Writers of Pro Football Prospectus 2008

Most Recent FO Features


» 2017 Adjusted Games Lost

Two NFC teams were hit hardest by injuries last year. One already set the AGL record in 2016, while the other has a coach with the worst AGL since 2002. Also: the Rams' incredible bill of health in L.A., and Tampa Bay's questionable injury reporting.

02 Nov 2010

Week 8 DVOA Ratings

by Aaron Schatz

Is it time to take the Kansas City Chiefs seriously? It sounds crazy, since they just barely beat the only winless team in the league, but the Chiefs actually move up after that win and stand first in the Football Outsiders DVOA ratings. When we predicted before the season that Kansas City would win the AFC West, that prediction was based on the idea of an average team with an easy schedule taking advantage of the declining Chargers. That prediction seemed pretty spot on after a couple weeks, but as more time goes on, it seems pretty clear that the Chiefs are not an average team.

Right now, shocking as it is to say, the Kansas City Chiefs may be the most balanced team in the league. The Chiefs rank sixth on offense, fifth on defense, and 12th on special teams. With all the success they are having on the ground, you probably expect their offensive DVOA to be tipped in favor of the run. That's not the case. The Chiefs rank fourth in rushing DVOA, but only one spot lower -- fifth -- in passing DVOA. Yes, they rank fifth with Matt Cassel at quarterback and with Chris Chambers and his horrid 41 percent catch rate in the starting lineup (well, until they got wise and pulled him out after Week 5). Actually, because of Chambers' struggles, Dwayne Bowe is the only Chiefs wide receiver with enough targets to be listed in our main wide receiver stats. The Chiefs are 31st in the NFL with 187 net pass attempts, but when they do pass, they've been effective.

Actually the unit behind the success of the offense is probably the offensive line. Cassel has taken only eight sacks and the Chiefs are fifth in Adjusted Sack Rate after ranking 25th a year ago. They are third in Adjusted Line Yards and lead the league with runners only getting stuffed at the line on 11 percent of carries.

The defense is as balanced as the offense; the Chiefs rank ninth against the pass and seventh against the run. A year ago, the Chiefs ranked dead last on defense in Adjusted Line Yards and Adjusted Sack Rate. This year, they are 13th in ALY and 21st in ASR. Cornerback Brandon "I've never visited Tulsa with Rufus Wainwright so don't look at me like that" Flowers has taken what Bill Simmons always calls "the leap," a big reason why the Chiefs rank fourth in DVOA against number-one receivers. Some people think outside linebacker Tamba Hali has also taken "the leap," but those people are wrong. As Bill Barnwell pointed out in FOA 2010, Hali actually took the leap last year, and this year is just a continuation of that fine performance.

Early on, we wrote off the Chiefs' win over San Diego as a bit of a fluke. Then they beat the Browns and 49ers, so wasn't their 3-0 start just the easy schedule? Turns out that schedule isn't quite as easy as we thought it would be. Kansas City's schedule ranks 23rd through eight weeks. That's still below average, but not the easiest in the league. Their two losses, to Indianapolis and Houston, were fairly close. The Houston game is the only one this season where the Chiefs have a below-average DVOA. As for those first three weeks, the Chiefs truly outplayed the Chargers, with 51.3% DVOA in Week 1 compared to -19.5% DVOA for San Diego. And Cleveland isn't a joke; the Browns actually moved up to 16th in this week's ratings and are playing reasonably well against one of the league's hardest schedules.

How did the Chiefs go up when they barely beat Buffalo? Actually, the Chiefs end up with a 28.1% DVOA for this week's game. The Chiefs demonstrated they could move the ball significantly better than the Bills, a fact that was obscured when Todd Haley made fourth-down decisions that either didn't work out (Cassel getting sacked on fourth-and-2 from the Buffalo 19) or were certifiably insane (punting on fourth-and-8 from the Buffalo 33 instead of either going for it or trying a long field goal). In addition, the Bills fumbled three times but recovered all three themselves. 

Now, I don't think that the Chiefs are truly the best team in the league, even if they have the best numbers so far. However, the Chiefs need to be taken seriously as a Super Bowl contender. The Chiefs are certainly much more real than the Raiders, the team they visit this week in a game that we are contractually required to refer to as "unexpectedly important." Kansas City isn't a shoo-in, especially since the game is in Oakland, but the Raiders are one of many teams who, unlike Kansas City, do seem to be the product of massive league-wide schedule imbalance. Seven different teams have a .500 record or better despite a DVOA below average. St. Louis, Seattle, and Tampa Bay have had the three easiest schedules in the league.

By the way, Kansas City's next five games include Oakland, Seattle, Arizona, and Denver twice. Average DVOA of those opponents: -21.6%. Maybe the schedule hasn't been as easy as you think so far, but that's okay, it's about to get easier.

Below Kansas City, this week saw a lot of movement in the DVOA rankings, although that's not necessarily because of heavy movement in the ratings themselves. Teams are so close that small changes in opponent adjustments will move teams up and down the table, and this week actually saw some surprisingly large changes in opponent adjustments. Perhaps the most interesting change is in the NFC East, with the Giants and Eagles. Even though both teams were on bye, the Giants dropped from first to sixth and the Eagles rose from sixth to second. How do two division rivals end up with opponent adjustments going strongly in opposite directions? Although the Giants and Eagles will end up playing 12 games against common opponents this year, they've actually played only two games against common opponents so far. Now look at what happened to those five opponents this week:

Previous Giants Opponents in Week 8 Previous Eagles Opponents in Week 8
CAR Lost to #28 St. Louis ATL Bye
CHI Bye GB Beat #9 Jets
DAL Crushed by #25 Jacksonville JAC Dominated #27 Dallas
HOU Cancelled out by IND SF Beat #29 Broncos
IND Cancelled out by HOU WAS Narrow loss to #19 Detroit

So Giants opponents had two bad losses, thus making the Giants' schedule look worse. Eagles opponents had two strong wins, one okay win, and one narrow loss, thus making the Eagles' schedule look better. And thus, the two teams basically switch places in the DVOA rankings. Of course, don't overlook the forest for the trees here. In reality, the top six teams in DVOA are really in one single tier, all basically equal in performance this season. In addition, DVOA can't change wins and losses on the scoreboard, and the Giants still have the advantage there. Combine that with the fact that it should be much easier to get out of the NFC playoffs than the AFC playoffs, and Big Blue is still the leading Super Bowl contender in our playoff odds at 16.0 percent.

* * * * *

This week we say goodbye to DAVE, our formula which combines preseason projection with in-season performance. The tables now list WEIGHTED DVOA instead, our formula which drops the importance of earlier games, although right now the difference between regular DVOA and weighted DVOA is minimal.

Before we get to the ratings, I want to clear up some misconceptions about DVOA that seem to have come up recently in the comment threads. On our "FO Basics" page, we note that "The total quality of an NFL team is three parts offense, three parts defense, and one part special teams." Some people are wondering: If that is true, why is total DVOA figured by using the full values of offense, defense, and special teams. Shouldn't total DVOA actually be 3/7 times offense, 3/7 times defense, and 1/7 times special teams? The answer is no. Special teams DVOA is figured differently than offense and defense, because teams on standard plays have two goals (get closer to a first down and get closer to the goal line) while on special teams you have only one goal (either get closer to the goal line or, for placekicking, put the ball through the uprights). Special teams DVOA takes a point value based on estimated field position gains and then translates it into DVOA, and the coefficient that translates it to DVOA is based on maximizing the correlation of total DVOA and winning.

Once we do that, however, we discover that in general offenses and defenses will range from about -30% to +30%, while special teams will range from about -10% to +10%. From this comes the idea that special teams are only one-third as important as offense or defense. Having the best offense or defense in the league will usually help a team about three times as much as having the best special teams.

I hope that helps clear things up.

* * * * *

These are the Football Outsiders team efficiency ratings through eight weeks of 2010, measured by our proprietary Defense-adjusted Value Over Average (DVOA) system that breaks down every single play and compares a team's performance to the league average based on situation in order to determine value over average. (Explained further here.)

OFFENSE and DEFENSE DVOA are adjusted to consider all fumbles, kept or lost, as equal value. SPECIAL TEAMS DVOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season. WEIGHTED DVOA represents an attempt to figure out how a team is playing right now, as opposed to over the season as a whole, by making recent games more important than earlier games.

Opponent adjustments are currently at 80 percent strength and will steadily grow stronger until Week 10. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.

To save people some time, please use the following format for all complaints:

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

1 KC 29.4% 4 29.2% 1 5-2 17.9% 6 -8.7% 5 2.9% 12
2 PHI 27.3% 6 27.5% 2 4-3 25.3% 3 -4.7% 9 -2.7% 26
3 NE 26.8% 3 26.7% 4 6-1 32.7% 1 11.4% 27 5.5% 5
4 TEN 26.8% 5 27.0% 3 5-3 5.9% 12 -16.4% 3 4.5% 7
5 PIT 25.8% 2 25.9% 5 5-2 3.7% 16 -19.9% 2 2.3% 13
6 NYG 25.3% 1 25.6% 6 5-2 11.8% 9 -20.6% 1 -7.1% 31
7 SD 17.0% 8 17.0% 7 3-5 20.9% 5 -12.4% 4 -16.3% 32
8 IND 15.5% 10 15.5% 8 5-2 22.9% 4 3.1% 19 -4.4% 28
9 NYJ 14.9% 7 14.4% 9 5-2 5.2% 13 -4.0% 10 5.7% 4
10 GB 14.1% 11 13.8% 10 5-3 12.7% 8 -7.0% 6 -5.6% 30
11 MIA 12.8% 13 13.5% 11 4-3 10.7% 10 -2.3% 13 -0.1% 17
12 ATL 12.1% 9 11.9% 12 5-2 15.1% 7 3.4% 21 0.5% 15
13 BAL 8.3% 12 8.2% 13 5-2 7.3% 11 2.5% 17 3.5% 10
14 NO 6.3% 14 6.4% 14 5-3 5.1% 14 -3.1% 12 -2.0% 24
15 HOU 2.6% 15 2.6% 15 4-3 27.4% 2 22.8% 30 -1.9% 23
16 CLE -3.6% 20 -3.5% 17 2-5 -3.8% 21 3.3% 20 3.5% 9
17 SF -4.1% 22 -3.4% 16 2-6 -7.5% 27 -3.7% 11 -0.3% 18
18 MIN -4.6% 18 -4.3% 19 2-5 -7.3% 26 0.7% 15 3.4% 11
19 DET -4.7% 21 -4.3% 18 2-5 -5.7% 24 3.9% 23 4.9% 6
20 CIN -7.1% 19 -7.3% 20 2-5 2.0% 17 7.3% 25 -1.8% 22
21 WAS -8.9% 23 -8.9% 21 4-4 -4.9% 23 3.6% 22 -0.4% 19
22 OAK -10.0% 26 -9.1% 22 4-4 -7.2% 25 1.2% 16 -1.6% 21
23 TB -11.0% 25 -11.2% 23 5-2 -0.4% 19 8.6% 26 -2.0% 25
24 SEA -11.8% 16 -12.4% 24 4-3 -18.1% 29 2.7% 18 9.0% 1
25 JAC -12.8% 28 -12.7% 25 4-4 -1.2% 20 18.9% 29 7.3% 3
26 CHI -14.8% 24 -14.5% 26 4-3 -29.5% 30 -6.9% 7 7.8% 2
27 DAL -15.2% 17 -15.8% 27 1-6 0.8% 18 16.3% 28 0.4% 16
28 STL -16.9% 27 -16.9% 28 4-4 -13.9% 28 -0.5% 14 -3.5% 27
29 DEN -24.0% 29 -24.2% 29 2-6 4.3% 15 23.6% 31 -4.7% 29
30 BUF -28.5% 30 -28.3% 30 0-7 -4.5% 22 28.0% 32 4.1% 8
31 CAR -37.2% 31 -37.2% 31 1-6 -42.4% 32 -6.0% 8 -0.8% 20
32 ARI -38.0% 32 -37.9% 32 3-4 -34.2% 31 5.3% 24 1.6% 14
  • NON-ADJUSTED TOTAL DVOA does not include the adjustments for opponent strength or the adjustments for weather and altitude in special teams, and only penalizes offenses for lost fumbles rather than all fumbles.
  • ESTIMATED WINS uses a statistic known as "Forest Index" that emphasizes consistency as well as DVOA in the most important specific situations: red zone defense, first quarter offense, and performance in the second half when the score is close. It then projects a number of wins adjusted to a league-average schedule and a league-average rate of recovering fumbles. Teams that have had their bye week are projected as if they had played one game per week.
  • PAST SCHEDULE lists average DVOA of opponents played this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • FUTURE SCHEDULE lists average DVOA of opponents still left to play this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • VARIANCE measures the statistical variance of the team's weekly DVOA performance. Teams are ranked from most consistent (#1, lowest variance) to least consistent (#32, highest variance).
1 KC 29.4% 5-2 26.9% 5.9 1 -2.0% 23 -10.1% 30 10.7% 17
2 PHI 27.3% 4-3 29.5% 5.9 2 3.2% 10 1.1% 15 8.1% 7
3 NE 26.8% 6-1 25.9% 5.8 4 1.8% 14 3.5% 12 8.5% 9
4 TEN 26.8% 5-3 23.8% 5.7 5 4.2% 8 7.1% 7 9.7% 13
5 PIT 25.8% 5-2 25.0% 5.4 6 7.4% 4 -4.8% 27 10.0% 15
6 NYG 25.3% 5-2 25.9% 5.8 3 -3.9% 26 0.7% 17 21.1% 24
7 SD 17.0% 3-5 14.7% 5.0 10 -0.8% 21 -2.7% 22 18.1% 23
8 IND 15.5% 5-2 15.5% 5.2 8 2.0% 12 8.8% 3 10.9% 18
9 NYJ 14.9% 5-2 25.4% 5.0 11 0.7% 18 1.0% 16 7.9% 6
10 GB 14.1% 5-3 16.4% 5.2 9 -0.8% 20 2.6% 13 9.3% 11
11 MIA 12.8% 4-3 10.7% 5.2 7 5.9% 7 1.7% 14 15.9% 20
12 ATL 12.1% 5-2 8.6% 4.9 13 0.9% 17 -10.7% 31 10.6% 16
13 BAL 8.3% 5-2 7.7% 5.0 12 0.6% 19 0.1% 19 4.7% 2
14 NO 6.3% 5-3 6.5% 4.6 14 -7.6% 29 -9.9% 29 9.4% 12
15 HOU 2.6% 4-3 -1.2% 4.0 16 7.4% 5 7.9% 4 24.5% 26
16 CLE -3.6% 2-5 -4.4% 3.8 18 9.1% 2 0.3% 18 8.4% 8
17 SF -4.1% 2-6 -3.5% 4.1 15 -1.0% 22 -12.7% 32 27.8% 29
18 MIN -4.6% 2-5 -10.7% 3.7 19 7.9% 3 -4.8% 26 2.7% 1
19 DET -4.7% 2-5 -5.2% 3.6 20 3.1% 11 -0.6% 21 9.1% 10
20 CIN -7.1% 2-5 -4.5% 3.3 24 1.2% 15 9.1% 2 5.4% 3
21 WAS -8.9% 4-4 -4.1% 3.3 23 1.0% 16 7.6% 5 6.0% 4
22 OAK -10.0% 4-4 1.9% 2.9 26 -6.0% 28 11.6% 1 36.7% 32
23 TB -11.0% 5-2 -4.9% 3.9 17 -10.1% 30 -3.1% 23 16.0% 21
24 SEA -11.8% 4-3 -6.0% 3.5 21 -13.0% 31 -3.8% 24 32.0% 31
25 JAC -12.8% 4-4 -17.6% 3.4 22 6.0% 6 6.3% 8 25.3% 27
26 CHI -14.8% 4-3 -10.7% 2.8 28 -5.5% 27 5.9% 9 23.0% 25
27 DAL -15.2% 1-6 -16.1% 2.9 27 2.0% 13 7.1% 6 26.4% 28
28 STL -16.9% 4-4 -7.6% 2.9 25 -13.1% 32 -4.3% 25 15.0% 19
29 DEN -24.0% 2-6 -23.1% 2.2 29 3.3% 9 3.8% 11 17.3% 22
30 BUF -28.5% 0-7 -29.3% 2.1 30 13.4% 1 5.1% 10 7.0% 5
31 CAR -37.2% 1-6 -36.2% 1.5 32 -3.2% 25 0.0% 20 9.7% 14
32 ARI -38.0% 3-4 -34.5% 2.1 31 -2.0% 24 -9.8% 28 28.0% 30

