's random if any specific defense does it, but I don't know if it's random if any particular offense does it.

My suspicion is that it isn't; pick 6s are big plays that meet highlights, but interceptions are not by themselves the biggest play ever, and I suspect we have a bad memory of this sort of thing. Still, if Aaron can check it out I'd love to see it.

65
by chris clark (not verified) :: Tue, 10/31/2006 - 7:23pm

re 58

Thanks for the GREAT graph Richard. It looks a little less cubic shaped than last week.

66
by zip (not verified) :: Tue, 10/31/2006 - 7:24pm

#59

Thats the thing, watching games, its clear that interception returns arent random

Are you sure? Can you up with any statistical evidence of this?

I'm not saying you're wrong, and I kind of think you're right, but if you can't provide some numbers to look at, we can't draw any conclusions from your anecdotes. "I remember Drew Bledsoe throwing a lot of pick-sixes" just isn't going to cut it.

Even if they aren't random, isolating something about interceptions that correlates to TDs, AND finding a way to extract that information from the play-by-play, is no small task.

67
by what do we have in common? (not verified) :: Tue, 10/31/2006 - 7:25pm

Chicago
Dallas
Kansas City
New England
New York Giants
San Diego

68
by Joe (not verified) :: Tue, 10/31/2006 - 7:26pm

Trying to decide if an INT to a CB or an LB is worse from a statistical grading perspective could be a nightmare. There's just too many ifs in there. Many INTs by a CB are because the WR fell down and/or misran the route after the QB threw the ball. Also you have to consider outside LBs vs inside LBs, outside LBs tend to get interceptions on screen passes, and intercepted screens turn into pick6 so often they might as well save all the running and make it by rule (kidding of course). Meanwhile inside LBs get a lot of picks due to tips at the line, so it's not entirely individual effort.

This of course is just one of the many reasons I read other people statistical analysis of football and attempt to make none of my own.

69
by Kal (not verified) :: Tue, 10/31/2006 - 7:27pm

On #64, I meant how many INTs are returned for TDs. Duh.

70
by MJK (not verified) :: Tue, 10/31/2006 - 7:30pm

Aaron's probably too busy to answer this, but maybe someone else knows.

How many iterations are there between VOA and DVOA? I know one of the major improvements in DVOA last year was iterating it again to get "2nd order" DVOA to compensate for really out-of-whack schedules, but does it stop there? Ideally, of course, one should probably infinitely iterate till convergence. Does DVOA converge after two iterations? 3? 4?

Probably not an issue this year, except for maybe New England, since none of the other really good or bad teams are playing a disproprotionately hard or easy schedule (i.e. Indy last year), but I was just curious.

If you're not a mathematical person and you didn't get what I'm asking, what I mean is that Team A outplays (on a play-by-play basis) Team B. So they get a high VOA. If Team B was really bad, then DVOA gets downgraded. But some of Team B being bad was that they had to play Team A, so maybe they're not as bad as their VOA implies. So their DVOA goes up from their VOA. But that means that maybe, since they're actually better, Team A's DVOA should not be downgraded as much. Etc.

71
by DavidH (not verified) :: Tue, 10/31/2006 - 7:31pm

so, anybody know where I can see data for how often each QB's INT's are returned for TD's? Cuz I'm not going through the play by play myself.

Hmm, maybe I should buy the FO package.

72
by Rich Conley (not verified) :: Tue, 10/31/2006 - 7:33pm

66

No, because I'm lazy. But I'd bet my income that a pick on a toss to a RB in the flat is significantly more likely to be run back for 6 than a pick 20 yards downfield from the LOS.

73
by Richard (not verified) :: Tue, 10/31/2006 - 7:33pm

Looking at z-scores... Chicago is the only team to crack 2.00. Baltimore, San Diego, Philadelphia and New York (Giants) made it over 1.00. Only Houston, Oakland, San Francisco, Tennessee and Arizona fall below -1.00. Teams ranked between 6th and 27th are all within one standard deviation of the league average Weighted DVOA.

74
by NewsToTom (not verified) :: Tue, 10/31/2006 - 7:33pm

Re #59/66
I think there are two things being confused here. One, what happens on a particular INT return isn't particularly random: players generally make a good return or not based on what sort of interception it is and their personal skills (Ronde Barber, or really any good DB with speed, jumping a short route is a good way to get a pick-6). Two, in what configurations interceptions fall is, using the available data, pretty random. Maybe, eventually, the charting project can produce good data, but it won't happen with the current setup. Frankly, to do a good job of it, you'd need full-22 camera angles and to chart all pass routes run by all receivers on every play, when a pass decision is made, what sort of coverage there is, and what happens. Thanks for volunteering.

75
by Thad (not verified) :: Tue, 10/31/2006 - 7:34pm

ok so i checked interceptions over the last three years.
All ints are sorted by position
dl 22.67 ints 3.67 tds 16%
lb 95.67 ints 13.67 tds 26.1%
db 404.0 ints 35.33 tds 14%
remember this is for the NFL as a whole
So defensive backs are much more likely to get ints but linebackers are more likely to score when they get their paws on one.

76
by Becephalus (not verified) :: Tue, 10/31/2006 - 7:36pm

I made a graph colored by divisions just for fun. Hope you al enjoy. I was a little disturbed by GB's proximity to MIN. An unacceptable situation for any true Minnesotan :)

77
by CA (not verified) :: Tue, 10/31/2006 - 7:38pm

Even if interception returns for TDs currently are essentially random, they don't necessarily have to be. I doubt most teams spend much time practicing blocking by their defenders in the case of a turnover, but what if they did? Well, for one, it probably would be a big waste of time, with the opportunity costs outweighing the benefits. But I could see teaching your starting defense special teams-style blocking techniques as potentially being an effective way to increase marginally the probability of returning an interception for a TD. Also, employing laterals probably slightly increases the probability of a return for a TD, although it also probably more than slightly increases the probability of an additional turnover.

78
by Becephalus (not verified) :: Tue, 10/31/2006 - 7:39pm

Damn Richard continues to beat me to my own idea :( First last week, now this week.

79
by B (not verified) :: Tue, 10/31/2006 - 7:41pm

76: Ooooh, pretty colors.

80
by Rich Conley (not verified) :: Tue, 10/31/2006 - 7:41pm

75

Where are you pulling your stats from, just out of curiousity.

81
by NF (not verified) :: Tue, 10/31/2006 - 7:41pm

#59: I agree with you that various plays will have interceptions thrown to some receivers returned for TDs more often. The way I see it, even if you could know the probability that an interception is run back for a long return for a play with an interception for a long return, what do you do with the information? Your looking at a subset of a subset of a subset, an interception to a type of receiver with a long return on a type of play. Even if you had a huge database that pinned down the likelihood of big returns to that accuracy, how often do you get multiple plays by one team that can be compared to one probability?

The idea that specific QBs are more likely to throw interceptions of a certain length or to a specific receiver, or just throw more interceptions that have a long return, is intriguing. Stats like "highest interceptions per pass in the 4th quarter in close games (1500 career passes minimum)" and the difference between that and interception per pass, or average length of intercepted throws would be interesting.

When Harrington/Boller/Plummer/Favre/Losman next throw 4+ interceptions, it would be neat to read a Too Deep Zone examining the interception tendencies of post-1978 NFL quarterbacks with major careers.

82
by Will Allen (not verified) :: Tue, 10/31/2006 - 7:44pm

Rich, I suspect you may be right, but a won-loss record doesn't establish it statistically, and if it can't be established statistically, it can't be reflected in DVOA. If DVOA is to be discounted in a non-arbitrary manner, it is necessary to have hard data to show where it is going off course, and a won loss record comprising a sample size of a single team's seven or eight games, or even 16 games, doesn't do so.

Aaron says he has run a very large number of simulations with all sorts of variables, and the current DVOA is what produces rankings which best correlates with outcomes. Does this mean DVOA is perfect? Of course not, but it does mean that we have to come up with something more detailed and comprehensive than, "The Steelers are #11 in DVOA, even though they are 2-5, and Roethlisberger has had some ints returned for touchdowns, so that means DVOA is flawed in how it evaluates ints".

Yes, I have some suspicions that something might be missed right now in how ints are treated, but I certainly can't say so with any degree of certainty.

83
by NewsToTom (not verified) :: Tue, 10/31/2006 - 7:45pm

#58 is clearly better than #76 because it uses weighted DVOA, which ranks my beloved Titans #31, instead of DVOA, which ranks the Titans #32. Because they're the Houston Oilers, Houston Oilers, Houston Oilers #1.

Sorry, I couldn't resist.

84
by Bobman (not verified) :: Tue, 10/31/2006 - 7:45pm

RE 26/32 and Rich C's risky play points, can we look at the red zone INT-TDs from the passer's perspective?

It seemed to me, years back (say his first three years), that every time Peyton Manning threw a pick, it was run back for six. Clearly this is not the case, but I can't remember the last time it has happened (i.e. since he's become super as opposed to just darn good), which leads me to this: Do rookie or poorly performing QBs have higher TD/INT ratios? My money says yes they do, and it's probably because they are throwing riskier balls, ill-advised passes into coverage where there's no safety net.

It's the flat pass I am thinking about, where if it's not ankle high, or way outside pass where only the intended receiver can get it, then the DB can catch it in his breadbasket in stride, streaking the other way for a TD.

(Now of course a pick returned for six would affect a QB's DVOA, but escluding that factor...) So if there is a correlation between the QB and the TD/INT ratio, then it's not random. But it's more a factor against the offense than in favor of the defense.

Simplest way to look just for quick confirmation would be the seemingly large number of rookies taking snaps this season: how does their INT/TD ratio look compared to the rest of the league?

New can of worms?

85
by ChrisFromNJ (not verified) :: Tue, 10/31/2006 - 7:46pm

#70:

Actually, Chicago is the team with the real outlier in terms of easy schedules, at almost -10% more than the next easiest. Considering that they're still in first by a huge margin, I think it's fairly safe to say that a) the iterations are working, and b) Chicago is just ridiculous this year.

86
by Bobman (not verified) :: Tue, 10/31/2006 - 7:47pm

Shit, Kal, I missed your comment at 64. Very similar point. Aaron's right, these threads are exhausting.