Posted by: Aaron Schatz on 02 Nov 2010

228 comments, Last at 08 Nov 2010, 9:25am by DeltaWhiskey


by Will Allen :: Tue, 11/02/2010 - 5:32pm

The Vikings defense has taken a very appropriate slide in the rankings in the past couple weeks. It was truly awful on Sunday.

by andrew :: Tue, 11/02/2010 - 7:31pm

Their run defense was bad, but it was almost entirely bad in the second half, as noted earlier.

This isn't the first time this has happened... in the opening loss to New Orleans the Saints had almost no rushing in the first half and ran out the game in the second half.

by dmstorm22 :: Tue, 11/02/2010 - 8:53pm

It also happened in the Jets game. The Vikings run defense seems to tire greatly as the game goes on.

by jmaron :: Tue, 11/02/2010 - 10:01pm

I watched a replay of that game and I think the defence played very well in the first half. The NE offence in that half came on two plays really:

1) Little screen pass to Branch - missed tackle by Cook (I think Welker blocked him in the back pretty clearly on that play).
2) Missed int by M. Williams - that resulted in a 30 yard gain.

Asher Allen also missed an int that went right over his head on an out that would have given the Vikings a first around the NE 20. He did the same thing in Dallas. That guy seems to have absolutely no ball skills.

They certainly weren't great in the 2nd half. Much of that was missed tackles and one huge breakdown on the long TD.

Allen was certainly the goat of that game.

You have to take into account that according to DVOA that was the best offensive team in the league they were playing.

by Will Allen :: Tue, 11/02/2010 - 10:07pm

A defensive back with no ball skills is part of a truly awful defensive performance. Simply catching very routine interceptions may have made them 4-3, or better, at this point.

by jmaron :: Tue, 11/02/2010 - 10:23pm

You could see the run defence slipping last year. It makes me realize now just how good P. Williams was as a run defender from 06-08. He was simply dominating.

I think Guion should be seeing a lot more snaps. I also think Robison should be seeing more snaps as well. Allen and Edwards are seeing 50+ snaps and Robison is seeing around 10. Neither of them is playing well enough to see that many snaps.

by andrew :: Wed, 11/03/2010 - 12:00am

Well last year they were on the field a lot less also... the offense consistently drove the ball... and by the second half they were often up by enough that teams wouldn't keep trying to run anyway.

But yeah, they need to rotate the line more. Pat Williams was indeed awesome, but he should be more of a situational player now I think...

by tuluse :: Tue, 11/02/2010 - 5:35pm

Is it time to start up the DVOA loves the Eagles mantra?

by Pass to Set Up ... :: Tue, 11/02/2010 - 6:44pm

No, let's wait and see how they look after the Indy game.

by Pat (filler) (not verified) :: Tue, 11/02/2010 - 11:09pm

They're going to get demolished by Indy, and it's not going to be pretty. Seriously, the Eagles defense is surviving on two things right now: turnovers and sacks. That's it. They happen to be pretty good at both of them.

Guess what Manning is ridiculously good at preventing? Oh $#!+, there goes the neighborhood.

by Independent George :: Tue, 11/02/2010 - 11:48pm

I think so, too, but watching the miscommunication between Manning and his receivers, and Ryan Diem alternate between getting beat and getting flagged, makes me think it's far from a sure thing.

by Dean :: Wed, 11/03/2010 - 9:08am

That, and Dmitry Patterson is getting his first career start at CB. Against Peyton Manning. Yikes.

by Kevin from Philly :: Wed, 11/03/2010 - 10:59am

Hey, everybody in town had this as a loss when the schedule came out in April, so no big surprise if it goes badly. On the other hand, the birds are always strong after the bye week, so who knows? Crazy things happen in the NFL, especially this year. Maybe one of the Williams sisters gives Payton a head slap during the weekly Oreo eating contest, and he throws 5 picks on Sunday. Maybe Justin Timberlake sings him a song and he can't play due to the migrains. Maybe his "G" is too cold and he gets a brain-freeze.

by chemical burn :: Wed, 11/03/2010 - 11:01am

The Eagles' only hope in this game is hanging 35 or 42 on a mediocre defense.

by Bobman :: Wed, 11/03/2010 - 12:52pm

I'm thinking that on the play where Ryan Diem holds, clips, AND facemasks, the whistle blows early because he also false-started--he gets a whole game's worth of penalties out of his system in one shot, but it's only five yards. Whew. It's what I'm hoping for, at least.

I thought Manning and the receivers looked out of synch as well, but maybe having more time under his belt with the same guys in practice will help. Uh, wait, wasn't Wayne one of the guys he missed a couple times? Well, maybe they'll iron things out by the time their next 10k yards and 64 TDs come along.... Once again, it's what I'm hoping for.

by bravehoptoad :: Wed, 11/03/2010 - 12:45pm

Maybe we need a zlions-fan template for Eagles complainers.

They're ranked too high because they complete lots of short passes, but their linebackers suck, their quaterback is over-rated, etc.

We could also design a counter-template.

DVOA does not especially reward small-ball offenses over long-ball offenses, not too long ago the Eagles were consistently under-rated every year, etc.

We could all just cut-and-paste and save a lot of typing.

by zlionsfan :: Thu, 11/04/2010 - 12:16am

I'll see what I can do. :)

by B :: Tue, 11/02/2010 - 5:37pm

Chefs ranked 1st Raiders rakned 22nd. DVAO Drunk too much Seirra Nevada.

On a more serious note, I know I haven't seen as many Chiefs games as DVOA has, but it's been my impression that every KC pass attempt is either a wildly inaccurate Cassel Pass or a drop by Bowe. I guess I must be missing something here. Still, their defense, rushing and special teams are all top notch.

by BigRedCadillac (not verified) :: Tue, 11/02/2010 - 6:44pm

Like almost all pass recievers Bowe does drop a few. I would say that some of those drops are due to problems with quarteback accuracy and timing. But beyond that I believe that Haley has empasized no fumble, no drops, no negative plays so much that it shines a highlight on Bowes drops and makes them seem more prevalent than other recievers. In fact Haley has begun to praise Bowe for the progress he is making in being an ALL round football player, blocking hustleing etc. And Bowe has made some spectacular catches as well as having very good yards after catch #s.

by NJBammer :: Tue, 11/02/2010 - 5:37pm

KC #1, PHI #2 - Either you guys are gonna look like geniuses or complete idiots by the end of the season. How much would $100 get you right now for a Cheifs/Eagles matchup?

by spenczar :: Tue, 11/02/2010 - 5:54pm

<-begin standard response->
Of course, there are 6 teams from 25% to 30%. I'd reckon the error in DVOA from its "real" value is probably at least 5% and probably closer to 10%, so ranking doesn't mean much in this case.
<-end standard response->

by NJBammer :: Wed, 11/03/2010 - 10:17am

If the "error" in DVOA is as much as 10%, why do they carry it out to a one-tenth decimal place?

by DeltaWhiskey :: Wed, 11/03/2010 - 10:24am

Because they're writers and not statisticians?

by NJBammer :: Wed, 11/03/2010 - 11:25am

It just reminds me a little of the old joke about the definition of estimates. "A wild a$$ guess taken to 2 decimal places."

by dbostedo :: Wed, 11/03/2010 - 12:42pm

Actually, I believe some of the writers would qualify as statisticians, and have some training to back it up.

WARNING : Thinking out loud (er... in print) ahead...contents may be garbage...

Either way, I'd think the tenths of a point are mainly just so that you can rank and compare teams. It is a formula, after all, and could be taken out to any amount of significant figures. I do wonder if anyone's ever figured out that taking it further doesn't improve its accuracy.

Plus, DVOA doesn't really have any accuracy in terms of an absolute because there's no absolute frame of reference to compare it to. Saying that it's got an error of 5% doesn't mean anything since you'll never know how good a team actually is - i.e. you can't play the same set of games thousands of times in real life to prove out how accurate DVOA is.

All you can really say is that it has a certain accuracy against actual records - but that's not the same thing as saying it's got a particular error margin. Since DVOA tries to measure actual team strength adjusted for opponent and luck, and a team's actual record is not adjusted for strength of schedule or luck or anything else, they're not directly comparable the way some people might think. Of course, changes to DVOA are to make it more accurate versus actual team records (I think) so they are definitely related.

For instance, if DVOA says that a team is the best in the league, but they have terrible luck and lose a bunch of games, they're record would be bad. So in that case, is DVOA inaccurate? Or is their actual record skewed? Since DVOA is based on actual records, the two are intertwined with no absolute frame of reference.

by DeltaWhiskey :: Wed, 11/03/2010 - 1:00pm

"Actually, I believe some of the writers would qualify as statisticians, and have some training to back it up"

Last time a I looked at the Bios, no-one on staff popped out as a statistician, nor did anyone site any particularly strong training. Moreover, Aaron repeatedly uses the line that they're not trained as statisticians.

Plain and simple, it is time that FO hire a trained statistician. There are enough readers on this site with a solid undestanding of statistics that FO must do this or let the readers see the man behind the curtains.

by dbostedo :: Wed, 11/03/2010 - 6:25pm

Ah...you know, I've never looked at the bios, and didn't realize they were there. Nothing jumps out to me either, but there's also no proof the other way I guess. I was just guessing based on comments and things remembered along the way.

by Aaron Schatz :: Thu, 11/04/2010 - 5:58pm

Hi ho. I don't think anybody at FO has ever claimed to be a trained statistician. Well, Ryan Wilson was an economist for the government, but other than him, nobody has claimed to be a trained statistician.

by spenczar :: Wed, 11/03/2010 - 5:25pm

I don't think you have a rigorous understanding of the terms you are using. The central limit theorem and the law of large numbers are pretty much enough to guarantee that samples converge towards the appropriate distribution function. Unless I'm mistaken, the conditions for both of those are satisfied in the case of football, since football is a finite sample space.... Then, all you need is a chi-squared test (or your personal favorite test of distribution goodness-of-fit) to find your error in measurement. Unless I'm missing something?

by RickD :: Fri, 11/05/2010 - 2:29pm

You're presuming that you're sampling from a fixed distribution. The most you could hope for from football is that it's a distribution that is fixed at a given point in time, but varies stochastically with time. (For example, the 1976 Steelers were better in 1976 than they are now.)
Sports statisticians reguarly mis-apply the Central Limit Theorem. You could, for example, say that Player X averaged p1 interceptions per game over his first five seasons. But he's only thrown p2 interceptions per game this season. Therefore we should expect his interception rate to increase, right?

Problem is that Player X is a much better QB now than he was as a rookie. A better model for his interception rate would be that it fluctuates over time. And thus there's no reason to expect his high rookie interception rate to be relevant.

People who only know a little bit about statistics are the worst at using poor models for physical systems. Personally, I'm not a statistician, but a mathematician (if the difference is important, and it is to statisticians), but I know enough about statistics to know that the practice of applying statistics intelligently to real-world numbers is an art that is easy to screw up.

by spenczar :: Sat, 11/06/2010 - 4:42pm

That's a really excellent point, Rick. I was mistaken.

However, even if the characteristics that correlate to winning are stochastic, they don't dramatically fluctuate year to year. That probably mostly introduces some unavoidable level of error, but isn't necessarily an insurmountable problem, right?

by DeltaWhiskey :: Mon, 11/08/2010 - 9:25am

"People who only know a little bit about statistics are the worst at using poor models for physical systems."

Point taken, and that's why I'm always excited about misapplying my little bit of knowledge.

by Groundskeeper (not verified) :: Tue, 11/02/2010 - 5:55pm

Great googly moogly!

by leviramsey (not verified) :: Tue, 11/02/2010 - 6:18pm

They're not the chefs...

by evenchunkiermonkey (not verified) :: Wed, 11/03/2010 - 3:50am

odds on KC winning the AFC are listed as 15/1 and Philly winning the AFC at 7/1. 30/1 on a KC SB win and 15/1 on a Philly SB win

by Guy#1 (not verified) :: Wed, 11/03/2010 - 9:40am

I think the odds of philly winning the AFC are alot longer than that.

Sorry couldn't help myself :D

by bravehoptoad :: Wed, 11/03/2010 - 12:39pm

KC #1, PHI #2 - Either you guys are gonna look like geniuses or complete idiots by the end of the season.

Half of this always happens. Any odd opinions in football are heaped with scorn until they're proven correct, at which point they do not get hailed as geniuses...mostly they get shrewd (or grudging) recognition.

That's still one of the more awesome kinds of recognition to get.

by Karl Cuba :: Tue, 11/02/2010 - 5:49pm

If only the niners had managed not to throw away a couple of those close games earlier in the year and had the 4-4 record FO estimates they should have, they'd look a lot better for the division. Sigh.

by bravehoptoad :: Wed, 11/03/2010 - 12:34pm

On the other hand, it was fun to be in the City when the Giants won the Series. It's so much more a part of the City, when the Giants are in the World Series than when the 49ers are in the Superbowl. I don't know why. During a World Series game, you can walk down the street in North Beach...in the Castro...in the Mission...and tell everytime a good pitch is thrown just by the sound. You can follow the tides of the game standing on Market Street with your eyes closed.

During the Superbowl...mostly silence. Everyone's hunkered down glued to the screen.

by spenczar :: Tue, 11/02/2010 - 5:52pm

SF has really been climbing in the past four weeks.
(31 -> 28 -> 32 -> 28 -> 27 -> 23 -> 22 -> 17)
Is that opponent adjustments kicking in, or is it that they've actually improved?

by Karl Cuba :: Tue, 11/02/2010 - 7:17pm

I reckon a little from column A and a little from column B. I do think that the offense has been better since Jimmy Raye got canned, which has helped to keep the defense rested too but the schedule looks harder now than it did.

by bravehoptoad :: Wed, 11/03/2010 - 12:29pm

Seriously, how's this for disrespect:

And Cleveland isn't a joke; the Browns actually moved up to 16th in this week's ratings and are playing reasonably well against one of the league's hardest schedules.

Not to mention SF at 17!!!! and 16 weighted!!!!!!



Break out the champagne.

P.S. Sorry if the tone of this message is confused. I'm grim.

by Bobman :: Wed, 11/03/2010 - 12:55pm

I think the jump is mainly attributed to Singletary dropping less trou than he had been earlier in the season.

by Mikey :: Tue, 11/02/2010 - 5:56pm

How do opponent adjustments actually work? The Bills and the Browns have faced the two hardest schedules, but their DVOA is only a tiny bit above their VOA. Does it tend to do more work blunting good performances than highlighting the better parts of bad ones?

by MJK :: Tue, 11/02/2010 - 7:44pm

I don't know for certain that this is what's up with the Bills and the Browns, but it is worth noting that opponent adjustments aren't the only difference between VOA and DVOA. I don't know everything that goes in, but, for example, fumble recovery precentage can also create a difference, and for special teams, I think weather (or at least kicking in Dome/Open/Denver) is factored in.

So it could be that the Browns might get a bump from VOA to DVOA because of their schedule, but a knock if they've recovered more than an expected number of fumbles...

Or there could be a ghost in the machine?

Also, I would image that early in the season (and we're still relatively early), opponent adjustments won't hit as strongly as they do later in the season, because the opponent adjustments are not iterative. Meaning, team A has to play team B. Team B is good, so team A gets a boost. But that boost means that Team B should look better as well, because they had to play Team A. Eventually, you converge, but I don't think DVOA runs to convergence. Later in the season, when team B is only a tiny part of team A's schedule (and vice versa), it probably doesn't matter, since just the one or two iterations DVOA does will get you close to convergence, but at this point in the season, when team A is 1/7 of team B's total schedule and vice versa (or more, if they've played twice), the opponent adjustments might be somewhat undervalued.

by Dr. Mooch :: Tue, 11/02/2010 - 10:49pm

My eyes are drawn to the VAR of the Bills and the Browns, ranked fifth and eighth. The Bills have been bad against good teams, which may help them some, but they've been bad against bad teams, too. Anecdotally, I notice that Oakland, the team with the highest VAR, has a huge 9% difference between DVOA and VOA. Seattle, next highest, has 5.8% difference. Just scanning the list, I'd guess the correlation between the absolute value of DVOA-VOA and VAR is somewhere above 0.25, maybe 0.3 or 0.4

by fluffy :: Tue, 11/02/2010 - 11:57pm

Jets actually have a 9.5% difference between DVOA and VOA

by Athelas :: Tue, 11/02/2010 - 6:01pm

So the Jets Offensive DVOA went from 14.8% to 5.2%.

Yup, that seems about right.

by Jimmy :: Tue, 11/02/2010 - 6:01pm

I know the drive stats page hasn't been updated as of writing but there does appear to a be a bit of incongruity about the Bears defense between the drive stats and their DVOA (and as the Bears didn't play this week their stats shouldn't change). What I mean is that the Bears D is ranked 2nd in yards per drive, 1st in points per drive, 1st in causing punts, 7th in turnovers and 1st in success rate. Their DVOA ranking, 7th (and at -6.9 not all that good a seventh). Is all that from adjustments based on schedule strength. Their VOA is -11% so I guess some of it must be, the two still don't seem to marry that well.

by Todd M (not verified) :: Tue, 11/02/2010 - 6:06pm

The Cowboys were so bad against the Jags that their DVOA dropped from 2.1% to -15.2% (I assume opponent adjustments are also factored in, but you get the idea).

by Raiderjoe :: Tue, 11/02/2010 - 6:11pm

Raiders at 22 and cheisf at 1 complete joke

Raiders goign to teach Chiefs and DVOA system good lesson on Sunday

Raiders 31, chiefs 10

by Chiefs (not verified) :: Tue, 11/02/2010 - 7:55pm

you might be right........... but I doubt it

welcome back to reality raiders

by NotRaiderJoe (not verified) :: Tue, 11/02/2010 - 8:05pm

This guy again.

by dbostedo :: Tue, 11/02/2010 - 10:46pm

Yes. Again. And hopefully again and again and again.

by bravehoptoad :: Wed, 11/03/2010 - 12:43pm

Is there a more popular poster on the site?

by Spielman :: Thu, 11/04/2010 - 10:32am

Not even close.

by Shattenjager :: Thu, 11/04/2010 - 5:03pm

Is there another "popular" poster at all?

by chemical burn :: Thu, 11/04/2010 - 5:08pm

zlionsfan has some cache, of course, for the template.

Also, bobman, Karl Cuba and Will Allen seem like really good dudes, but that probably doesn't generate "popularity."

by Pat (filler) (not verified) :: Thu, 11/04/2010 - 10:15pm

See, of course, what I should've done is somehow name ROBO-PUNTER after myself. Damn zlionsfan and his secretly forward-thinking self-aggrandizement.

by Jerry :: Fri, 11/05/2010 - 2:35am

(filler) (not verified) (punter?)

by Shattenjager :: Sat, 11/06/2010 - 12:09am

Most of the posters here generally seem to be decent people and not stupid. However, there are only three posters I specifically look for besides the FO staff. I find it difficult to imagine that most posters specifically look for lots more than three people.

Incidentally, as someone who was not reading FO in the ROBO-PUNTER days, I actually had thought that one of the FO writers made him up. I apologize, Pat (filler), for not giving you proper credit in my mind.

by ammek :: Wed, 11/03/2010 - 1:41pm

Signs the Raiders' luck may change:

— 32nd in variance;
— Of their 18 fumbles, they've lost only two;
— They're the healthiest team in the league.

by Pat (filler) (not verified) :: Tue, 11/02/2010 - 6:12pm

Seriously, San Diego, WTF. I mean, *really*, WTF. Special teams is fluky. A crazy-bad performance doesn't usually continue to repeat itself. But at this point, you have to say that San Diego's special teams are just historically awful. By this time in the season, the godawful Minnesota special teams from 2008 were climbing back to single-digits.

And my guess is that they're actually worse than they appear, because their worst unit - their punting unit - isn't seeing the field as much as on an average team.

by Formersd (not verified) :: Tue, 11/02/2010 - 6:25pm

I've been saying it for a few weeks, but the Chargers really need to consider just going for it on 4th down and stop punting. The punt unit is just that bad.

by GoDillos (not verified) :: Tue, 11/02/2010 - 7:58pm

And we are less than 2 years from when Mike Scifres turned into robo-punter in the wild card round. All Colts fan sigh.

by coboney :: Tue, 11/02/2010 - 8:09pm

Sadly even Robo-Punter needs 10 other guys to play semi-competently. So does Scrifes.

by tally :: Tue, 11/02/2010 - 8:20pm

Actually, one of the features of Robo-Punter is that he doesn't need 10 other guys--he just punts the ball right through any potential punt-blocker.