87
by Will Allen (not verified) :: Tue, 10/31/2006 - 7:57pm

Again, Rich, I agree that some routes are more likely to have a int returned for a td, compared to other routes that produce ints, but the question is if it can be established statistically that quarterback x's ints are significantly more likely to be that especially harmful int, compared to quarterback y's ints. Yes, bad qbs throw more ints than good qbs, but I don't think anybody has established yet that a higher percentage of a bad qbs ints are more likely to be returned for tds.

88
by Rich Conley (not verified) :: Tue, 10/31/2006 - 7:59pm

Will, some of the Pittsburgh thing may be accounted for in the fact that DVOA correlates to Points, not wins.

Mind you, I'm cherry picking here, but Pittsburgh has scored

157 points and given up 145. That should make them slightly better than baseline, IE maybe 5% VOA (theyre 7.5 actually)

Now, if you take out that KC game, we're looking at a team thats scored 112 points, and give up 138, which is about exactly where cleveland is.

I've just got this feeling that the KC game is a huge part of Pittsburgh's DVOA, and that its going to prove to be an outlier.

89
by zip (not verified) :: Tue, 10/31/2006 - 8:05pm

All this talk about TDs and INTs is trending away from DVOA. Remember, the question is basically "Do INTs returned for a TD have any predictive value for the offense or defense?"

Clearly screens that get picked are going to go the other way more often than deep balls that get picked. But what does that mean from a predictive standpoint? Maybe DVOA overvalues screen plays because it's removing the worst case scenario from any offensive pass (pick 6) in a situation where that worst case scenario is significantly more likely to happen?

My hunch is that no team runs enough screens for this to make a difference, and no team is good enough at defending screens to make a difference. The sample data is so small we'd need years of charted screens to really be able to say anything...

Of course now I'm really curious. Is the FO game charting data complete from last season? Is it available to the public, even at a price? I'd love to be able to look at every pick six that happened last year.

90
by Tones (not verified) :: Tue, 10/31/2006 - 8:09pm

Thanks for the feedback on game by game DVOA.

I am kind of curious about doing some of my own number crunching on the NFL's PBP data.
What are the legal implications of using the PBP data on NFL.com? I assume if I use it and then publish the results, that's not a huge deal. But what about a site like this that generates revenue via advertising? Or does the NFL just consider the PBP data to be public domain?

91
by Thad (not verified) :: Tue, 10/31/2006 - 8:12pm

They’re doing everything right that you historically have needed to win games, and still losing.

Not evrerything.
Over the past 10 years there have been 311 team.
17 of them have run the ball 39% of the time.
5 of those teams had winning records
44 of those teams have run the ball 38-40% of the time, 6 of them had winning records,(three went to the Super Bowl.)
On the other hand, there have been 45 teams that have won 12 or more games in the past 10 years.
They ran the ball 47% of the time.
They gave up an average of 31.5 sacks.
The Eagles are on pace for 40
The teams that won 12 games threw an average of 198 incomplete passes.
The Eagles are on pace for 232.
Not bad, but not great.

Now I am not saying the Eagles can't win, they have a very good offense and McNabb is just having a super year.
But historically most good teams, the really good ones, pass the ball effectively and run a lot, the Eagles pass the ball very effectively.

92
by NewsToTom (not verified) :: Tue, 10/31/2006 - 8:16pm

Re #89
The complete FO game charting data from last season is available on the FO Online Store. Check out under About This Site. It's $60. I've been contemplating getting it myself, but haven't yet.

#90
So long as you don't try to publish the raw PBP data collected by the NFL in and of itself, I wouldn't worry about it. Sports stats are public domain.

93
by Will Allen (not verified) :: Tue, 10/31/2006 - 8:17pm

Yeah, Rich, Aaron has said repeatedly that early season blowouts skews DVOA significantly, and we are still only in week eight, so less than half the games have been played. Who knows? Maybe Pittsburgh will go 5-11, but I think 8-8 is far more likely, and having an 8-8 team among the top 12 or so in DVOA certainly is not a strong indicator of a weakness in DVOA.

94
by B (not verified) :: Tue, 10/31/2006 - 8:28pm

77: By that logic, teams with lots of starters who play special teams should be better at getting/preventing Int TDs than average teams.

95
by Thad (not verified) :: Tue, 10/31/2006 - 8:32pm

Stats inc.

96
by Jim (not verified) :: Tue, 10/31/2006 - 8:41pm

Did anyone other than me think that Belichick continuing to employ 5 wides every down with 8 minutes to go and up 31-7 was kind of bush league. The Pats completely embarrassed the Vikings on national tv - didn't he fear for his QB's safety?

97
by AD (not verified) :: Tue, 10/31/2006 - 8:44pm

#91. The "run to win" fallacy, just thinly disguised.

98
by AlexDL (not verified) :: Tue, 10/31/2006 - 8:45pm

resorted data table and superimposed graph of WEI. click on name for a look.

99
by admin :: Tue, 10/31/2006 - 8:50pm

Golly, almost 100 comments before the commentary went up on FOX. Click my name and enjoy some good Colorado-specific humor.

100
by Pat (not verified) :: Tue, 10/31/2006 - 8:52pm

But historically most good teams, the really good ones, pass the ball effectively and run a lot, the Eagles pass the ball very effectively.

Sigh. Philly hasn't run the ball much this year because they've never been ahead significantly. In the games where they have been ahead significantly (Houston and Green Bay) they've run about 50% of the time, and it happens in the second half. Can we please let this inane idea that you have to run to win die?

101
by NF (not verified) :: Tue, 10/31/2006 - 8:52pm

#91: It's been well established by this site that teams that run the ball a lot are not necessarily good. Teams that are good run the ball a lot because they are always running out the clock at the end of games they are winning. SEE 2004-2005 Pittsburgh Steelers, 2004 New England Patriots.

Also, incomplete passes are obviously a proportion of passes thrown, and if you are throwing the ball a lot in the second half trying to catch up, you will throw a lot more incomplete passes than teams that have a lead in the second half and are running the ball mostly to run the clock out.

102
by Thad (not verified) :: Tue, 10/31/2006 - 8:55pm

2005 int ret for tds
Collins, Kerry 4
Favre, Brett 3
McMahon, Mike 3
McNair, Steve 3
Delhomme, Jake 3
Manning, Eli 2
Bledsoe, Drew 2
Wright, Anthony 2
Brooks, Aaron 2
Carr, David 2
Brady, Tom 1
Green, Trent 1
Brees, Drew 1
Holcomb, Kelly 1
Maddox, Tommy 1
McCown, Josh 1
Volek, Billy 1
Smith, Alex 1
Cassel, Matt 1
Harrington, Joey 1
McNabb, Donovan 1
Bouman, Todd 1
Simms, Chris 1
Boller, Kyle 1
Brunell, Mark 1
Bollinger, Brooks 1
Losman, J.P. 1
Banks, Tony 1
Palmer, Carson 1
Griese, Brian 1
Garcia, Jeff 1

103
by Thad (not verified) :: Tue, 10/31/2006 - 8:58pm

2004 int ret for tds
Feeley, A.J. 5
Griese, Brian 3
Delhomme, Jake 3
Collins, Kerry 3
Testaverde, Vinny 3
Palmer, Carson 3
Fiedler, Jay 2
Rattay, Tim 2
Harrington, Joey 2
Volek, Billy 2
Dorsey, Ken 2
Roethlisberger, Ben 2
Rosenfels, Sage 1
Henson, Drew 1
Blake, Jeff 1
Favre, Brett 1
Plummer, Jake 1
Hutchinson, Chad 1
Bulger, Marc 1
Brooks, Aaron 1
Brady, Tom 1
Hasselbeck, Matt 1
Brunell, Mark 1

Garcia, Jeff 1
Bledsoe, Drew 1
McCown, Josh 1
Leftwich, Byron 1
Carr, David 1
Holcomb, Kelly 1
Green, Trent 1
Kitna, Jon 1
Johnson, Brad 1
Maddox, Tommy 1

104
by Thad (not verified) :: Tue, 10/31/2006 - 9:00pm

2003 int ret for td's
Johnson, Brad 4
Harrington, Joey 4
Maddox, Tommy 3
Blake, Jeff 3
Carr, David 3
Ramsey, Patrick 2
Stewart, Kordell 2
Fiedler, Jay 2
McNair, Steve 2
Kitna, Jon 2
Holcomb, Kelly 2
Couch, Tim 2
Delhomme, Jake 2
Favre, Brett 2
Plummer, Jake 1
Brooks, Aaron 1
Culpepper, Daunte 1
Green, Trent 1
Griese, Brian 1
Garcia, Jeff 1
Hasselbeck, Matt 1
Collins, Kerry 1
Carter, Quincy 1
Bulger, Marc 1
Kittner, Kurt 1
Brees, Drew 1
Testaverde, Vinny 1
Manning, Peyton 1
Volek, Billy 1
Palmer, Jesse 1
McCown, Josh 1
Pennington, Chad 1
Brady, Tom 1
Bledsoe, Drew 1
McNabb, Donovan 1
Leftwich, Byron 1

105
by Will Allen (not verified) :: Tue, 10/31/2006 - 9:02pm

The Vikings run a lot because they have a pretty good line, but also because their wide receivers and qb have huge deficiencies, and the only way for them to pass with any effectiveness is to make play action as credible as possible. Thus, when they fall behind by multiple scores to a team with any significant talent, and have to pass a lot, they have no chance at all. If they had better talent at quarterback, and especially at receiver, they would throw more.