Scifres doesn't have that feature installed.

by OKChristian (not verified) :: Wed, 11/03/2010 - 10:26am

And he also gets down field and tackles the guy at the 1 yard line by himself.

by DeltaWhiskey :: Wed, 11/03/2010 - 10:41am

When would he ever have to do that?

by tally :: Wed, 11/03/2010 - 4:15pm

He doesn't. Any ball punted by Robo-Punter™ is automatically downed by burying itself at the 1.

by chemical burn :: Wed, 11/03/2010 - 4:19pm

I thought he put a special spin on his kicks that they are impossible to field, so every robo-punt leads to a fumble if handled.

by Bobman :: Wed, 11/03/2010 - 12:58pm

Sigh... bang head on concrete floor... it's a fine line, really.

by Mr Shush :: Wed, 11/03/2010 - 7:57am

The Texans do too, though in their case it's the defense that should never be allowed to see the field if it can in any way be avoided.

by bag (not verified) :: Tue, 11/02/2010 - 6:24pm

I know we all thought the NFC was worse than the AFC, but the top 15 has only 5 teams from the NFC. If playoff spots were purely created from DVOA, then Indy is the #6 seed in the AFC with a DVOA of 15%. The NFC would have to put in SF with a solid negative 4% DVOA and 2-6 record.

by Formersd (not verified) :: Tue, 11/02/2010 - 6:27pm

There's going to be some good teams in the AFC that won't be making the playoffs. My guess is that Miami, Tennessee, San Diego and Houston end up staying home and being very disappointed.

by Mr Shush :: Wed, 11/03/2010 - 8:03am

I'm emphatically not buying on the Chiefs as a top team (as opposed to a solidly decent one, which they are), and still expect SD to catch them. Miami and Houston will both stay at home, and in fairness they are both definitely less good than whichever of NYJ, NE, BAL, PIT, IND and TEN misses out. I think I agree that the Titans are the most likely most unlucky.

by Jetspete :: Wed, 11/03/2010 - 10:03am

SD's schedule will prevent that. KC should get to 11 wins, and SD still has to go to Indy and Houston (where KC lost). Plus, unless KC loses a division game SD will lose any tiebreaker because of that loss to oakland.

by WhyCantIJustBuy... :: Wed, 11/03/2010 - 8:20am

I thought that, then I saw the Jets and Steelers play some NFC underachievers on the weekend. I don't think the Falcons or Giants would be losing to the Colts or Dolphins.

by Marcumzilla :: Wed, 11/03/2010 - 6:22pm

On the contrary, I believe the Giants could lose to the Colts. In fact, it wouldn't even be close if the Giants stick to a dime defense.

by ammek :: Wed, 11/03/2010 - 3:06pm

The big difference is in pass offense. Nine of the worst 10 pass offenses are in the NFC, including all three total stinkers.

Who will be the three NFC probowl QBs? Brees, certainly. Who else? Rodgers? Ryan? Freeman?

The best left tackle in the NFC might just be Jeff Backus. Yes, of Detroit.

by chemical burn :: Wed, 11/03/2010 - 3:13pm

Manning will go if the Giants keep winning; he seems even more of a shoo-in than Brees. If one of the two Eagles QB's gets a decisive amount of playing time, he'll go - especially if it's Vick. Rodgers and Ryan are strong possibilities, especially if they manage to get high TD totals. Freeman I think would have to catch fire in a crazy way and lead the league in some stat to get real consideration...

by Independent George :: Tue, 11/02/2010 - 6:27pm

Yes! The Giants are out of the top 5! I can't begin to tell you what a huge relief that is.

by iapetus (not verified) :: Tue, 11/02/2010 - 6:52pm

Everyone knows that Tampa Bay is the best team in the NFC.

by Jeff Fogle :: Tue, 11/02/2010 - 7:23pm

Evidence against Kansas City having one of the top six offenses in the league:

*They've bludgeoned three teams, but struggled significantly versus four teams. Two of the three they bludgeoned rank 29th and 30th in defensive DVOA.

*In the four poor outings:
4.0 YPP, 7 points on drives of 60+ yards, 1 of 11 on 3rd Downs (SD)
4.6 YPP, 6 points on drives of 60+ yards, 5 of 15 on 3rd Downs (Cle)
4.5 YPP, 0 points on drives of 60+ yards, 1 of 10 on 3rd Downs (Indy)
5.6 YPP, 7 points on drives of 60+ yards, 4 of 15 on 3rd Downs (Buff)

Opposing DVOA defensive rankings in those games
San Diego 4
Cleveland 20
Indy 19
Buffalo 32

So..less than 5.0 YPP, 5 PPG on long drives (abysmal), and 11 of 51 on third downs (22%--really abysmal) vs. a collection of DVOA defenses that averages 18.75 in rank and splits the median at 19-20.

If you abuse bad defenses, but struggle vs. average or good defenses (general rules through 7 games...with counter-exceptions being a good game vs. SF and a horrible offensive game through 5 quarters vs. 32nd ranked Buffalo), that's not generally considered a top six performance. It's more obvious if you're playing a balanced schedule. If you're playing a tough schedule, everyone thinks you have a horrible offense.

*Jim Armstrong's Drive Data here at the site ranked Kansas City 15th in Drive Success Rate before they scored 13 points in five quarters on the Bills. Note that Armstrong doesn't adjust for strength of schedule.

Also wanted to add...

Aaron said:
"The Chiefs are 31st in the NFL with 187 net pass attempts, but when they do pass, they've been effective"

16-27-1-250 (SF)
20-29-0-189 (Houston)
13-18-0-190 (Jax)

49-74-1-580 in the three good games (66% completions, 7.8 YPA)
56-105-2-522 in the four bad ones (53% completions, 5.0 YPA)

When they passed vs. SD, Cleveland, Indy, and Buffalo, they obviously weren't effective or they would have been converting third downs, putting long distance points on the board and posting strong YPP numbers (plus good passer ratings for Cassel, who posted 68.0, 46.1, 70.5, and 84.1 in them).

Aaron said:
"the Chiefs need to be taken seriously as a Super Bowl contender."

If you can't drive the field for points unless you're playing lousy teams, how are you going to score in January on Pittsburgh (2nd in DVOA Defense), the NY Jets (10th in DVOA defense), or Indianapolis (9 points already in a road meeting). It will be tough at home, more difficult if KC can't earn home field advantage after the first round. The Chiefs could be one of those "win three low scoring coin flip" type Super Bowl contenders. Those aren't the ones you take seriously...particularly when inexperienced players (in terms of the postseason) are the ones you're asking to win coin flips.

The schedule is soft the rest of the way (note that Sagarin ranks their schedule 31st to this point, compared to 23rd for FO). They could definitely put up some big numbers vs. the softer defenses they face (two games ahead with Denver). Make them do something vs. a top 10 defense before you call them a top 6 offense. To me, this ranking is a huge strike against DVOA. A preponderance of evidence does not suggest KC is a top 6 offense. It suggests they're an average offense that can abuse bad defenses because of a strong running game...but has trouble vs. quality in a way that virtually prevents them from driving the field for points more than once a game. That's my take anyway...

by Tracing plan (not verified) :: Tue, 11/02/2010 - 8:25pm

Well you don't really get what objective analysis is vs. cherry picking numbers to make a case, but I'll start with your first problem.

You have essentially argued that KC isn't good on offense vs. top defenses (which may or may not be true, but to be relevant the question is how they do compared to other offenses against top competition) and concluded that they are not a "top 6" offense. But the first part is irrelevant. You instead need to name six teams that are better.

by Jeff Fogle :: Tue, 11/02/2010 - 9:22pm

Appreciate the response TP. I would disagree that DVOA is an objective analysis. And, to make a case, you have to focus on the points that best make the case or your case would last several thousand words.

I haven't essentially argued that KC isn't good on offense vs. top defenses. I've argued that KC isn't good on offense vs. most everyone above horrible. Buffalo ranks 32nd, and KC scored 13 points in five quarters. Two of the other three from the bad games were at 19 and 20. There is one sample of KC playing well offensively vs. a good defense (SF, ranked 11th in DVOA). That's the outlier right now for performance vs. teams ranking better than bottom of the barrel.

The Drive Data here at the website from Jim Armstrong lists 14 teams better than KC at drive success rate even before adjusting for schedule (well, that's before the Buffalo game is thrown in, that page hasn't been updated yet). I'll just name those: Indianapolis, San Diego, New England, Houston, Atlanta, New Orleans, Dallas (with Romo), Green Bay, Baltimore, Jacksonville (with Garrard), NY Giants, Philadelphia, Miami, Cincinnati.

Now, I'm not fond of the Cincinnati offense...and a healthy Jax, Philly, or Miami would certainly be debatable within the hunk of "okay" so to speak. I don't think it's hard to pick six teams from that listing who are better offensively than KC. Perhaps KC's big numbers vs. the awful defenses of Houston and Jacksonville are weighing too heavily in the DVOA process. If you take their "median" performance rather than letting those big games warp the averages or full season volume...it paints a different picture.

by Tracing plan (not verified) :: Tue, 11/02/2010 - 10:33pm

Lots of logic problems still.

DVOA is objective. It is not subjective. So is drive success. Both are numerical measures. You saying KC is good or bad--that is not objective.

You say KC is no good against "most everyone above horrible." They had mediocre games against 2 middle of the road teams, a good game against a good team. You say this is an outlier. After 7 games there is no outlier. We just barely have trend as of yet.

The unadjusted Drive Data is problematic--it is not adjusted for opponent. You wonder if DVOA is not adjusting ENOUGH for the poor quality of opponents, yet you site completely unadjusted stats. Very curious.

You could certainly argue about several teams below KC in DVOA, but none would be a slam dunk. And KC is likely to be on an upswing as they get more and more practiced in their offense.

by Jeff Fogle :: Tue, 11/02/2010 - 11:03pm

Looks like we'll be disagreeing about a lot TP...even about what logic means.

The choices that were made in designing DVOA were subjective. The output reflects that subjectivity. DVOA is a formula that starts with subjective opinions about the value of plays/performance/production and uses them to tabulate a numerical value. The fact that a final number is expressed numerically doesn't erase the subjectivity that went into the process.

If you're restating what I said, be sure to properly characterize it. The games against 2 middle of the road defense weren't mediocre (your subjective word). The numbers were poor (my subjective word).

You can have outliers after a few games. It's obviously dangerous to draw too many conclusions from the outliers. It's also dangerous to ignore the outliers.

What's curious? I show poor performance in more than half the schedule even BEFORE accounting for a weak strength of schedule. It's unlikely performance would get BETTER vs. a tougher schedule. It's not curious at all. It's a quick way to make the point. With logic.

KC is likely to be on an upswing given the last data point was 13 total points in five quarters against the 32nd ranked DVOA defense? They are likely to bounce back from that poor offensive outing. The trend, to the degree one exists, isn't showing an upswing. It's showing two extremely strong showings vs. bad defenses, one strong showing vs. the 11th defense, and four poor showings vs. a mix of defenses that centers around 19th. If you're going to deduce a trend, you have to do it from those data points in my view.

Unlikely we're going to agree on much I'm afraid...

by DeltaWhiskey :: Wed, 11/03/2010 - 9:56am

"DVOA is objective"

This is not true. The "success points" that DVOA is based on were not empirically derived. There is no evidence that the various tweaks to DVOA over the years were derived in an emprically valid manner.

"The unadjusted Drive Data is problematic--it is not adjusted for opponent"
DVOA's greatest claim is it's adjustment for opponents; however, historically VOA has a slightly better correlation with wins than DVOA...I don't know if that's a statistical artifact, but it suggests to me that the "D" adjustments are over-rated in their value.

by DeltaWhiskey :: Wed, 11/03/2010 - 9:57am

Damn, should have read along farther.

by Arkaein :: Wed, 11/03/2010 - 11:56am

VOA is supposed to correlate higher than DVOA to wins in the current year, because VOA only considers play relative to the opponents faced.

If a team is average and plays an average schedule resulting in a 0.500 record they should have VOA and DVOA of 0%. However if a team is above average (say DVOA of 20%) and plays an above average schedule (say this is also opponent DVOA of 20%), you would still expect a 0.500 record. In both cases the team's are playing even with their opponents and would expect a 0% VOA.

The difference is that the team that has the higher DVOA should be expected to look better against a future schedule of opponents. DVOA was designed to correlate with wins in the second year. So VOA's higher correlation with wins in the same year is entirely expected. Just don't use it to predict future success.

by tuluse :: Wed, 11/03/2010 - 4:56pm

The "success points" that DVOA is based on were not empirically derived.

I'm pretty sure they are, they got the basis for the formula from The Hidden Game of Football.

by bravehoptoad :: Wed, 11/03/2010 - 12:44pm

I'm shaking my head, hearing SF called the "good team" here.

by tuluse :: Tue, 11/02/2010 - 11:31pm

DVOA is objective in that it treats all teams and all plays equally.

by greybeard :: Wed, 11/03/2010 - 2:06am

You are confusing consistency with objectivity.
Passer rating treats all QBs, all passes, all touchdowns, all interceptions and all completions equally, yet not many people would not call it objective measure of the quality of passer because it does not take into sacks and opponents into account. DVAO is even more subjective than passer rating in that it assumes that there is a success criteria such as 5 yards on first down whether the play was a pass or run, that there is a sane way to adjust for opponent, etc.

by qed :: Wed, 11/03/2010 - 8:16am

I would say QB rating, DVOA, etc. are all "objective" measures of quality in that they don't include any manual adjustments by human beings. That doesn't mean they are GOOD measures of quality, just that they don't contain any subjective input.

If I decide to rank all RBs purely by their fumble rate I'd still call that an objective measure of quality, it's just a poorly-designed metric.

by RichC (not verified) :: Wed, 11/03/2010 - 9:31am

DVOA is subjective. Its consistent, but its fundamental design features the designers chosing that certain things are more important that others.

DVOA assumes certain things, and weighting those is subjective.

Your "Fumble metric" would be objective if you said "I'm measuring how often runningbacks fumble". If you are using it as a measure of running back QUALITY, its now subjective.