106
by Thad (not verified) :: Tue, 10/31/2006 - 9:02pm

2002 int ret for td's
Harrington, Joey 3
Warner, Kurt 3
Lucas, Ray 2
Kitna, Jon 2
Maddox, Tommy 2
Griese, Brian 2
McMahon, Mike 2
Brunell, Mark 2
McNair, Steve 2
Culpepper, Daunte 2
Redman, Chris 1
Johnson, Brad 1
Johnson, Doug 1
King, Shaun 1
Hasselbeck, Matt 1
Brady, Tom 1
Bulger, Marc 1
Miller, Jim 1
Gannon, Rich 1
Carr, David 1
Plummer, Jake 1
Collins, Kerry 1
Wuerffel, Danny 1
Smith, Akili 1
Couch, Tim 1
Ramsey, Patrick 1
Testaverde, Vinny 1
Hutchinson, Chad 1
Brees, Drew 1
Bledsoe, Drew 1
Martin, Jamie 1
Peete, Rodney 1
Brooks, Aaron 1

107
by Fnor (not verified) :: Tue, 10/31/2006 - 9:07pm

Aaron's point about zone defences is interesting and very well taken. I think the real culprit is less the zone scheme itself and more the problems of a zone scheme without a powerful enough pass rush behind it.

108
by BadgerT1000 (not verified) :: Tue, 10/31/2006 - 9:08pm

Really enjoyed the "Top Ten" list.

Though instead of a real answer for number 1 I would have enjoyed seeing "Plummer sucks" or some other rude variation.

But that's just me.

109
by AlexDL (not verified) :: Tue, 10/31/2006 - 9:14pm

so, Aaron you're basically saying that Indy has no chance to win the game because they play the cover2 and that New England plays the 3-4?

110
by Carlos (not verified) :: Tue, 10/31/2006 - 9:17pm

#96 - I don't know whether it was "bush league," but I'm not at all sensitive to claims of running up the score in professional sports. But I was shocked Brady was still in. He took at least one big, completely unnecessary hit. What if he'd gotten a concussion?

111
by Richard (not verified) :: Tue, 10/31/2006 - 9:23pm

109: It sounds about right, doesn't it?

112
by Thad (not verified) :: Tue, 10/31/2006 - 9:26pm

So now I looked at number of ints per position and ave ret yardage per int.
I looked at the past 4 years, 128 teams
dl 22.5 ints 13.6 y/ret
lb 96.75 ints 14.85 y/ret
db 404.5 ints 14.85 y/ret

113
by Josh D (not verified) :: Tue, 10/31/2006 - 9:28pm

Love your statistical analysis and you should just ignore ignorant comments which you have heard repeatedly. Hopefully the following will not be ignorant nor repetitive.

Have you thought about ranking your tems by DVOA with a variance weighting? From my reading on your site, it does not appear that variance is part of the DVOA rankings. In short, I would rather have a team with slightly lower DVOA if it was more consistent (i.e. Denver over Pittsburgh currently).

Secondly, I believe you measure variance on a per play basis, however it may make sense to measure variance on an aggragated game-by-game basis versus play-by-play basis because I think I would rather have a team which had one bad game out of four versus one that has the equivalent of a bad quarter every game. (...or it may be the other way around) Have you compared team's variance at these different levels (on a per play basis versus a per game basis) to see if it tells a different story?

114
by Bobman (not verified) :: Tue, 10/31/2006 - 9:32pm

RE the commentary. Aaron, Indy's problem at NYJ was not the Jets D. It was the Colts D and a 103 yard KR for a TD.

"eventually" sure looks like a line in the sand. I'm gonna call Rodney Harrison and tell him you're respectin' him. That'll take his edge off. Just enough....

115
by JonL (not verified) :: Tue, 10/31/2006 - 9:35pm

The KC comment got me thinking - has there been any tracking of whether specific coaches overuse their #1 running backs, regardless of what team they're with? I'm thinking of a baseball parallel with Dusty Baker and overusing young pitchers.

In other words, did Herm Edwards also use Curtis Martin too much in NY?

116
by Thad (not verified) :: Tue, 10/31/2006 - 9:36pm

re 100
ok Pat,
first of all I did not say have to.
Second of all, do you have any evidence that it is inane?
I looked at 10 years for this data, what do you have.?
The 2004 Eagles ran the ball 39% of the time, won 12 games, and at the end of the third quarter of the Super Bowl were tied.
I am not saying it can't be done and I am certainly not saying that this years Eagles won't.
I am saying that it is done infrequently.
Also they had a fine lead over the Giants and had plenty of chances to run.

117
by Marko (not verified) :: Tue, 10/31/2006 - 9:37pm

I don't think that fumble recoveries are random, nor do I think returns for touchdown (whether fumbles or interceptions) are random. Under Lovie Smith, the Bears really focus on creating turnovers and scoring on returns. They do practice blocking on returns.

The Bears swarm to the ball on defense and actively try to strip the ball, with great success. They do the same thing on special teams. I don't know the exact number of fumbles they have forced and recovered this year, but I know that they are definitely among the leaders in those categories. And we all remember the two touchdowns they scored on fumble returns during Arizona's Monday Night Meltdown. Those weren't random; they were looking to score on defense, which was the only way they were going to win that game. Arizona knew that, but the Bears still were able to do it.

Since Lovie Smith became head coach in 2004, the Bears lead the league in defensive touchdowns. This is not a coincidence. If I recall correctly, they also led the league by a wide margin last year in interception return yardage. Again, not a coincidence.

118
by Bobman (not verified) :: Tue, 10/31/2006 - 9:37pm

109, No, what Aaron's saying is Ken Lucas's groin is this season's MVP. (Mrs. Lucas isn't commenting.)

119
by Pat (not verified) :: Tue, 10/31/2006 - 9:37pm

#113:

From my reading on your site, it does not appear that variance is part of the DVOA rankings.

That's what Estimated Wins does.

In short, I would rather have a team with slightly lower DVOA if it was more consistent (i.e. Denver over Pittsburgh currently).

Well, it depends. If you've got a team which fluctuates between 100% DVOA and 40% DVOA, that's hugely inconsistent. Inconsistently beating the utter crap out of teams.

And variance is also better for bad teams, as well.

120
by Pat (not verified) :: Tue, 10/31/2006 - 9:39pm

I looked at 10 years for this data, what do you have.?

This.

121
by Pat (not verified) :: Tue, 10/31/2006 - 9:42pm

So now I looked at number of ints per position and ave ret yardage per int.
I looked at the past 4 years, 128 teams

You'd really want to look at net yardage per interception - that is, the yardage change in possession. A 14 yard runback for a DB if it's on a 40 yard pass is slightly different than a 14 yard runback by an LB on a 14 yard pass.

122
by Kevin (not verified) :: Tue, 10/31/2006 - 9:42pm

#100... What about the NYG/PHI game? The Eagles were up 17 points with 4 minutes left in the third quarter. From that point on, they ran the ball 11 times for 20 yards. Those twelve carries included a fumble and a stop on 4th and 1. Good teams are capable of running for three yards on two tries.

Please forgive me Aaron, but I must mention this again. The Plaxico Burress fumble recovery didn't end the game!!! The Giants still needed to score 10 more points after that fumble recovery just to tie the game. The Eagles could've scored on one of the 4 drives they had after the score became 24-14. The Eagles got the ball at midfield in overtime, but couldn't even get a single first down. Luck isn't why they lost to the Giants. Their complete inability to make even one play at several key junctures in the game was why they lost.

123
by Fnor (not verified) :: Tue, 10/31/2006 - 9:48pm

#117: Everyone runs those drills. Everyone drills their players to chase the ball. Many teams run a "swarming" offence. You talk about a couple years with one team as definite proof that all those things aren't random, and you don't really give any evidence that any of them aren't random.

124
by Blackmallow (not verified) :: Tue, 10/31/2006 - 9:50pm

Come on Aaron, that first int in the Saints game was right in the receiver's hands, but he heard footsteps and turned before he caught the ball. Result: the ball hits his helmet and bounces 8 feet in the air in the defender's hand.

Does FO count bs ints like that to the qb's record or is it seperated out like yards after the catch is?

125
by Josh D (not verified) :: Tue, 10/31/2006 - 9:54pm

#119

Pat, thanks for your answer.

If estimated wins includes variance, shouldn't the teams be ranked by estimated wins then (or at least show a second ranking by those results)? How did they determine the variance weighting in estimated wins?

126
by Thad (not verified) :: Tue, 10/31/2006 - 9:54pm

re 121
Yeah Pat on that we agree! :)
I don't have that info otherwise i would have listed it.

127
by Julio (not verified) :: Tue, 10/31/2006 - 9:57pm

It's pretty obvious what the problems with DVOA are: it takes into account things that are only indirectly related to the outcomes of games. The current offensive/defensive rankings are essentially what you would get if you ranked teams by yds/game allowed/earned, rather than pts/game allowed/earned. I think it's pretty obvious to everyone that up to this point, Chicago, Indy and NE are the best teams (and maybe Denver too).

128
by Thad (not verified) :: Tue, 10/31/2006 - 9:58pm

re 120
No, I said nothing about establishing the run.

129
by Starshatterer (not verified) :: Tue, 10/31/2006 - 10:01pm

Bobman (#114 )--

It's funny that both the Colts' and Pats' have an offensive strength to match against the opposing defense's weakness. The Colts' weakness might be weaker, but their offensive strength sure looks stronger, too.

Interesting matchup. DVOA + homefield made me take Denver in my picks pool for week eight. I suspect the same will suggest picking New England this week. Beatpaths slightly favor Indy (#1 vs tied for #3 with Chicago).

You have to careful with Rodney Harrison. Call him, and he'll peg you for an Indy fan, and take it as a sign of disrespect that you'd think a show of respect would throw him off him game. The man can play the disrespect card.

130
by Kevin (not verified) :: Tue, 10/31/2006 - 10:07pm

#127. How is it obvious to everyone that NE is among the best teams in the NFL right now? It can't be based on performance against top/winning competition. The best team they've beaten are Minnesota and a Cinncinati team that's an end zone INT away from losing three straight home games. Your opinion must be solely based on the fact they're the Pats with Brady and Belichick. For example, the NYG are better than NE based on the quality of opponent that they've faced this season.

131
by Pat (not verified) :: Tue, 10/31/2006 - 10:19pm

No, I said nothing about establishing the run.

It's this part:

The correlation between first quarter rushing attempts and team wins is a measly .171. That means there is almost no connection between running a lot in the first quarter, and winning a lot of games. The correlation between fourth quarter rushing attempts and team wins, on the other hand, is .750. That’s a sizeable relationship.

The reason why teams that rush more win more is because you rush when you're trying to keep a lead. Virtually the entire correlation between winning percentage and rush percentage is contained within fourth-quarter rushing.

That is, the data that you gave before is entirely due to the fact that those teams were rushing in the fourth quarter with a lead.

See here.

Those twelve carries included a fumble and a stop on 4th and 1. Good teams are capable of running for three yards on two tries.

Actually, no - read the article I linked in #120. It doesn't matter how much you rush for in the fourth quarter. It just matters that you run.

Of course, running for positive yardage makes it easier for you to keep running, but you don't have to be able to run well in order for it to matter.

Incidentally, the Giants have a really, really good run defense. There's a reason that Philly wasn't able to run on them.

132
by Richard (not verified) :: Tue, 10/31/2006 - 10:20pm

127: Getting people with the same biases to come to the same conclusion doesn't make said conclusion obviously correct.

133
by Pat (not verified) :: Tue, 10/31/2006 - 10:21pm

If estimated wins includes variance, shouldn’t the teams be ranked by estimated wins then (or at least show a second ranking by those results)?

I'm guessing Aaron vetoed that due to the complexity of it. People are used to the concept of power rankings. But I've always kindof been for using estimated wins rather than weighted DVOA, but... eh. Honestly, not much difference. You'll get people who complain either way. :)

How did they determine the variance weighting in estimated wins?

Click on Estimated Wins. Click on "here" in the glossary.

134
by manning,e (not verified) :: Tue, 10/31/2006 - 10:27pm

shouldnt the eagles and giants past opponent strength of schedule put them ahead of chicago? While Chicago has a dvoa 11 percent higher then the giants they have been accumlating that against opponents 23 dvoa percent weaker. The giants two loses came against two opponents who at the time had high dvoa. The seattle team the giants had just "lost" the Superbowl and had their starting qb and runningback. While Chicago has been almost losing every week to a poor class of opponent, vikings and cardinals. Giants did what the Bears did (massive comeback)only the bears did it against the currently number 31 dvoa team while the giants did it against the number 3.

135
by anonymous coward (not verified) :: Tue, 10/31/2006 - 10:30pm

I think that some of the confusion might be over what exactly DVOA measures. I am probably wrong, but doesn't DVOA measure how the teams are likely to perform moving forward, not how well they have played? Since that is the case, I don't think the argument is that Philadelphia and Pittsburgh have played well so far, instead it is that they have done nothing to indicate that they won't play well in future games.

136
by B (not verified) :: Tue, 10/31/2006 - 10:31pm

134: Opponent adjustments aren't at full strength yet, because there isn't enough information to know just how good or bad everybody is.

137
by Pat (not verified) :: Tue, 10/31/2006 - 10:32pm

#134: the D in DVOA means "defense-adjusted". The numbers are already adjusted for opponent.

138
by ChrisFromNJ (not verified) :: Tue, 10/31/2006 - 10:37pm

I know this isn't a DVOA question, and is therefore deletable spam, but my browser appears to be crashing whenever it comes into contact with FOX Sports. Is anyone else having the same problem?

139
by Fnor (not verified) :: Tue, 10/31/2006 - 10:46pm

Chicago nearly losing? Wha? They've only nearly lost two games. They probably should have lost those games (saved by some awesome Manning-luck vs MIN, against the Cardinals they were saved by... uh... the Cardinals), but that doesn't erase the stomps they've had along the way.

If you want to criticize certain things about the Bears, go ahead (I sure as heck have). If you want to say they've benefitted from a heaping helping of luck (looking at you, Rexy), go ahead. But you can't just completely ignore whole games!

140
by MJK (not verified) :: Tue, 10/31/2006 - 10:49pm

Josh D (113)

Interesting points. I'm not sure that I agree that I'd rather have a team that has one bad game out of four than one that has one bad quarter per game...in the first case, you're likely to go 3-1, and in the second case 4-0 is still a pretty good chance--one bad quarter and three good quarters will lead to a win more often than not. But I definitely agree that I'd rather be a team that is consistently very good than one that is, on average, slightly better, but wildly inconsistent. At least, I'd perfer the former come playoff time.

Incorporating variance into a rating seems like a personal choice, though, because I'm sure there are some people (namely, fans of variable teams like the Chiefs and the Jags) that would disagree.

Also, how would you do it? Something along the lines of defining a normally distributed random variable for each team's DVOA on any given day with the appropriate mean and variance, and then calculating the pairwise probability that each team is better than each other team? I'm not really a probability person...

141
by Thad (not verified) :: Tue, 10/31/2006 - 10:53pm

Virtually the entire correlation between winning percentage and rush percentage is contained within fourth-quarter rushing.

That is, the data that you gave before is entirely due to the fact that those teams were rushing in the fourth quarter with a lead.

That was my original point, I guess I did not explain it very well.

142
by Fnor (not verified) :: Tue, 10/31/2006 - 10:56pm

#141:

Then your entire argument is conclusory. "The Eagles don't win because they don't run enough. They don't run enough because they don't have comfortable leads. They don't have comfortable leads because they're losing."

143
by B (not verified) :: Tue, 10/31/2006 - 10:58pm

139: If they keep wearing their orange jerseys, I can. Stomp or no, I refuse to acknowledge such bad ascetic sense.

144
by Fnor (not verified) :: Tue, 10/31/2006 - 11:01pm

#143:

Yeah, I think they own too much crap, too.

("ascetic." hehe)

145
by dryheat (not verified) :: Tue, 10/31/2006 - 11:03pm

#96 Not bush league at all. The Pats were playing against the clock, and they were better equiped to keep the clock running, and getting first downs, via the passing game. Aside from the interception, I don't think Minnesota's secondary defensed a ball all game. Brady's incompletions were drops or throw aways under pressure.

146
by kleph (not verified) :: Tue, 10/31/2006 - 11:04pm

while i also agree these comment threads have become exhausting, i always make sure to skim them no matter how many comments have accrued. i learned my lesson after missing out on the legendary ROBO-PUNTER thread.

147
by doktarr (not verified) :: Tue, 10/31/2006 - 11:20pm

Aaron, who came up with that Denver humor? That stuff is hilarious. Do you have a local plant?

RE: Variance discussion,
If you think variance should play a role, then you want the estimated wins stat.

RE: Bobman and Starshatterer,
I think it's reasonable to expect that the Pats offense will tear through the Colts defense. I'd be surprised if the Pats defense really slows down the Colts offense, but I'd expect them to do a little better than DVOA would suggest (i.e. poorly, but not +31.6%).

148
by Thad (not verified) :: Tue, 10/31/2006 - 11:24pm

re 141
no, I was responding to this

They’re doing everything right that you historically have needed to win games, and still losing.

They are certainly doing some things right,
not all.

149
by Will Allen (not verified) :: Tue, 10/31/2006 - 11:31pm

Consistently kicking the living hell out of mediocre and bad teams is a very good indicator of quality. The fact that Chicago played one bad game against Arizona, and that Minnesota played them essentially even, doesn't outweigh the fact that they have just totally crushed everybody else. They are an extremely good team with a high degree of variability at the most important position, which is why it ought to be very interesting to see them in the playoffs.

It wouldn't surprise me in the least to see them get really thumped in the Conference championship game, maybe even not even in the divisional round, depending on personnel match-ups, but I wouldn't be shocked to see them thump three teams in succession either.

Yeah, there are a lot real dogs at the bottom this year, but I also think the the matchups at the top are going to be more interesting than normal come playoff time. It'll be fun to watch.

150
by dbt (not verified) :: Tue, 10/31/2006 - 11:35pm

Will, I'm very much looking forward to the Giants and Patriots as barometer games, especially on the road.

151
by Will Allen (not verified) :: Tue, 10/31/2006 - 11:46pm

dbt, what is crazy about the Giants matchup is that they have a high degree of variability at qb as well, albeit not as high as the Bears, so we might not learn a damned thing when they meet, and if they meet again in the playoffs, it wouldn't be too surprising to see their performances reverse, even if the first game resulted in one of them getting blown out.

152
by CA (not verified) :: Tue, 10/31/2006 - 11:59pm

I guess I am the only one who likes the Bears' orange jerseys. Football uniforms are one of the few places where the color orange just seems right. But, then again, I also really liked the much-maligned Tampa Bay creamsicle uniforms, and that's going to disqualify my opinion on all uniform matters in a lot of people's eyes. Tell you what: I'll accept that the Bears never wear the orange jerseys again if the Broncos are required to return to their classic orange home jerseys permanently.

153
by BadgerT1000 (not verified) :: Tue, 10/31/2006 - 11:59pm

As much as it pains me, I think the Bears are the perfect storm of players with experience, being well-coached, at their physical peak, and highly motivated to have success. If they get mediocre play from the QB position they will be in a good position to win a game. If they get GOOD quarterback play they WILL win a game.

Just look at the defense. Guys all over the place who have multiple years of experience but still under 30 and completely in tune with each other on the field. This guy or that guy can sometimes take a risk or freelance because he knows, KNOWS, that players X and Y are here and HERE and should he miscalculate those guys will still be able to make a play.

A lot of guys around the league want to win. The Bears as a group are playing like they HAVE to win. HAVE TO. As in MUST. No exception.

I don't think this is a hot streak or jsut a case of mauling bad teams. Like I wrote, this is the perfect storm of an NFL team. And I don't think it will be over until it destroys everything in its path.

154
by Rollo (not verified) :: Wed, 11/01/2006 - 12:05am

It would be neat to see variance broken down by offense and defense. Jacksonville has been an interesting team to follow. I think the high variance is part coaching - the Jags have a disturbing tendency of laying down the week after a big victory and also consistently blow off the Houston game - and part personnel. When the offensive line is blowing people out of the water, and a rejuvanated Fred + Maurice Jones Drew can slice and dice, the defense (in particular the deplted D-line) gets plenty of rest and plays lights outs, and Jacksonville looks almost Super.

155
by D (not verified) :: Wed, 11/01/2006 - 12:14am

#154
It is, check the offense and defense sections.

156
by Ryan Mc (not verified) :: Wed, 11/01/2006 - 12:53am

Any obvious reasons for the massive discrepancy in Cincy's first half versus second half offensive performances?