Why does FO have porn adds in the sidebar now?

by OKChristian (not verified) :: Wed, 11/03/2010 - 2:52pm

From my perspective, DVOA is clearly an objective stat because it treats all input elements the same regardless of anyone's personal opinions.

All measurements and all systems are created with a human hand, so there is always a human element. But to say the DVOA is subjective because it was created by a human with inherent subjectivity is just wrong. That is like saying that measuring how many yards someone ran is a subjective assessment because someone alone the way subjectively decided exactly how long a yard is. Somebody had to make the final decision about the exact length of a yard, just like Aaron makes the final decisions on what the measuring stick of DVOA is.

And DVOA is not a judge of quality as someone suggested above. It is a measurement of how much better or worse a team performs against an average represented by a percentage score. When someone looks at these numbers and starts making quality judgments based on the numbers, that is the individual's subjective opinion of an objective stat.

Now you are free to make your own subjective opinion about the quality or value of DVOA, but you can't say that DVOA is not an objective stat. It is no more subjective than stats like completed passes, points per drive, or the Patagonian win expectation. It is just way more complex.

But hey...I guess all this is just my subjective opinion on the definitions of two words created by humans.

by dmb :: Wed, 11/03/2010 - 3:59pm

Jeff Fogle and DeltaWhiskey are correct: there's more than one definition of "objective," and DVOA does not meet all of them. It certainly is objective in the "consistency" sense, as greybeard and RichC helpfully put it. And by that definition, every common quantitative measure is objective.

But people frequently use "objective" to mean something akin to "devoid of human / discretionary input," and DVOA is most certainly not objective by that standard. As DeltaWhiskey noted, there are discretionary judgments (e.g., the standards for a "successful" play) that are built into DVOA. Aaron has also stated frequently that every offseason he experiments with different modifications, and decides which ones will remain a permanent part of the stat. These factors influence the output, and at some level, are based on a person's subjective evaluations. (This is what distinguishes DVOA from, say, completions. Although subjective evaluations are sometimes required to figure out if a play meets the definition of a completion, the definition itself comes from a rule that provides an inherent standard. DVOA, on the other hand, uses standards that are not literally built into the structure of the game.)

That's not necessarily a bad thing. In fact, almost all worthwhile quantitative analyses have some built-in subjective assumptions, and any productive discourse about such work requires acknowledgment of those assumptions. Around here, those subjective evaluations are (typically) informed by empirical evidence and sound logic, which helps strengthen the case for DVOA as a useful stat.

by Aaron Schatz :: Thu, 11/04/2010 - 6:07pm

Hey folks.

I think we addressed these issues in the FO Basics series in the preseason:


by DeltaWhiskey :: Fri, 11/05/2010 - 8:13am


Thanks, hadn't seen these. Probably b/c when they ran I was w/o consistent computer access. I'll definitely read them here soon.

by AudacityOfHoops :: Tue, 11/02/2010 - 8:28pm

I am by no means saying you are wrong, but I'd like to see some context for most of these numbers. Do you have a source where you can look them up easily, or do you compile them yourself? I'm just wondering what a top 6 offense would look like in these metrics:

YPP, PPG on long drives, and 3rd down conversion rate in 4 worst games

For that matter, I want to know what any team looks like on average in the "PPG on long drives" metric. That one's new to me, but I can see how it would be useful.

by Jeff Fogle :: Tue, 11/02/2010 - 9:28pm

Thanks for asking for clarifications AoH. I keep 60+ scoring by hand. I'll try to write up full season numbers in this space later tonight. Tougher to go through and cross-index performances vs. caliber of opposing defenses for all 32 teams. But, the full season averages will provide some context.

It's a little dicey to use "four worst games" for everyone obviously, because not everyone has played as weak a schedule as Kansas City has (23rd by FO, 31st by Sagarin). Some "four worst" games are going to come against all top 10 teams I believe (extreme imbalances in schedule this year). KC's come against 4-19-20-32 in opposition defensive DVOA. I'll see what I can do to provide context in that area...maybe direct comparisons to teams that rank closely to KC in the FO rankings above...

by AudacityOfHoops :: Tue, 11/02/2010 - 9:54pm

Thanks, that will be awesome to see. In the meantime, I looked at the NFL.com gamebooks for the Pats and Colts, and jotted down their numbers (I probably made a mistake somewhere in here).

6.5 YPP, 21 points on drives of 60+ yards, 9 of 13 on 3rd Downs (Cin)
5.1 YPP, 14 points on drives of 60+ yards, 5 of 11 on 3rd Downs (NYJ)
6.7 YPP, 28 points on drives of 60+ yards, 7 of 12 on 3rd Downs (Buf)
4.5 YPP, 10 points on drives of 60+ yards, 5 of 11 on 3rd Downs (Mia)
5.4 YPP, 20 points on drives of 60+ yards, 4 of 14 on 3rd Downs (Bal)
3.1 YPP, 7 points on drives of 60+ yards, 3 of 12 on 3rd Downs (SD)
6.7 YPP, 21 points on drives of 60+ yards, 6 of 11 on 3rd Downs (Min)

6.7 YPP, 17 points on drives of 60+ yards, 4 of 12 on 3rd Downs (Hou)
5.9 YPP, 14 points on drives of 60+ yards, 4 of 11 on 3rd Downs (NYG)
5.6 YPP, 14 points on drives of 60+ yards, 6 of 14 on 3rd Downs (Den)
6.2 YPP, 21 points on drives of 60+ yards, 7 of 10 on 3rd Downs (Jax)
4.5 YPP, 13 points on drives of 60+ yards, 8 of 17 on 3rd Downs (KC)
6.9 YPP, 14 points on drives of 60+ yards, 3 of 10 on 3rd Downs (Was)
4.9 YPP, 17 points on drives of 60+ yards, 7 of 15 on 3rd Downs (Hou)

by Jeff Fogle :: Tue, 11/02/2010 - 10:15pm

Thanks for that AoH!

If we grab the "worst four" for both...you see that there's still long distance scoring volume...and something that's at least better than 20% on third downs if not particularly impressive. Of course...these are great offenses led by Hall of Fame QB's. But, it gives you a sense of how far KC is behind these guys in terms of consistency...and these guys played tougher strengths of schedules.

by Jeff Fogle :: Tue, 11/02/2010 - 10:07pm

Went through my spreadsheets for scoring on drives of 60+ yards. Grabbed averages and medians. These aren't adjusted for strength of schedule. Medians help in excluding outliers...but that middle game obviously doesn't come against the 16th ranked team for everyone. Not endorsing using this stat in a vacuum. I do like the combination of this stat, third down conversions, turnover potential because of style, and a few others in terms of building an overall profile of an offense.

What follows is two lists of 32 teams, so I'll try to cluster for easier digestion.

Best: San Diego 18.6, New England 17.3, Atlanta 17.1, Indy 15.7, Houston 15.3, Oakland 14.8, Baltimore 14.3, NY Jets 14.0, Philadelphia 13.3, Tennessee 13.0

Middle: Denver 12.8, Kansas City 12.3, NY Giants 12.3, Dallas 12.0, Green Bay 11.6, New Orleans 10.8, Buffalo 10.4, Miami 10.2, Cincinnati 10.1,

Worst: Cleveland 9.4, Washington 9.3, Minnesota 9.1, Tampa Bay 9.0, Detroit 9.0, St. Louis 9.0, San Francisco 8.8, Pittsburgh 8.7, Arizona 8.4, Seattle 7.7, Chicago 7.7, Jacksonville 7.4, Carolina 7.0

Kansas City ranks 12th in this stat. You'll see in a moment they were helped by the huge games vs. Houston and Jax.

20: New England
19: San Diego
17: Houston
14: NY Jets, Indianapolis, Philadelphia, Dallas, Green Bay, Cleveland
13.5: New Orleans
13: Baltimore, Atlanta
12: Tennessee
11.5: Oakland, Denver
10: Miami, NY Giants
9: Minnesota
8.5: Washington, St. Louis
7: Kansas City, Buffalo, Cincinnati, Pittsburgh, Jacksonville, Chicago, Detroit, Carolina, Tampa Bay, Seattle, San Francisco, Arizona

Here, Kansas City is tied for last (21st) with a hunk of teams that aren't setting the world on fire offensively.

The consistency isn't there. Two huge games pumped up their volume in a way that suggests KC deserves to be considered better than the others at 21-32 in median. But, all the way up to top six? When all this is happening against a weak schedule?

If I had to pick offenses I think are better than KC (and I'll leave out Pittsburgh with Roethlisberger because DVOA has to use the stats compiled with other QB's for Pittsburgh to this point): New England, San Diego, Indianapolis, Houston, New Orleans, Atlanta, Baltimore, NY Jets, Tennessee, NY Giants, Dallas with Romo (irrelevant now, but looking at earlier numbers). That's more than six. Trying to answer multiple questions with this post I guess.

Regarding the article as a whole...I don't think KC has a top six offense. And, I think their general struggles offensively do disqualify them from serious Super Bowl consideration. I think the game-by-game evidence is persuasive that they'd struggle in January vs. most or all of the opposing defenses they'll likely be facing (there's still a mathematical chance of course that Jax and Houston could earn a Wildcard, and visit KC in a first round game). NE's defense has graded out poorly in most metrics...which would give KC a chance in Foxboro. The poor showing at Indy is a bad strike against them in my view. What would that project to at Pittsburgh or at the NYJ?

by Jerry :: Wed, 11/03/2010 - 2:17am

I haven't watched the Chiefs at all, and if you'd asked me about them before this article appeared, I'd have said that they're doing well enough against crappy opposition. Maybe they're not the sixth-best offense in the league, and I doubt they're the best team, but I'm now reasonably sure that they're better than I thought they were. That's a big reason why I appreciate this site.

By the time January rolls around, we'll have a better handle on every team, and we still won't know what's going to happen in the playoffs.

by RichC (not verified) :: Wed, 11/03/2010 - 9:35am

The problem with judging by "Points scored on drives longer than 60 yards" is that you're essentially penalizing KC for having a much better defense than Indy or NE (and they do).

As to struggling in january, don't, by definition, all playoff teams "struggle" except one?

by AudacityOfHoops :: Wed, 11/03/2010 - 10:41am

You might also penalizing them for having a strong rushing offense, which could reduce the number of drives per game (I can't find this number anywhere to check).

A fairer measure would be "points per drive that starts at the teams own 40 or worse."

Also, just for fun:

Ridiculous: Detroit 17.1, Tennessee 15

Best: Jacksonville 13.2, NY Giants 12.7, Pittsburgh 12.3, New England 12.0, Indy 11.9, Oakland 11.7

Upper Middle: Philadelphia 11.3, Kansas City 11.0, Cincinnati 10.8, Arizona 10.6, Green Bay 10.4, Tampa Bay 10.4, Chicago 10.3, New Orleans 10.1, Washington 10.1, Dallas 10, Seattle 9.9

Lower Middle: Minnesota 9.3, Houston 9.0, Miami 8.8, NY Jets 8.7, St. Louis 8.5, Buffalo 8.3, San Francisco 8.3

Worst: San Diego 7.6, Cleveland 7.5, Atlanta 7.0, Baltimore 7.0, Denver 6.4,

Also Ridiculous: Carolina 5.1

by Jeff Fogle :: Wed, 11/03/2010 - 3:12pm

Agree AoH that it would be a fairer measure. Not sure if it's worth the time on my end to go through all the drives to get a ratio though. The raw counts have always done such a good job of paiting a picture...particularly when used in tandem with other indicators stats like third down conversions...that it didn't seem worth the trouble.

Now, if I were going to come up with a stat I was calling the "ultimate" long distance scoring stat or something, I'd clearly have to do that. Agreed.

I started keeping these raw counts back when the USA Today first went to their expanded boxscores a couple of decades ago. Didn't want to do the extra math per game to estimate drives at the time. As internet resources developed that showed drive charts...I didn't see the value at the time of going the extra step in terms of my own purposes with the stat. Old habits die hard (lol), especially as your energy drains as you get older. Looks like the Drive Data here at FO is doing what you suggest...which saves us all the trouble!

In terms of penalizing strong rushing offenses...I don't believe that to be the case over time. If you have a strong rushing offense, you can drive the field for 13-14 points per game and register well in this stat. There just hasn't been a case during my studies of a high quality team with a strong running game that graded out poorly at driving the field for points. A strong running game helps you in the red zone, and gives you big play potential to pop a long run for a TD. I'd agree that it's hard to "blow up" to a super-high average because you don't play many shootouts. But, you'll have a solid average, and an excellent median if you're a quality grinder...it that makes sense.

by Jeff Fogle :: Wed, 11/03/2010 - 3:03pm

Not sure why you'd draw that conclusion RichC. Why would having a much better defense than Indy or NE (which is debatable given the strength of schedule differences involved) keep Kansas City's offense from driving for two touchdowns per game on long drives? A good defense means the other team isn't running out the clock on you. A good defense means the other team is punting and you're getting the ball with a chance to drive the field.