157
by Ryan (not verified) :: Wed, 11/01/2006 - 1:45am

You know what I find interesting about post 134? If the poster had looked at the UNAJDUSTED total VOA he would see that the bears VOA is significantly higher than the other teams, and that the STRENTGH of schedule is pulling their total DVOA Way down...What's the point of reading just half of the statistics that are listed? He would have answered his own question. FCOL The Bears' VOA is almost 40% higher than PHI's, but their DVOA is only 11 higher, which is why they're still ahead in DVOA but by not quite as much. Oh well; just needed to rant on people who don't pay attention.

158
by Ryan (not verified) :: Wed, 11/01/2006 - 1:49am

BTW. to the poster who asked how many fumbles CHI has recovered, it's an exceedingly and ridicuously high 14, which would have tied them for sixth most fumbles recovered last season. They're on pace to recover 32 fumbles, and I'll eat my hat if they actually do that.

159
by MdM (not verified) :: Wed, 11/01/2006 - 1:58am

Is it just me, or are there are much higher than normal amount of people leveling criticism at DVOA lately? I mean, it seems like open season. Not that it shouldn't be discussed, but seems like every sentence is "It seems a weakness of DVOA is..."

A lot of the comments seem to be way off the mark, imo.

By the way, Pat, how the heck do you know so much about it? How come your writing isn't green? :)

160
by Andrew (not verified) :: Wed, 11/01/2006 - 2:03am

ChrisfromNJ:

My browser at work is constantly crashing on FoxSports.com. At home, no problem.

Rich Conley and Aaron:

Just from watching games, it is painfully obvious that certain defensive backs are just worlds above most other defenders with their on-field ability if they happen to get the ball in their hands. Generally, these are guys with pretty good hands and the ability to make cuts and the patience to allow blocks to develop - i.e. guys with punt return skills. Just from who I watch, I am going to suggest from the past Deion Sanders and Rod Woodson as being exemplars of this phenomena, while a more current defensive back with serious ball running skills would be Lito Sheppard who has 12 interceptions (with 3 returned for touchdowns) since 2003 and a total of 409 yards of interception returns, which translates into an unreal 34 yards per return average - I don't think anyone else in the league is even close.

With most defenders, the return yards they seem to get are simply the yards from where they get their hands on the ball to the point where they collide with the nearest offensive player while running headlong in the direction they originally started. Most defenders are defenders because they have bad hands (which is part of why we see so many dropped interception opportunities), and poor field awareness in attempting offensive/special teams play, which is what an interception return essentially is.

For most defenders, if they happen to run a ball back to the other end zone, its because they intercepted an out pattern or slant and were simply able to run unmolested the opposite direction because no offensive player was in any position to touch them. Had someone been in front of them, they would have been tackled.

Kevin and others commentin on the Eagles-Giants game:

The fumble recovery in the endzone was the critical play of the game (along with the Trent Cole kick in the nuts). The Eagles were attempting to run out the clock in the 4th quarter, not to score more points, since they had a comfortable 3 score lead on the Giants - the important thing was to end the game as quickly as possible, not to run up the score. Without the oddity of that fumble recovery, the Giants would probably have had to use up several more minutes of clock to score (and might have only gotten a field goal), and thus probably would not have gotten the ball back at the end of the game to move downfield for the game tying field goal. I suspect the game would have ended in a 24-17 Giants loss but for the fumble recovery, with Eli being praised for a heroic 4th quarter effort that just came too little, too late. The randomness of the event was shown by the exact same play occuring the following week in Seattle, with Buress again fumbling the ball on a 20 yard reception across midfield, except that in Seattle, instead of bouncing wildly 20-30 yards downfield into the endzone, the ball dropped dead at Buress' feet and was recovered by Seattle.

Did the Eagles have opportunities to attempt to score in the 4th quarter? Sure, but they were playing conservative to end the game. Who expects to see a 17 point 4th quarter lead dissappear? That was obviously a mistake, especially with the attempt to run on the Giants, which is quite futile, and Andy Reid admitted as much later. He should have kept the foot on the accelerator and kept on passing to run up the Eagles score to 27 or 31 points. That plan probably would have lead to at least one more Eagles score (and also a win).

161
by andrew apold (not verified) :: Wed, 11/01/2006 - 2:21am

Is there any analysis of DVOA affect of non-standard uniforms (anything other than the team's base home/away jerseys)? Do nfl game records show the uniform? I think the gamebook did, but is it in a searchable format?

162
by Travis (not verified) :: Wed, 11/01/2006 - 2:36am

Re: 160

Sheppard's return average is better than anyone since 2003 with a similar number of interceptions, though Travis Fisher (5 int, 47 yards per return) is better, as are the Derrick Dockery-types (1 int, 96 yards).

Chris McCallister, Ronde Barber, and Phillip Buchanon each have 3 int return-TDs in that time, while Darren Sharper has 4.

Just to clarify:

Eagles-Giants gamebook

1. The Eagles had a 17-point 4th quarter lead, but never had the ball in the 4th quarter with a 17-point lead - the Burress fumble-TD came 3 plays in, at the 13:53 mark.

2. The Giants would never have punted on 4th and 7 at the 6:16 mark had the Eagles recovered the Burress fumble. I was surprised they did so in the first place.

3. A huge unmentioned key was the Westbrook fumble on the next drive, which gave the Giants a short field down 10 with 4+ minutes to go.

163
by Lou (not verified) :: Wed, 11/01/2006 - 3:05am

I also really like orange jerseys. I think the Bears look very sharp in them. I also believe they should never wear them again. The Navy blue just looks too good. I second the motion to get Denver to change their jersey color back (and stop stealing our color scheme.)

164
by Ian (not verified) :: Wed, 11/01/2006 - 3:18am

I haven't read every single post, but surely a major factor for giving up TDs off interceptions (from passes originating in the redzone) is the offensive package. If a team like Pittsburgh has a jumbo package with 1WR and two TEs and runs a play action pass that gets intercepted, surely that has to be more of a risk than a team that has 5 WR set on the field? Once the CB/S/LB steps into the ball, it is unlikely that anyone other than a fast RB or WR will be able to catch him if he gets by the first offensive player. In this case, I don't see how this can be quantified? Thoughts?

165
by Torn (not verified) :: Wed, 11/01/2006 - 3:19am

That top 10 on why DVOA hates Denver is hilarious...impressive too considering it seems you guys did your homework...

166
by Andrew (not verified) :: Wed, 11/01/2006 - 3:21am

Travis:

Yes, the Westbrook fumble was key, but the occasional fumble by a running back being recovered by the defense is not particularly flukey. Recovering it in scoring position almost guarantees some points.

Kicks to the nuts on the 3rd to last play of the game and balls fumbled 20 yards forward into the endzone are flukey.

167
by NewsToTom (not verified) :: Wed, 11/01/2006 - 3:33am

Rod Woodson over the course of his career had 71 interceptions, of which he returned 12 for touchdowns, a rate of 16.9%. Deion Sanders over the course of his career had 53 interception, of which he returned 9 for touchdowns, a rate of 17.0% Referring to numbers earlier in the thread (#24), these are both better than the rate put up by Ty Law, but worse than that by Chris McAlister and Nate Clements, let alone Mike Brown. So, they're good, but not really that good. And if you remove 1 piddling TD return from Deion Sanders' 188 game career, he suddenly becomes a lot closer to league average (15.0%). Pardon me if I seem skeptical of assigning these any great predictive value.

168
by manning.e (not verified) :: Wed, 11/01/2006 - 3:40am

heres what im saying about dvoa and the bears versus giants.doesnt the fact that the giants have proven themselves to be as good or better than the best teams in the league as well as having utterly dismantled the only bad team theyve played,tampa bay who arent really that bad,...outweigh the blowouts and two near loses the bears have put up against much lower quality oppents. I understand that dvoa takes this into account but its the rate of weighting opponent strenght that needs to be taken into account between teams with close dvoa when one has team has performed at a similar level but against a much higher quality opponent. Would you not expect the bears to have the same record as the giants if they had played the giants schedule. Shouldnt the two games that the bears should have lost against mediocre opponents weigh more heavily against their total dvoa then the three games the giants should have lost against a much higher quality opponent (though one could make an argument given the massive dvoa differential between giants-colts that the giants should have won that game, eagles-giants, that the eagles should have won that game and then the giants should have only lost two games.)

saying that we dont know how terrible a team is until later is false math because the only number that really matters in predicting or evaluating a team results is how good they are in the "right now" of when the two teams met, that is all that can be calculated since each time two teams play it is a single instance which can never be repeated.

its the same problem that there is with dave..dave means absolutely zero after the first play of week one because all teams have entered into a new set, and numbers from old sets do not apply. Probabilities do not actually work in progressions, when you flip a coin it is always going to be 50/50 unless of course , its a trick coin. Is dvoa claiming the nfl is a trick coin?

169
by Ryan (not verified) :: Wed, 11/01/2006 - 4:02am

sigh...please..just...LOOK at the VOA's..you'll..see...that..the..Bears...are heavilly dropped down by their weak schedule. An 18% difference in VOA and DVOA through seven games is a pretty damn large difference. If you look at all the DVOA's through the years you'll see that through 16 games, most teams are within 5% in terms of DVOA and VOA, so an 18% difference is huge. In fact, if the Bears keep up the 18% gap between VOA and DVOA it would be one of if not THE largest gap between VOA and DVOA in the DVOA era.

170
by Richard (not verified) :: Wed, 11/01/2006 - 4:16am

169: I believe you're fighting a losing battle against the Giant fan's willful ignorance.

171
by pcs (not verified) :: Wed, 11/01/2006 - 4:19am

Sam Adams has 3 career interceptions, of which he brought back 2 for TDs. From this I infer that the Bengals need to be getting Sam Adams more into pass coverage.

172
by Travis (not verified) :: Wed, 11/01/2006 - 4:34am

Re: 166

Is the what-the-hell-was-he-doing false start on Donovan McNabb two plays before Westbrook's fumble, forcing the Eagles into 1st-and-15, fluky enough? Sure, it was McNabb's own fault, but I don't think I've ever seen that before.