I understand this is something that people have to take my word for because these numbers aren't widely circulated. I've been doing this since USA Today went to the expanded boxscores a couple of decades ago (before the internet boom). It hasn't been true that having a quality defense penalizes your offense in terms of its ability to score points on long drives. Good offenses score points on long drives. They get enough opportunities over a series of games, a half season, or a full season to prove whether or not they can do it consistently.

Quick examples from last year, using performances only in this stat:
Indy finished 5th in long distance scoring defense, 2nd in offense
Green Bay finished 7th in defense, 5th in offense
Dallas finished 8th in defense, 9th in offense
Minnesota finished 9th in defense, 3rd in offense
Baltimore finished in a tie with Minny at 9th in defnese, 6th in offense

The best teams are good at BOTH of these things...and being good in defense last year didn't prevent the offenses from putting up numbers. Minnesota was a top 10 defense yet scored 17.5 ppg on drives of 60 yards or more. Indy had the best split in the NFL at 9.1 on defense and 18.3 on offense.

In terms of struggling in the playoffs. The topic here is Kansas City's OFFENSE, not the whole team. Kansas City's OFFENSE is likely to struggle in January vs. playoff caliber defenses (particularly on the road) based on the evidence to date...unless they get to face Houston or Jax in a playoff game (meaning somebody from NE/NYJ or Pitt/Baltimore didn't make it), or New England. And, it's obviously not true that 11 offenses struggle in the postseason and one doesn't.

Green Bay didn't struggle offensively in its playoff loss last year
Minnesota didn't struggle offensively outside the awful late pick from Favre
Indy didnt' struggle offensively for the playoffs as a whole, but they weren't the ultimate winner.

Did Arizona's offense struggle in the playoffs two years ago when they made it to the Super Bowl before losing a thriller to Pittsburgh?

Kansas City's offense is likely to struggle in January based on the evidence so far in my view. That means they don't yet deserve serious consideration as a Super Bowl threat.

by B :: Wed, 11/03/2010 - 3:33pm

A team with a good defense will generally have better starting field position than a team with a lousy defense. And that means the team with a good defense will have less opportunities to score long touchdowns. For example, say Team A has 5 drives that start on their 40 or worse, and they score on 2 of them, and team B has 2 drives that stat on there 40 or worse, and they score on 1 of them, by long scores per game, Team A is better, but on long scores per opportunity, team B is better.

by AudacityOfHoops :: Wed, 11/03/2010 - 3:51pm

The concept makes sense, but I don't think the effect is strong enough to make the kind of difference we see in the Chiefs' numbers. I mean, look at FO's drive stats. KC has only the 12th-best starting field position. With that slightly-better-than-average starting position, they are still only 17th in points per drive, and they are 12th in drive success ratio. All against a relatively weak schedule.

I hope I'm wrong, but I just don't see it.

Here's an interesting thought. Tell me if this seams off-base: A team can be better than average at rushing AND at passing, but still worse than average overall on offense, if they run too much (if passing has an overall better success rate, which I believe it does).

by B :: Wed, 11/03/2010 - 11:33pm

Maybe. A teams passing DVOA is the average of all their passing plays, and rushing DVOA is the average of all their rushing plays. However, total offense isn't just the two added together, it's weighted by the percentage of plays that are rushing vs passing. So the Chiefs, who run more, will have their offensive DVOA weighted more towards rushing plays. What you will see with teams that favor rushing or passing is defense will key on stopping that, making passing or rushing easier, so I'm sure the Chiefs passing DVOA is inflated by that effect, but the total offensive DVOA isn't inflated as much.

by Jerry :: Thu, 11/04/2010 - 3:59am

Simpson's paradox. The guys at the p-f-r blog did a few posts with it.

by Jeff Fogle :: Wed, 11/03/2010 - 3:51pm

A team with a good defense creates the possiblity for MORE possessions for their own offense...by forcing punts and taking the ball away. So, better field position, but more overall possessions. Somebody will have to show that teams with good defenses have fewer long distance scoring opportunities. Many of last year's best defenses saw their offenses light up the board in long distance scoring.

Also, having a BAD offense can create more opportunities because the flip flop of possessions goes quicker. Indy has only had 71 possessions this year, but rates out near the top of the league again in driving the field for points. It's a complicated mix of stuff, so be careful trusting assumptions.

Jim Armstrong has updated his drive data here at the site.
Best field position this year: Tennessee, average own 32.55 yard line
Worst field position this year: Denver, average own 26.58 yard line

All teams average starting on their own side of the 40. In terms of opportunities, beginning a real count seems like it's going to be worthwhile if this kind of stuff is going to become part of mainstream discussions.

by AudacityOfHoops :: Wed, 11/03/2010 - 3:41pm

I basically agree with your opinion on the Chiefs. However, as a Chiefs fan, I find it awesome that "do the Chiefs deserve serious consideration as a Super Bowl threat" is even a possible topic of discussion. In past years, that conversation would be over in one word.

by Jeff Fogle :: Wed, 11/03/2010 - 4:26pm

Went to grab Chiefs starting yardage from the drive charts at ESPN. Not going to do this for all 32 teams (lol), so we have to kind of ask if this is "consistent" with what we perceived to be the performance of a top six offense based on what we see with our own eyes from NFL offenses:

Versus San Diego: 12 of 13 drives start inside the 40 (one TD drive).

At Cleveland: 11 of 13 drives start inside the 40 (two field goal drives)

Versus SF: 9 of 12 drives start inside the 40 (one TD drive)

At Indy: 10 of 10 drives start inside the 40 (no scoring)

At Houston: 10 of 10 drives start inside the 40 (4 TD Drives and a FG)

(Stopping here for a second, does anybody want to still try and make the case that Kansas City isn't getting enough chances to drive the field because their defense is somehow getting them the ball outside the 40 or something? Only 5 of 58 drives started outside the 40 to this point...and it's going to end 6 of 78).

Vs. Jacksonville: 7 of 8 drives start at the 40 or closer (4 TD Drives)

VS. Buffalo: 12 of 12 drives start at the 40 or closer, (1 TD Drive)

That's two excellent performances vs. the horrible defenses of Houston and Jacksonville. Then, 3 TD's and 2 FG's out of about 54 drives that start inside the 40 vs. everyone else (I'll say about because it's possible I miscounted while doing this by hand, or that espn mislabeled some drives. I caught one they mislabeled but may not have caught others). Is that enough of a sample size to draw conclusions about KC's inability to drive the field for points when facing defenses who are better than horrible to this point in the season? Do I really need to go through and show anyone that more than five other offenses were better than that?

And, after reading through five games of clear and unmistakeably obvious struggles driving the field for points vs. the teams who weren't Houston and Jacksonville, do you accept DVOA's numerical assertion that Kansas City is one of the best offenses in the league? Two huge outings vs. bad defenses have created that illusion. If KC doesn't face lousy defenses in the playoffs, it's hard to take them seriously as a Super Bowl candidate based on this evidence in my view.

by AudacityOfHoops :: Wed, 11/03/2010 - 5:09pm

Not that it makes a difference in the conclusion, but shouldn't they have 3 FG's above vs. Indy? Or if those were short field, fewer drives starting inside the 40.

But still, yep, does not look too hot to me. But the KC homer in me is holding out hope that the pattern we see is just due to chance. (If you switch their production in the Jacksonville and San Diego games, their numbers all look basically the same - same production vs same opponents. But now all of a sudden, their big games don't solely come against bad offenses, and the conclusion isn't as strong.) If I have to put money down, I don't believe this theory. But I can still hope it's true.

by AudacityOfHoops :: Wed, 11/03/2010 - 11:36am

You mentioned "turnover potential by style," and I think that's something that's being overlooked in this conversation. The Chiefs' turnover rate this year has been historically low. Here are all the teams with a TO/drive lower than 0.09, in the time period covered by FO's drive stats (1998-2010):

2010 KC .051
2008 MIA .073
2006 SD .073
1998 DAL .074
2008 NYG .077
2002 KC .079
2009 GB .081
2002 JAC .081
2007 NE .082
2005 DEN .082
2010 NYJ .083
2005 SEA .083
2005 JAC .087
2003 KC .087
2004 NYJ .088
2010 NE .088
2006 STL .088

For reference, this year's median is .125. So the difference between 2010 KC and 2004 NYJ (15th best over the past 13 years, i.e. a typical league leader) is as large as the difference between 2004 NYJ and this year's median. Obviously we expect some regression towards the mean, but I don't think DVOA does. DVOA just sees that KC never turns it over.

Now, this hasn't translated into points as well as one would hope, but one of the key tenets of this site is that points aren't the only measure of an offense. Chiefs opponents have had the worst average starting field position in the league, in large part due to them rarely being handed a short field. I would love to see some stats on how often a team's opponents have taken over the ball in the wrong half of the field, and how many defensive touchdowns have been scored against an offense. I imagine KC would rank highly in both of these.

by Bobman :: Wed, 11/03/2010 - 1:12pm

To add something to the turnover luck view, see how well NO did last year compared to this year, in traditional stats, record, FO stats... you name it. Then look at their TO success in 2009 vs 2010 (incl pts scored off TOs). It's not the only reason for their 09 success nor for their 2010 "good but not great" status.... but it contributes to each side.

If KC's TO luck changes for the worse--doesn't make them a lesser team--but could make the road a lot tougher for them. (Caveat--timing helps--if they lose six fumbles in a big loss to their toughest opponent, that may even out the TO situation and get them back to "par," but does not necessarily affect their record and post-season chances. But if they start losing 1-2 more TOs per game for the next 1-2 months, they might lose an additional couple games that their previous TO luck would have had them pulling out, put their D in deeper holes, etc, and tarnish their record, traditional stats, and FO metrics along the way.)

by Jeff Fogle :: Wed, 11/03/2010 - 3:19pm

Agree AoH that turnover avoidance has been a key element of the KC offense so far. How much of that is from playing a crappy schedule?

Also a danger of comparing 7-game stats from one season to full season stats from others. Less time for the math to work itself out. Though, if KC's schedule is going to be weak all year, that may not matter here. Would be a bit of a headache to go back and check the strength of schedules for all of those teams through 16 games.

How many teams on that list made the playoffs...made the Super Bowl...won the Super Bowl? Does a low TO/drive ratio make you a serious Super Bowl contender? I see some success on the list at a glance that would provide the Chiefs some hope. I would still see them as a longshot though, rather than a serious contender.

Agree with you that KC's turnover avoidance hasn't been correlated with scoring production on a game-by-game basis.

Thanks VERY much for bringing stats into these comment discussions. Extremely helpful and informative as everyone talks through the issues.

by Jordan Smith (not verified) :: Wed, 11/03/2010 - 5:07pm

Being an avid Chiefs watcher, I see where you're coming from Jeff. Clearly there are probably 6 better offenses in the NFL that are not named the Kansas City Chiefs (purely from an objective standpoint). However, when you knock on the efficiency of a Kansas City's ability to score points on things like 60+ yard drives and call their pass game inefficient there are some things you need to remember. First is that Todd Haley has massive brass balls. He will go for it on 4th down in or close to opponenents territory at least once a game. Things like this account for scoreless drives, because in most situations a team would do what most teams do: kick a field goal. This last Sunday Todd Hayley kicked (his most questionable call of the season) a punt from Buffalo's 33 instead of attempting a long field goal. This is where I think you're overlooking alot of the Chiefs ability to move the ball and be "good even against opponents who are not terrible" and saying "according to statistics, they're bad at scoring points." Secondly you also have to realize one of the most important statistics for the passing game to be efficient: interceptions. The passing game is efficient because they avoid interceptions, pure and simple. They avoid it against bad teams, they avoid it against good teams like the chargers. It's important to know when the pass game is efficient also. Matt Cassel sports a very nice 109.7 QB rating on 3rd and managable (3rd and between 3-7 yards) averaging 7.9 yards an attempt.

I'm not trying to create an argument for why the Chiefs have a top 6 offense. I am simply pointing out the reason why it may seem in pulling certain stastics for driving that KC may not appear either efficient or good at scoring points. You must realize the philosophy of the coaching style and team in order to garner a more accurate gauging. DVOA and other statistics are meant to supplement the evaluation process, not determine it. Take it with a grain of salt. Afterall, the bills only managed 10 points against Kansas City's defense after putting up 30+ on Baltimores (with Ed Reed playing like his usual self).

by Jeff Fogle :: Wed, 11/03/2010 - 5:38pm

Appreciate your comments Jordan.

*Let's say Haley has brass balls. I can see where that leads to a decrease in field goals. But, unless his gambles NEVER work, it should also lead to an increase in 7's. Some gambles win...some lose...3's disappear off the long drive slate, but 7's and 0's are there. There aren't many 7's when KC isn't playing Jax or Houston, and there are a zillion zero's. If KC can't drive the field for points now with brass-balled Haley as coach vs. everyone but Jax and Houston, how are they going to do it in the playoffs to a degree that would make them deserve serious consideration as a Super Bowl threat?