173
by tunesmith (not verified) :: Wed, 11/01/2006 - 8:25am

Beatpaths power rankings are up now, too (click my name to see) - I have the top five as IND, DEN, CHI, NE, MIN.

If rock/scissors/paper holds true next week, look for Denver to take a tumble!

174
by manning.e (not verified) :: Wed, 11/01/2006 - 8:59am

my ignorance isnt willful its accidental and mostly caused by the socratic nature of life, but i digress. The problem with voa versus dvoa differential is that as i said before,the weighting must be way off if dvoa voa differential ends up being so small when they have played the absolute easiest schedule in the entire league whereas a team playing the fifth most difficult schedule has half the voa of the other team but doesnt get an increase in terms of voa. variance can reach absolutely no statistical significance whatsoever because the teams dont play the same teams over and over again. Variance can only be relevant when conditions are replicated which is impossible. Making believe for a moment that variance could reach significant, the fact that the bears dont vary in their play proves that the variable is insignificant..of course they dont have variance-they play awful teams every week and play against those awful teams the same way- which only proves they are better than awful. if one were to consider variance a factor then combined with strength of schedule the jaguars would have to be ranked the number one team in football because are numero uno in variance, have played the third hardest schedule, posted a winning record and will in the immediate future be playing the 27th easiest schedule.

175
by Rick (not verified) :: Wed, 11/01/2006 - 9:29am

Re: 171
I think you need to be drinking more Sam Adams Boston Lager.

:)

176
by Aragorn (not verified) :: Wed, 11/01/2006 - 10:40am

On Brady and the Tampa-2 and Zone:
From the game comments...
"In the next four weeks, the Patriots play the two undefeated teams and both of them play the Tampa-2 scheme. Do you see where I am going with this?"

I think the Bears enjoy people seeing them as the Tampa 2. Sure they play it but they have the personnel to play man-to-man. Patriots vs Colts D/Minnesota D will be very different from Patriots vs Bears D. Not saying Brady won't put up some points but I expect the Bears will play a healthy dose of man-to-man vs Patriots and if not their "Tampa-2" matchups are better than the other teams mentioned.

177
by azibuck (not verified) :: Wed, 11/01/2006 - 10:41am

"At this point, Jason Witten has taken over for Jeb Putzier as "fantasy tight end who teases you with yardage but never seems to get a look in the end zone."
Well timed. Why don't you just write "I hate the Cowboys" for your comment every week.

178
by Andrew (not verified) :: Wed, 11/01/2006 - 11:25am

NewsToTom:

Deion Sanders had 1331 return yards to go with those 53 interceptions, a 25 yard per interception average - double the league average.

Rod Woodson had 71 interceptions and 1483 return yards (NFL record), a 21 yard per interception average or about 50% higher than league average. His 12 TD returns are also an NFL record.

Both these guys are obvious first ballot hall of famers.

Mike Brown has 14 interceptions in 5 1/2 years of play, and just under 18 yards per return. Nate Clements has 20 interceptions in 5 1/2 years of play and is below league average in return yards at just over 13. Chris McAllister has 19 interceptions in 7 1/2 years, but a much more impressive 21 yards per return. When he gets the ball he clearly knows what to do with it, but he doesn't get the ball often enough to put up great career numbers (just a little over twice per year on average).

None of these guys are obvious Hall of Famers, so I wouldn't mention them in the same breath with Woodson and Sanders.

Actually, the most impressive active defensive back ballhawk is Darren Sharper - in 9 1/2 season, 47 interceptions, 953 yards, 20 yards per return, 7 TD's. He's the only active guy with a clear shot at career trifecta statistics (total interceptions, return yards, and return TD's) up there with Rod Woodson and Deion Sanders. A young guy like Lito Sheppard or DeAngelo Hall or Nathan Vasher might also do it, but all of them have a long way to go careerwise.

179
by Pat (not verified) :: Wed, 11/01/2006 - 11:55am

By the way, Pat, how the heck do you know so much about it? How come your writing isn’t green? :)

I've been here since early 2003. I've read most of the articles. That's pretty much enough.

Plus, it should note that almost everything people talk about when trying to suggest things has been done before. It usually falls into one of several categories - either "X isn't luck", "team Y is clearly not that good, they've only won Z games" when team Y has a low estimated wins and a high DVOA, "blowouts shouldn't be included", "run to win", etc.

Plus the search feature is nice when you remember the titles of the original articles. :)

Thad:

They are certainly doing some things right, not all.

They're doing something not right. Obviously. They're losing. But "not running more" isn't something they're doing wrong. It's a symptom, not a cause.

Really, if I had to guess, it's penalties. Philly's among the league leaders in penalties/game currently, and the company they keep is not good at all. This could be luck as well - as in, the bad luck to have certain officiating crews for games, though there's blame to be placed on the team for not handling that well. And also, you hate to say "this team's not good" when they're losing due to penalties, because even though it's their fault, it's not a good measure of how they'll do in the future, because in most cases, it's either outside the game itself (like personal fouls) or it's essentially random (like holding).

180
by MCS (not verified) :: Wed, 11/01/2006 - 12:13pm

OK Pat. How do you do it? How do you find all the different articles from years past?

Do you have them book-marked somewhere? Do you use the search engine? I haven't found it very helpful.

Every time I want to find an old article, I have trouble finding it.

181
by Mike B. (not verified) :: Wed, 11/01/2006 - 12:22pm

Buffalo is clearly ranked too high because they are clearly given a boost by not sucking horribly in their bye week. Drinking heavily and trying to block out the horror is way better than this. Brady Quinn for us? No chance, that would make sense...

182
by Pat (not verified) :: Wed, 11/01/2006 - 12:35pm

MCS: I know, I kindof agree. I think one of the best things Aaron could have one of the interns do is try to organize past articles by author, and then by type of article.

In my case, I just remember what the article was called, roughly, or one of the catchphrases in it.

Bookmarking things would probably make sense, but I've never really used bookmarks.

183
by Tom Kelso (not verified) :: Wed, 11/01/2006 - 1:09pm

The only reason I can think of right off the top of my head for the Chargers being unusually bad in the red zone is poor tackling.

Closer watchers of the Bolts can verify this, but the best evidence of this is Todd Heap's winning TD, where he ran about 7 yards into the end zone after catching the ball, and taking a hit from Merriman. If Merriman had gone for a tackle, rather than trying to blow him up or jar the ball loose, no way does Heap make that score. And who tries to blow up a tight end anyway? Does Marty call those plays, too?

184
by MCS (not verified) :: Wed, 11/01/2006 - 2:25pm

Ahhh. . . memory. I used to have one.

I think.

185
by Pacifist Viking (not verified) :: Wed, 11/01/2006 - 2:35pm

Sorry, Pat: no more questions after skimming.

186
by cttb (not verified) :: Wed, 11/01/2006 - 2:51pm

159: To the degree that people are criticizing DVOA more specifically and intently, it's a good sign. This site is seeing less commentary of the zlionsfan template variety, where there is no comprehension of DVOA at all, and more directed commentary that shows at least minimal understanding of DVOA's foundation. When people are asking, "How come DVOA doesn't cover aspect X? Doesn't X have a statistical impact?", they're actually comprehending some of what DVOA brings to the table and are asking a more advanced question. That's education in action.

187
by zip (not verified) :: Wed, 11/01/2006 - 3:44pm

178:

Chris McAllister has 19 interceptions in 7 1/2 years, but a much more impressive 21 yards per return. When he gets the ball he clearly knows what to do with it, but he doesn’t get the ball often enough to put up great career numbers (just a little over twice per year on average).

I disagree that a yardage return average, especially with just 19 returns, means anything. Lets look at McAlisters picks for TD (http://www.nfl.com/players/playerpage/133350):

98 yards, 83 yards, 60 yards, 51 yards,
Career pick-six return yardage, 292 yards.

Career INT return yardage, 397 yards.

So 73% of his INT yardage comes from 4 picks. On the other 15 picks he's made that weren't returned for TDs, he's averaging a scant 6.8 yards per return.

Sorry, but this is a case of small sample size -- all of his pick-sixes have been at least 50 yards or more. It has nothing to do with him "clearly knowing what to do with the ball."

188
by Bobman (not verified) :: Wed, 11/01/2006 - 3:53pm

PCS #171--excellent, but jerseys numbered in the 20's don't fit him. What to do... stitch two together so he's number 2728?

Starshatterer, Rodney Freakin' Harrison won't return my calls! How can I pay him respect, and recalibrate his respset-o-meter, if he won't let me? Oooh, I got it: Have Manning send him a bottle of champagne with a note congratulating him on the win--before the game. Now THAT shows respect, so long as it's the real stuff, you know, from France or someplace, and not that Idaho bubbly I drink. Because Rodney would take second-tier sparkling wine as, well, you know....

I expect a game much like the season-opener a few years back, when NE uncharacteristically (for then) opened with an empty backfield, and Indy opened with about 8 straight runs by James, and both scored on their opening drive. With no James to fumble at the goal line and Manning protecting the ball better, maybe I get a win, maybe not. I'm not counting on one. A tie would be pretty freakin' funny, really.

189
by B (not verified) :: Wed, 11/01/2006 - 4:13pm

188: I kinda have a feeling this game will come down to a Vinatieri field-goal

190
by MJK (not verified) :: Wed, 11/01/2006 - 4:15pm

Why would NE open empty backfield given (1) They just did that, and Belichick almost never repeats game plans from week to week, and, more importantly (2) Indy's run defense this year?

I expect NE to come out in their 2 TE I set. They may very well run passes out of this set, precisely because it LOOKS like a running set but they're actually pretty good at passing out of it, and because everyone expects them to start pounding Maroney and Dillon into Indy's run D, but I would be shocked if they came out empty.

191
by MJK (not verified) :: Wed, 11/01/2006 - 4:20pm

Scratch that--I forgot Dan Graham was injured (and maybe Stephen Neal, too). I'm sure missing their two best run blockers was part of the reason why they came out throwing against Minn, in addition to Minn's run D.