*I'm all for avoiding interceptions. The best teams drive the field for points while also avoiding interceptions. Avoiding interceptions by itself isn't enough to constitute an efficient passing game. It all still has to lead to the end zone in one form or another.

*Do you have a link for that Cassel stat? Would be interesting to see the number of samples...then, if possible, go back and figure out what part of that came agaisnt Jax and Houston, and what part came in the other five games. If the bulk came in the good games, it's data in favor of my point. If it came in the others, then we have to determine why he's able to post a good passer rating on third and manageable while the team as a whole has a very poor third down composite outside the good games. Didn't see the stat in the splits at nfl.com, so I'm not sure where you got it from.

Thanks again for joining the conversation. Helps to hear from somebody who watches the team a lot because it adds depth to what the drive charts and game stats are suggesting.

by DeltaWhiskey :: Thu, 11/04/2010 - 8:46am

"Let's say Haley has brass balls. I can see where that leads to a decrease in field goals. But, unless his gambles NEVER work, it should also lead to an increase in 7's. Some gambles win...some lose...3's disappear off the long drive slate, but 7's and 0's are there."

Not necessarily: Fourth and five from the 30, yields a 47 yd FG attempt. Brass Balls goes for it and is successful X% of the time. Field goals are succesful from this point X% of the time as well. So at this point it's a wash; however, more TDs are not the necessarily the logical follow-on, b/c Brass Balls may not go for it after the subsequent series moves them closer to field goal range where on 4th and 5, the success rate remains X%, but now the FG success rate is X%+Y%.

Eschewing a FG, does not guarantee a TD.

by Jeff Fogle :: Thu, 11/04/2010 - 12:42pm

No guarantees...and agree with the not necessarily theme because you can just get closer to a field goal. But, over a large sampling, I'm confident that the distribution would go from something like 3-3-3-0-0-3-3-3 (throwing in a couple of missed field goals on the field goal tries) to something like 0-0-7-7-0-3-0-7 or something.

Went to check Kansas City's actual 4th down plays this year to see what it looked like. NFL.com has them going 6 for 11 on the year on 4th downs. I counted 6 of 12, but there was a sequence at Houston where they converted on a pass interference call, then later scored a TD on a 4th down run. That's probably where the difference lies.

Of the 11 different drives that had 4th down attempts, four weren't really "brass balls" in terms of whether or not the drive was likely to end in points:

*The first came in the final seconds of the first half of the rainy opener vs. San Diego. It was 4th and 10 from the SD 39 in the rain. They didn't convert and SD took a knee to end the half. Nobody tries a 56-yard field goal on a sloppy field in the rain. Punting didn't make much sense with so little time left. Scoring was a longshot anyway in that situation.

*The second came late at Cleveland...4th and 1 at the Cleveland 36 where converting would clinch the victory because Cleveland was out of timeouts. KC converted, then took a knee twice to win. This was a drive to end the game rather than a drive to score points, so it wouldn't influence conclusions about KC's ability to drive the field for points in the big picture.

*The third came when KC was sitting on a 31-3 lead on SF in the fourth quarter. They went for it on 4th and 3 from the SF 31-yard line and didn't make it. It did turn a possible 3 or 7 into a zero, but it wasn't really a brass balls choice. I could see including it in the "the coach does things that keep them off the board too often" line of thinking though.

*The fourth was late in the Indy game, down 10 points, 4th and 12 from their own 40 yard line. EVERYBODY goes for it there and you're still a longshot to score anyway. KC actually did convert this, but ultimately missed a long field goal at a time where they had to rush "two scores down" to try and get something on the board.

The rest I think could reasonably be called "brass balls" choices:

*4th and 1 at the 50-yard line vs. SF. They converted off a direct snap to T. Jones. But, the Chiefs would throw an interception a few plays later, taking a 0 for the series. Of course, this was a go for it or punt choice. The coach didn't take a field goal off the board so to speak.

*4th and 2 from the Houston 17-yard line (pass interference), then 4th and 1 from the 2-yard line. KC would convert, and turn what would have been a 3 from the 17-yard line into a 7.

*4th and 1 from the Jacksonville 2-yard line. KC would convert and get a TD on this drive. Had they settled for the field goal, this would havebeen a 3 instead of a 7. Many coaches would go for it in that spot though I guess. Maybe not brass balls, but stronger than aluminum.

*4th and 3 from the Jacksonville 24-yard line, failed to convert...turning a probable 3 into a 0.

*4th and 2 from the Buffalo 19-yard line, failed to convert...turning a probable 3 into a 0.

*4th and 1 from the Buffalo 30-yad line. Did convert, ultimately kicked a field goal.

In scoring position (the last five examples above), what could have been
3-3-3-3-3, became
15 points one way, 17 the other

If you want to throw in the failed conversion with the 31-3 lead vs. SF, it's
3-3-3-3-3-3 if all the FG's are good, becoming
18 points one way, 17 the other

I don't think this would influence analysis of their long distance scoring performance. Though, the two TD's did come against Houston and Jax. Could see that a different distribution of when the successes happened might make things seem less dismal in the games not involving Houston and Jax. We're just not talking about significant volume. KC doesn't become a top six offense at driving the field if they take the FG's every time...or even if they convert on every brass balled gamble.

Getting kind of talked out on this theme, but appreciate everyone who made comments one way or the other. That's how science works. Would be fun to see the Chiefs pick up their pace offensively and become a true threat. Surprise teams are fun to root for. The KC/Oakland game could launch an ultimate surprise that makes either Raiderjoe or DVOA happy. I don't think the evidence suggests to this point that KC is ALREADY one of the best offenses in the league, or should rank second in Super Bowl victory likelihood at this stage of the season.

by Kisersosay (not verified) :: Wed, 11/03/2010 - 9:34pm

Ok Jeff point of the offense struggling against good defenses is correct. I think they have proved that. But a good team wins with offense defense and special teams while playing mistake free football. That is how KC is winning. You talked about how KC's offense struggled against Buffalo's poor defense but you didn't say anything about holding a team that put 30+ on the Ravens to only three points into the fourth quarter and 10for the game. Or holding Peyton to no touchdowns losing on a late fourth quarter touchdown run. Don 't be so one sided if you say they will struggle in they playoffs. Bring the whole picture. I heard somewhere that defenses win championships. You don't have to score a lot of points. You just have to score more than your opponents.

by Jeff Fogle :: Wed, 11/03/2010 - 9:59pm

Kiser, I'm specifically talking about Aaron's comments on the KC offense. Why would I bring up the defense when refuting his comments about the offense? I said the offense is likely to struggle in the playoffs.

In terms of the Super Bowl, could you provide a list of recent AFC Super Bowl entrants who survived the playoffs despite an inability to drive the field vs. defenses who were better than horrible?

by Xian :: Tue, 11/02/2010 - 8:35pm

Unsure if the Guts & Stomps article is actually on the web, but see http://www.footballoutsiders.com/info/glossary and read the "Guts" definition.

I think this is what you're getting at, and prior research has indicated the opposite of what you are suggsting. (I think.)

by AudacityOfHoops :: Tue, 11/02/2010 - 8:50pm

I wish people would stop citing that article (which is here, by the way. I don't think it means what people think it means. They barely touch on the issue of opportunities and skip home field advantage entirely (i.e. it might just be that playing bad teams leads to more wins, which leads to home field advantage, which leads to wins.). Plus, they use a 0.500 record to define who is good and bad - which is defensible, but I would have rather seen DVOA, SRS (from PFR), scoring margin, etc.

Just about the only thing I think I can take away from it is that NOT stomping bad teams is not a good sign.

by RichC (not verified) :: Wed, 11/03/2010 - 9:36am


That article doesn't prove anything, and it goes strongly in my "Bad stats are worse than no stats" category, because people use it to justify things that it doesn't by any means prove.

by Kisersosay (not verified) :: Wed, 11/03/2010 - 8:52pm

Funny. You can't just pull passing stats and make a case for them being a bad offense. Sure they are certainly not dynamic in the passing game but they are in the running game. Being elficient and not turning the ball over makes for a solid approach to win games I have watched several of their games this year and I am amazed at how many times they have 3rd and 5 or more and they run it. And they make a first down most of the time. Your passing stats are going to struggle with that approach but your wins are what count in both their losses they were in position to win in the fourth quarter on the road against very good teams but could not quite pull it off. I am amazed at the turn around. They are young and improving. Good enough to make the Super Bowl? Not yet.....but keep watching you never know.

by AudacityOfHoops :: Wed, 11/03/2010 - 9:43pm

You can't just pull passing stats and make a case for them being a bad offense.

A) He didn't. The first section deals with the whole offense, not just passing stats.
B) The passing stats are used in a section where he argues that they are not that great at passing. He pulls the passing stats to make a case for them being a bad passing team. Seems appropriate.

I have watched several of their games this year and I am amazed at how many times they have 3rd and 5 or more and they run it. And they make a first down most of the time.

He lists 4 games where they are a combined 11 of 51 on 3rd down (22%). Overall, they have a 36% 3rd down conversion rate - 24th out of 32 teams. Now, this doesn't directly refute your statement - maybe all the failures were passing attempts. But considering that they only have a 61% success rate when they run on 3rd/4th & 2 or less (see here), I highly doubt their success on 3rd & 5+ is better.

Good enough to make the Super Bowl? Not yet.

Looks like you and the first poster agree. And I think I'm on the same page, here. I mean, they COULD, if they get lucky, or get great playoff matchups. But I don't have them anywhere near the favorite. Then again, I am usually pessimistic about teams I root for.

by Jeff Fogle :: Wed, 11/03/2010 - 9:48pm

"You can't just pull passing stats and make a case for them being a bad offense."

Could you show where somebody did that? Don't think anybody's said KC has a bad offense. The dispute is whether or not they're one of the best six in the league. To the degree passing stats were "pulled" it was to dispute the assertion from Aaron that "when they do pass, they've been effective." Passing was effective vs. the horrible defenses of Jax and Houston. It's hard to make the case that passing has been meaningfully effective in terms of driving the field for points in the other games.

"Sure they are certainly not dynamic in the passing game"

Does this mean you agree with me that Aaron has overstated the success of the passing game thus far?

"I am amazed at how many times they have 3rd and 5 or more and they run it. And they make a first down most of the time."

Kansas City is currently tied for 23rd place in the NFL in third down conversions with 36%, while playing one of the weakest schedules in the league. They were particularly bad on third downs in a few of the games listed earlier. Would have disagree with "most" of the time unless you can show that they have a great percentage when running on 3rd and five or more, but incredibly horrible in all other situations. That's what it would take to fall to 36% for the year.

"Your passing stats are going to struggle with that approach"

Aaron suggested that their passing game wasn't struggling.

"they were in position to win in the fourth quarter on the road against very good teams but could not quite pull it off."

They couldnt' pull it off at Indy because they only had three field goals on the day...and in the fourth quarter, had two three and outs as Indy was building their lead to double digits. Indy's defense ranks a shade below average right now in DVOA.

Agree with you that avoiding turnovers is a good thing...that the team is young and improving...and worth watching...but not yet good enough to reach the Super Bowl in the AFC...

by AudacityOfHoops :: Wed, 11/03/2010 - 9:49pm

Jinx, you owe me a Coke.

by Jeff Fogle :: Wed, 11/03/2010 - 9:52pm

I edited in a PS about the simultaneous post...but you responding to my post took away edit mode! You win, I definitely owe you a coke...

by nat :: Tue, 11/02/2010 - 7:20pm

They may be the most balanced team in the NFL...
No, that would be the Redskins, at 23 on offense, 22 on defense, and 19 on special teams.

And a 4-4 record.

by Bobman :: Wed, 11/03/2010 - 1:14pm

Well said. What's their expected wins? Without checking, I'll assume something like 3.8-3.9 since their rankings are a little below average.

by Rick M. (not verified) :: Tue, 11/02/2010 - 7:45pm

Ok, that does it.

It took all of 8 weeks, but Philadelphia again inexplicably creeps up into the top 5 DVOA yet AGAIN.

I have to think that this whole DVOA system is bunk and is somehow biased in favor of the Eagles franchise. It doesn't make sense. Philadelphia is a terrible team with a slightly above-average offense and a horribly porous defense. They were blown out by the Titans, blown out by the Packers, and dominated by the Redskins. How in holy heck can they possibly be #2 in DVOA?

I'm seriously beginning to question this whole system after inexplicably seeing the awful Eagles teams of the past 5+ years consistently finish in the top five of the DVOA rankings.

by chemical burn :: Tue, 11/02/2010 - 8:22pm

How do you figure leading in the 4th quarter versus the Titans and losing by a touchdown that came down to a 4th and 1 in GB territory to be blowouts? They did get dominated by the Redskins... for 1 quarter and then for the rest of the game they played slightly better than the Redskins (not enough to overcome the early domination, but objectively better than their opponent.) Also, the Eagles have made it deep into the playoffs and won the division several times in the last few years, so I'm not sure what counts as "awful."