192
by jimnabby (not verified) :: Wed, 11/01/2006 - 4:21pm

Echoing 177: Very strange team comment on the Cowboys (especially with far more obvious things to discuss).

With Bledsoe in there, Witten didn't even tease us with yardage. He's got over a third of his catches, almost half his yards and his only TD in the last 3 halves with Romo. He went from non-entity with Bledsoe to prime target with Romo. That's the story on Witten.

193
by B (not verified) :: Wed, 11/01/2006 - 4:34pm

190: Ahhh, but if nobody expects Bellicheck to go out with five-wides because he never repeats himself, then the Colts won't be prepared for it, so it'll be a surprise when he does it.

194
by MJK (not verified) :: Wed, 11/01/2006 - 4:38pm

Ahh, but knowing that everyone knows that he knows that everyone knows...

195
by mactbone (not verified) :: Wed, 11/01/2006 - 4:43pm

INCONCIEVABLE!

196
by dryheat (not verified) :: Wed, 11/01/2006 - 5:35pm

So he should poison both?

197
by Marko (not verified) :: Wed, 11/01/2006 - 6:48pm

194 - That reminds me of a line from The Warriors: "I know it. But now he knows I know it."

198
by Starshatterer (not verified) :: Wed, 11/01/2006 - 6:50pm

B (#189 )--

No. It's coming down to a Gostkowski fielg-goal.

199
by MJK (not verified) :: Wed, 11/01/2006 - 7:04pm

Sorry this is off topic, but...

The Pats just released their injury report. SIXTEEN players (plus Brady, with a "probable" as always) are listed as "questionable" with no further information other than a general body part. Anyone think Belichick is trying to play with Dungy's head? Anyone think the league might get on the Pats case again? This is a new level of not giving useful injury information.

200
by Bobman channelling Inigo Montoya (not verified) :: Wed, 11/01/2006 - 7:07pm

195

I don't think that word means what you think it means.

201
by Bobman channelling Inigo Montoya (not verified) :: Wed, 11/01/2006 - 7:12pm

Re 156,

Check out Indy's 1st half and 2nd half offensive performances as well. Night and day. (Actually, it's that way on D too, which makes me think it's the old "halftime adjustment" chestnut.) I seem to recall the concept of halftime adjustments getting slapped around (was it here) a year or two ago, but damn if it doesn't look that way in Indy. Pat, you seem to have a monster memory for these articles... am I recalling correctly?

Which of course means that this week Belichick will to exactly what they think he won't do because they're expecting him to do what they expect him to NOT do in the first half (you get it so far?) and then in the 2nd half, he'll do exactly the same thing, when they are expecting some sort of adjustment. Devilishly clever! (brain hurts, must stop.)

And I bet that all this time he's been building up a tolerance to iocane powder.

202
by Bobman channelling Inigo Montoya (not verified) :: Wed, 11/01/2006 - 7:19pm

MJK, in short, no, the Colts have been doing that all season as well with the injury list. Granted, their D looks like a MASH unit, but last week they had about 15 guys listed compared to Denver's 3. Does anybody really think they were 5 times as injured?

Looks like they've learned from the master. This week, each team will play their only healthy QB, with Steve Grogan facing off against Bert Jones.

Speaking of Bert Jones, I have never seen anyone throw a football as a "fastball" quite like him until Brady the other night. Looked like he was trying to hurt his receivers a few times. Is that typical? Was he trying to thread the needle more than usual? I always think of him as a guy with good touch and was a bit surprised.... Now Jones used to leave gaping exit wounds in his receivers back in the days of stickum, and often enough, the recievers could barely handle them. That was not a problem for the Pats Monday night, but just wondering.

203
by Bobman (not verified) :: Wed, 11/01/2006 - 7:20pm

Crap, forgot to fix my name. Maybe I should just change it like William James of the Eagles. (William James... wasn't he a philosopher? Brother of Henry James?)

204
by Bobman (not verified) :: Wed, 11/01/2006 - 7:24pm

This quadruple post thing is awful, but I just saw Marko's comment at 197:

Reminiscent of Prince Geoffrey in "The Lion in Winter" about who knows what in a palace intrigue. "We know. He knows we know. We know, he knows we know. We're a very knowledgeable family."

Kings, queens, legends, swords, knights, back-stabbing, and treachery around every corner: I highly recommend it.

205
by CaffeineMan (not verified) :: Wed, 11/01/2006 - 7:41pm

Re: Pats injury report

Nah, I'd be surprised if Belichick figured Dungy would be fooled by that. To me it's clearly a dig at Polian, rather than a serious tactic.

206
by MJK (not verified) :: Wed, 11/01/2006 - 7:53pm

On Brady throwing hard:

Despite his much hyped "poise", I would imagine that some of it probably had to do with Bellichick telling him before the game "We're not even going to try to run it. We're going to throw on every offensive down, even after they figure that out and start blitzing you. Oh, and we'll be going 5 wide most of the time, so you don't get a RB to help protect you back there. Have fun!" If I was Brady, I might be eager to get the ball out just a little bit faster in that case, and maybe put a little more heat on the balls.

Some of it also might be still getting used to the new recievers. There's less chance for confusion about a reciever's speed and route to mess up a pass if the ball gets there faster. To my eye, the extra hard throws seemed to be mostly to Gabriel and Jackson and Caldwell--Faulk and Watson were getting balls with better touch. Of course, that may be revisionist memory now that you have brought it up.

207
by Peter (not verified) :: Wed, 11/01/2006 - 8:46pm

Where can I find a team's rushing offense DVOA, or anything specific like that?

Or if this is easier, what's Jacksonville's rushing offense DVOA?

208
by DavidH (not verified) :: Wed, 11/01/2006 - 9:08pm

top of the page, Just The Stats --> Team Offense

209
by NewsToTom (not verified) :: Wed, 11/01/2006 - 9:09pm

Re #207
Check out "Team Offense" under "Just the Stats." JAX's rush O-DVOA is currently -1.0%, 11th in the NFL.

210
by Pat (not verified) :: Wed, 11/01/2006 - 9:20pm

I seem to recall the concept of halftime adjustments getting slapped around (was it here) a year or two ago, but damn if it doesn’t look that way in Indy. Pat, you seem to have a monster memory for these articles… am I recalling correctly?

Not that I saw. There was an article by Phil Simms a while back, though, on NFL.com. Is that the one you're thinking of?

211
by Andrew (not verified) :: Thu, 11/02/2006 - 12:38am

zip #187:

I disagree that a yardage return average, especially with just 19 returns, means anything. ... Career pick-six return yardage, 292 yards. Career INT return yardage, 397 yards. ... the other 15 picks he’s made that weren’t returned for TDs, he’s averaging a scant 6.8 yards per return. ... It has nothing to do with him “clearly knowing what to do with the ball.�

You have to remember that some interceptions go for zero yards because the defensive player makes the play by leaping up and is immediately tackled or downed by contact or runs out of bounds by momentum. The lack of return yardage there is not the fault of the defender in terms of being a poor field runner but is purely situational.

As far as McAlister goes, he had since 2001, for which the NFL provides by game data:

2001 - 1 INT, 0 yards
2002 - 1 INT, 0 yards
2003 - 3 INT, 83 yards (TD), 6 yards, 4 yards
2004 - 1 INT, 51 yards (TD)
2005 - 1 INT, 0 yards
2006 - 3 INT, 60 yards (TD), 0 yards, 0 yards

So of 5 returnable interceptions, he ran 3 into the endzone, and averaged just shy of 41 yards per return. Yes its a small sample, but it definitely shows talent in terms of field awareness and elusiveness.

McAlister also holds the NFL record for the longest return of a missed field goal (107 yards). A guy with no field awareness would not have these returns.

The same arguments could be made against Lito Sheppard's yardage, considering that two of his returns were over 100 yards. He has:

2003 - 1 INT, 34 yards
2004 - 5 INT, 101 yards (TD), 64 yards (TD), 7 yards (voluntarily went down to force the end of game with plenty of room to run), 0 yards, 0 yards
2005 - 3 INT, 34 yards, 33 yards, 5 yards
2006 - 3 INT, 102 yards, 29 yards, 0 yards

With 9 returnable interceptions, Sheppard averaged 45 yards per return, took 3 to the endzone, voluntarily gave up significant additional yards on one return to end a game, and holds the NFL record of being the only player to ever return 2 interceptions in his career over 100 yards (he accounts for 2 of the 6 interceptions returned over 100 yards since 2000!).

Yes it is a small sample, but all you have to do is watch these players when they get a return to see that they have more than simple straight ahead speed until they run into an offensive player and are tackled.

212
by Catfish (not verified) :: Thu, 11/02/2006 - 12:47am

"McAlister also holds the NFL record for the longest return of a missed field goal (107 yards)."

Wasn't Vasher's return in last years tornado bowl against the 49ers 108 yards?

213
by dbt (not verified) :: Thu, 11/02/2006 - 1:44am

#212: Yes it was.

On a completely unrelated note, click my name if you have firefox 1.5+ for my contribution to the "graph DVOA" competition.

214
by D (not verified) :: Thu, 11/02/2006 - 2:48am

#213
Nice job. Being able to pick different stats is nice, and being able to sort them makes it easy to the effect opponent adjustments can have on a teams relative ranking. It looks like the Ravens were mislabled as the Colts, but other than that, good work.

215
by dbt (not verified) :: Thu, 11/02/2006 - 2:56am

#214: thanks -- fixed. Just typed in Baltimore, and Colts came out without thinking. I actually caught myself typing Houston Oilers...

216
by Moridin (not verified) :: Thu, 11/02/2006 - 3:22am

#213
That is a nice little setup there.Interactive graphs are always a a good feature.

217
by Bobman (not verified) :: Thu, 11/02/2006 - 5:15am

210 Pat, I don't think that was it. But it's midnight, I am trying to get actual work done, and I have no time to google. Maybe I'll shoot off a question for the mailbag (and some offseason study)--along the lines of "Which teams have the most DVOA improvements after halftime and is this an indicator of superior coaching? Is it consistent all season? Year after year?"

218
by Tom Kelso (not verified) :: Thu, 11/02/2006 - 2:21pm

#215:

Nothing wrong with that, dbt; it just means you have an innate sense of things as they should be.

219
by Fnor (not verified) :: Thu, 11/02/2006 - 2:22pm

dbt: You totally broke it. It looks like you've got a parent object with a bad type definition... I can try to take a closer look later, when I don't have to pay attention to important people.

220
by DavidH (not verified) :: Thu, 11/02/2006 - 3:27pm

Awesome job on the sortable charts, dbt.

Looking at Offensive DVOA sorted by Estimated Wins, it seems like good offensive teams are less variable than normal.

Looking at Defensive DVOA sorted by Estimated Wins, it seems like teams at the extremes (very good or very bad defenses) are more variable than normal.

Don't have the energy right this second to look at past years to see if this holds up.

221
by Erik Smith (not verified) :: Thu, 11/02/2006 - 4:13pm

I need to find a way to get to these threads before Pat. That way, I won’t have to start posts with, “Pat’s right…�

I propose that instead of posting "first" people post "Pat's right." :-D

Thanks, Pat, for your excellent posts.

222
by Erik Smith (not verified) :: Thu, 11/02/2006 - 4:21pm

I suspect that, at all levels of the game, teams and players that practice the “scoop and score� fumble recovery technique recover a lower percentage of fumbles but return a higher percentage for TDs than teams that practice the “fall on the ball� fumble recovery technique.

What technique was Michael Lewis using on the Plaxico fumble in week 1?

OK, I know I should get over week 1. I'm glad the Eagles have a bye. I need it.

223
by Pat (not verified) :: Thu, 11/02/2006 - 4:25pm

Looking at Offensive DVOA sorted by Estimated Wins, it seems like good offensive teams are less variable than normal.

Nope - teams with a good offensive DVOA tend to win more than teams with a good defensive DVOA. It's in the estimated wins article. Offense is more important to winning than defense is, in general, probably because good offenses can build leads larger than good defenses, and are therefore less susceptible to fluke losses. (Unless you're Philadelphia.)

224
by Erik Smith (not verified) :: Thu, 11/02/2006 - 4:31pm

58:Richard

I enjoyed the graph. Are you able to plot standard deviation on the graph as well?

Thanks!

225
by Wanker79 (not verified) :: Thu, 11/02/2006 - 4:39pm

Re: Variation

I took TOTAL DVOA (WDVOA only flip-flops NO & JAX) and divided it by VARIANCE. I don't know if the actual numbers mean anything (probably not, so Rick will be happy), but intuitively the ordinal rankings seem to have some meaning. I guess it would be best described as "Forthcoming Excellent Archetypical Result". Here are the rankings according to "FEAR":

PHI
SD
NYG
NE
BAL
IND
CHI
DAL
DEN
CIN
PIT
STL
NO
JAC
KC
ATL
MIN
CAR
GB
WAS
BUF
NYJ
SF
ARI
SEA
TEN
HOU
OAK
CLE
TB
DET
MIA

That Arizona game is really screwing Chicago's number, but other than that, I kinda like it. Especially for the bottom-dwellers. Even Aaron commented on how consistently bad Miami has been. And I don't know about you, but I'm much more afraid of a team that almost beat Chicago or Indy or Jacksonville, than I am of a team who's most impressive game so far may have been only losing to NE by 10 or a team who beat a Seattle team that both DVOA and FEAR say is in the "really bad but not quite repugnant" tier.

226
by Erik Smith (not verified) :: Thu, 11/02/2006 - 5:10pm

91, Thad

Historically most good teams, the really good ones, pass the ball effectively and run a lot, the Eagles pass the ball very effectively.

I agree. This is what continues to baffle me about the Eagles. According to DVOA, the Eagles have the #1 rushing offense.

What this says to me is that while the Eagles are not running the ball often, they are running the ball *successfully.*

I've long felt that Andy Reid uses the run as a counterpunch, mostly running on typical passing downs.

In the past this drove me crazy because I didn't feel the strategy was successful. This year, I can say that I am amazed at how successful his run play calling strategy has been.

It's staggering to me how often the Eagles series (forgive me for not having done the full analysis on this) have gone:

1st and 10: Incomplete

2nd and 10: Run up the middle for 6 yards:

3rd and 4: Completion for 1st down

Plays 2 and 3 are successful (good for DVOA) and play 1 is unsuccessful. In general, I believe the Eagles have maintained a high percentage of successful plays.

I'll give them credit. They're running the ball successfully. But I can't hang a #1 running team label on them by any stretch. The run is a well-used weapon, but it isn't a strength (hence its low mix in the play-calling).

That said (as a fan) I'll be happy if the Eagles make the playoffs as a wild card.

227
by DavidH (not verified) :: Thu, 11/02/2006 - 5:21pm

223:
I don't know how "Estimated Wins" got in my comment. Totally meant to say that I sorted them by Variance. But I plopped all the teams from 1998-2005 into Excel and looked at some correlation coefficients.

(measure1 / measure2 ..... corr. coef. ..... non-directional significance)
offDVOA / totVAR ..... -0.06 ..... 0.373
abs(defDVOA) / totVAR ..... 0.12 ..... 0.049

So, the offensive pattern I saw was nothing. But, strangely, the defensive pattern is still there. Not strong enough to make much difference, but still, it seems weird. Especially since...

abs(defDVOA) / defVAR ..... -0.09 ..... 0..177

the correlation between "extreme" defenses and defensive variance isn't as strong. Which would seem to indicate that having a very good or very bad defense actually makes your offense more variable. But no:

abs(defDVOA) / offVAR ..... -0.01 ..... 0.872

At any rate, none of this correlates enough for me to care at all. I just am a dork who likes spreadsheets.

PS I don't know how to calculate significance, so I used the applet in the linked web page.

228
by Wanker79 (not verified) :: Thu, 11/02/2006 - 6:26pm

What does it say (if anything) that when you take ALL adjustments out of the equation (Off. VOA - Def VOA + ST VOA W/HIDDEN) Chicago just looks stupid good and every other team is practically a straight line (R^2=0.9921)?

229
by B (not verified) :: Thu, 11/02/2006 - 7:05pm

228: It means that Chicago has had the best DVOA and the easiest schedule so far this year.

230
by chris clark (not verified) :: Thu, 11/02/2006 - 7:55pm

re 67:

ok, what do
Chicago
Dallas
Kansas City
New England
New York Giants
San Diego
have in common? I've been wondering for a couple of days now and haven't come up with anything that rings true to me.

231
by DavidH (not verified) :: Thu, 11/02/2006 - 11:28pm

We're the only teams that have a positive offensive DVOA, negative defensive DVOA, and positive special teams DVOA.

232
by Chicago, Dallas, Kansas City, New England, New York Giants, (not verified) :: Thu, 11/02/2006 - 11:31pm

OK, pretend I posted that last comment under this name.

233
by Pat (not verified) :: Fri, 11/03/2006 - 1:53am

Damn Philly's uncharacteristically mediocre special teams!

Then again, a lot of that has to do with the fact that Andy isn't taking the *!#^(!ing field goal this year. That and the punt coverage (and punting in general) has been uncharacteristically poor.

234
by brian (not verified) :: Fri, 11/03/2006 - 2:48pm

how can you have two teams that the colts beat ahead of them. i bought the pro football prospectus 2006 , but i look at this and the system is clearly flawed. Philly is 2nd ahead of philly and jax. you got to be kidding me. So how far will indy move up if they beat ne. To 10

235
by Wanker79 (not verified) :: Fri, 11/03/2006 - 4:26pm

I find it hard to believe than anyone who spent money on a PFP would have so little comprehension of what this site is all about. I'm frankly dumbfounded.

236
by chris clark (not verified) :: Fri, 11/03/2006 - 8:05pm

re 234/236:

I suspect this problem (and a common problem) is reconciling actual w/l records v. DVOA. I suspect that there would be a lot less grief if the "power" rating system (ordinal rankings) were based on an average of expected wins and actual wins (i.e.(expected wins + actual wins + 1/2 actual ties)/2) with DVOA used to sort teams with identical scores on said ("power" ranking) metric. Doing so, would pull down the rankings of "unlucky" teams such as PHI and PIT whose actual record was under-performing the DVOA expectations and artificially inflate teams like IND and DEN whose record was better than the DVOA would predict.

I'm not suggesting the site change to such a metric. Any metric is going to rate some team in such a way that someone is going to disagree with.

However, I do think that it would lessen some of the "how can you rate team X over team Y who they beat" arguments, since teams with better records would float higher in the standings, even if that float were not reflective of their ability to sustain that record over the course of the season. Of course, one could even take that farther and rank the teams by first sorting on record and then using DVOA only to sort teams with the same records. But, while either of the above methods would lessen the "team X beat team Y" arguments, they wouldn't be reflective of what DVOA is trying to capture. So, I'm happy that Aaron hasn't caved in and adopted some w/l record based comparison.

If you want a w/l based comparison, look at beatwins.

Finally, how far IND moves up by beating NE depends on how they do it. After all, DEN moved up after losing to IND (DVOA 15th->12th), and did so because their stats improved relative to other teams.

237
by B (not verified) :: Fri, 11/03/2006 - 8:37pm

236: Why would bringing DVOA closer to conventional wisdom help things? If I need a ranking system based on wins and losses, I can check every other ranking system available.

238
by chris clark (not verified) :: Fri, 11/03/2006 - 9:00pm

re 237:

I didn't say it would help anything (other than cutting complaints) and said it would defeat the purpose of the site. Perhaps, in my long-winded-ness you missed that.

239
by B (not verified) :: Fri, 11/03/2006 - 9:11pm

Oops. Next time I'll read the whole post before commenting, then.

240
by Pat (not verified) :: Sat, 11/04/2006 - 1:24am

I suspect that there would be a lot less grief if the “power� rating system (ordinal rankings) were based on an average of expected wins and actual wins

Actually, if you want a real way to make a ranking more akin to conventional power ratings, what you'd do is something like VOA*(games played/16)+DVOA*(games to be played/16). That's ranking teams by a combination of how they have played (VOA) and how they will play (DVOA).

Doesn't completely fix Pittsburgh or Philly, but I doubt anything would.