I agree this is the worst Eagles defense in a long time though and they have an unbelievably propensity for late game meltdowns...

by Tracing plan (not verified) :: Tue, 11/02/2010 - 8:33pm

Agree with chemical burn but lets add on the Redskins: they out gained by 60 yards, out first downed by 5, out possessed by 5 minutes, and had both hands on the ball in the endzone for the win on the last play...and if they hadn't screwed up before half and gotten a TD instead of FG, they would have been kicking to win rather than throwing up a pass for the win. I'm a Redskins fan but in no way shape or form could you say the Redskins dominated.

by Snack Flag (not verified) :: Tue, 11/02/2010 - 8:39pm

My 2 cents for you:

DVOA loves offenses that either pass a lot (hence Philly, SD, NE, IND being top offenses) or have a ridiculously dominant running game (KC or NYG). The great running games succeed a lot on first down (think about Jamaal Charles collecting 10 yards at will on Sunday) and the offenses that pass a lot just pick up much more yardage (and "succeed") by virtue of the fact that they pass a lot.

On defense, I think DVOA doesn't factor in big plays enough. It chalks them up as "fluky" if I understand the system correctly. So a day that Kenny Britt had against Philly doesn't seem repeatable, and it doesn't accurately value that. But it dislikes a team like NE that will let its opponents collect a lot of yardage and first downs, but not touchdowns.

Since Philly passes a lot, DVOA likes the offense. I'm not 100% sure on this but, from my subjective view of the Eagles, I would imagine that the Philly defense gives up a bunch of big plays, but minimizes the first downs, so DVOA doesn't think the defense is that great of an issue.

I'm 100% positive that Aaron and the FO guys did not set up an intentional bias for the Eagles in DVOA. I believe that Andy Reid's scheme and team identity are exactly what DVOA values, so each year the Eagles are up near the top even though they may not seem to be that good.

by chemical burn :: Tue, 11/02/2010 - 8:50pm

Not sure how it has been for the past two years, but the Eagles defense actually worked in the opposite fashion under Jim Johnson: they gave up a ton of small plays, ranked among teams with the fewest long plays and were one of the only I think 2 or 3 teams that FO's researched showed could actually be labelled bend but don't break defenses year after year.

From my subjective view, I would think that the Eagle's DVOA is probably a little out of whack from what happened vs. the Titans. the Eagles really did dominate that game for 3 quarters, but settled for a bunch of short field goals instead of touchdowns, so the score was reasonably close. Then they lost the game on essentially 4 plays: a fumble from the Tenn 4 yard line and 3 huge passes the Britt. The Titans basically had a handful of good plays that went REALLY well for them and one of them involved fumble luck. I think DVOA sees this as a game that says essentially good things about the Eagles... but the reality is that the Eagles have a huge problem in their #2 CB and their young safety CANNOT be relied on the shore up those problems.

For sure, there's no way to feel good about the Eagles going into this weekend's game: Peyton Manning + glaring secondary hole + team the generates pressure through blitzes does not equal a win. ever.

by Snack Flag (not verified) :: Tue, 11/02/2010 - 9:12pm

Ah, well, there goes my theory about the Eagles and DVOA. Like I said, I had no proof really to back it up just subjective opinion.

I've just always believed that DVOA doesn't properly weigh big plays and punishes teams or players for having big play ability, as if it's better to have 8 10 yard gains rather than 1 80 yard play. That's akin to saying 4 singles is better than a home run.

Maybe I don't understand DVOA though. That's possible.

by AudacityOfHoops :: Tue, 11/02/2010 - 9:19pm

It's a per play metric, so after you have your 80 yard play (which is subject to some kind of diminishing returns, so it's not THAT much more valuable than a 40 yard play), what you do on your next 7 (next drive) matters. If you throw 7 straight incompletions, then yes, DVOA will like the 8 10-yard plays better. Which makes sense - a bunch of 10 yard plays translates into 7 points per drive, while one 80 yard play every 8 downs translates into something like 3 points per drive.

By the way, 4 singles IS better than a home run. :)

by chemical burn :: Tue, 11/02/2010 - 9:29pm

That all said, there really seems to be something happening this year where late in games, the opposing offense realizes that the Eagles have real issues in the secondary and they can just keep chucking the ball up. Detroit, SF and Tenn all did it - at a certain point, they just said, "Oh, these guys have real coverage issues on deep passes, we can just keep throwing it long and they can't stop it."

by Pat (filler) (not verified) :: Tue, 11/02/2010 - 11:04pm

No, with Detroit/SF it was really the linebackers that were the issue: Detroit hit short passes and linebackers were out of position, SF hit the hole between the MLB and the safety. The Detroit late TD was coverage, but it was set up by a ton of shorter passes.

The Titans game was really the first one where Hobbs seriously looked like a liability.

by chemical burn :: Tue, 11/02/2010 - 11:09pm

Well, I was going to say "out of position safeties," but just decided to keep it down to "secondary issues."

by Pat (filler) (not verified) :: Tue, 11/02/2010 - 11:12pm

Out of position safeties, linebackers, and corners. But not all at the same time. For that, we'll have to wait for the Indianapolis game.

by chemical burn :: Tue, 11/02/2010 - 11:19pm

Well, the whole problem for Mikell has been that Sims and Jordan were way out of position, so he was compensating and trying to do too much (like last year.) Allen looked much worse than he actually played against Tenn & GB because Hobbs played so flat in bracket coverage and just couldn't do anything, so Allen had to make up for the terrible coverage underneath. I actually think Mikell and Allen are really solid and that even Fokou and Sims are finding their footing. The problem really still is the #2 CB... and the Eagles agree since Hobbs got benched and apparently might even be cut this week. For what it is worth, Hobbs wasn't really out of position, per se, just two steps too slow, playing underneath coverage insanely shallow and completely lacking in ball skills. Allen, Mikell, Sims and Jordan have just been in entirely the wrong place at times - Sims in particular has had me saying "Where the hell is HE going?" I think he was initially given some freedom to freelance and that's getting reigned in...

by Karl Cuba :: Wed, 11/03/2010 - 12:59am

Haven't linebackers in coverage been a weakspot for the Eagles for a while? They just don't invest in them.

by Whatev :: Tue, 11/02/2010 - 9:43pm

The problem is that even that is contextual; if you've got a lineup with one guy who singles 100% of the time and 8 other batters who can't hit for beans, you never score. If you've got a lineup with one guy who hits home runs 25% of the time and gets out the other 75% with the 8 other batters who never get on base, you average a little less than a run a game.

by AudacityOfHoops :: Tue, 11/02/2010 - 9:52pm

True, but I thought we were talking about stringing together plays/hits. (A drive of eight 10-yard plays vs. one 80-yard play).

Anyway, you're right. Nothing is that simple.

by chemical burn :: Tue, 11/02/2010 - 9:54pm

Why is everyone talking about baseball here? It's simple for football: if you consistently have a 30 yard pass followed 3 plays that gain no yards, you will have the big play then punt (or kick a field goal - or even score a TD depending on field position.)

If you can make 3 plays that gain 10 yards followed by 1 play that gains no yards, you will eventually score a touchdown. Both sequences of 4 plays gain 30 yards. One is clearly a better long-term strategy to pursue.

by Whatev :: Wed, 11/03/2010 - 4:14am

The point you seem to have missed here is that which is better depends on where you set your mean. You've picked a context where you succeed 3 times out of 4 and that gives you the ability to move the chains cconsistently. But increase the quality of the opposing defense, say, to the point where your average pickup is only 3-4 yards, say, they force you to check down a lot. Then your "consistent" offense is going nowhere. You might move the chains once or twice, but you'll be forced to punt before you get into the red zone. But a big play offense will occasionally pull off two big plays in a drive and grab a touchdown.

by chemical burn :: Wed, 11/03/2010 - 10:48am

Yeah, but moving the ball 4 yards on 3 out of 4 plays will still result in first downs and touchdowns. If we take the 30 yarder followed by 3 zero yard plays and similarly "adjust for defense" it looks even worse. Then you get a, say, 20 yard pass followed by 3 negative plays. That won't even get you in field goal range.

Also, please note that this is not an either/or proposition. It is better to be what the Eagles ACTUALLY are: a team that gets consistent chunks of positive yardage and is also near the top of the league in number of big plays.

Also, as much as everyone sees having to make more plays as a greater opportunity for failure, for some reason no one is acknowledging that having more plays means a greater opportunity to break big ones (e.g. more opportunities for the Rb to break a tackle and take it to the house, more opportunities for a CB to slip and let a WR get into wide open space, more opportunities for a defense to blow an assignment, more opportunities for the offense to take a risk on a deep shot downfield.)

by Whatev :: Wed, 11/03/2010 - 5:12pm

Moving the ball 4 yards on 3 out of 4 plays will earn you a first down LESS THAN HALF of the time, assuming you punt on 4th down, because you can NEVER have a failed play. If you need to drive 70 yards for a touchdown, you will average a touchdown on less than ONE PERCENT of your drives. Even if you go for it all four downs, you'll still turn the ball over on downs about 25% of the time.

by chemical burn :: Wed, 11/03/2010 - 5:26pm

Well, this is all too theoretical to apply that kind of logic to it: if you could be guaranteed 4 yards on 3 out of 4 plays, you would obviously never punt. And that made-up number of 4 yard on 3 out of 4 is conspicuously (and arbitrarily) below 5 yards on 3 out of 4 which would be the threshold for TD certainty (assuming a punt on 4th down.) Compare the threshold for TD certainty in the "1 big play, 3 zero yard plays" scenario: a single 100 yard play every 4 downs.

Also, how does the TD % of 4 yards on 3 out of 4 that compare to the 20 yard play plus 3 negative plays scenario? It still seems like the better option.

by Whatev :: Wed, 11/03/2010 - 10:18pm

The answer is that A) both of them suck, and B) depends on exactly how many yards you have to go. Also, you're wrong about both "arbitrarily" (because 4 yards is about what you'd get from an average running play in an average year) and "certainty" (because even if you got exactly 5 yards per successful play, you'd still have to succeed twice in three tries per set of downs, something that is far from guaranteed). Given the same EV per play and ordinary field position, and assuming that the EV is large enough for a first down on average, the consistent offense is better simply because 70-80 yards is a long way to go; but that's not a fair comparison either because yards are so much easier to gain once you hit the second level.

On the practical level, as has been mentioned, touchdown drives very often feature at least one gain of 10 yards or more, because if you keep running plays, somebody will make a mistake. If that's the defense, there will be a big gain. If it's not, the offense will probably make one before they go the whole length of the field.

by chemical burn :: Wed, 11/03/2010 - 11:01pm

Listen, man, you're arguing nuanced realities against a pure theoretical. Of course they both suck and neither would ever happen in reality. But the point is that consistent gains are more effective for scoring than erratic gains even if the consistent gains are smaller. Do you deny that?

Of course, an increase in plays seems like it would mean an increase for mistakes, but no one presented any evidence that those mistakes favor defense or offense (or that any such increase in "mistakes" even actually happens on longer drives.)

And, yeah, TD drives generally feature gains longer than 10 yards because good offenses can generally generate plays longer than 10 yards. Of course. And DVOA "likes" those longer gains, too. But that's not the question, the question is: is erratic play more beneficial than consistency in terms of generating points? The answer is obvious. The only way it can even be disputed is if you posit the consistent gains as very small and the erratic gains as very big.

If you're going to dispute any of this based "on reality" you need to start presenting some facts from reality. Do teams that generate big plays but few small plays score more frequently than teams that rely on consistent smaller gains? Do teams that sustain longer drives commit more mistakes (e.g. sacks, interceptions, penalties, fumbles)? Do defenses give up bigger plays later in long drives (i.e. they "eventually" make mistakes.) And what are our definitions for big plays, consistency, mistakes, small plays, negative plays?

by zlionsfan :: Thu, 11/04/2010 - 12:39am

Well, if your SLG is around .143, you're screwed no matter how you look at it. In general, you want to be getting on base, and breaking up that HR into singles turns out to be better for the offense.

It's not quite the same idea, though ... the idea here is that you can't draw up a play to gain 80 yards, mostly because you only really have control of the first 20 or so. Beyond that, it's not quite reproducible ... you might have a player who's fast enough to outrun everyone, strong enough to break tackles 50 yards downfield, or maybe even both, but there's no way to plan how that will work. That's particularly true of runs, of course, given that again you're only working out the first 15-20 yards and then leaving the rest to open-field running (hopefully).

Passing, sure, you can plan for a bit more field, but there's still a point past which it becomes "just get past anyone who's left".

I think that's one of the things about DVOA that draws criticism: some people (not necessarily you or people above you) seem to feel that DVOA ought to account for that sort of thing somehow ... but I don't think it can. Diminishing returns on a play aren't really that different than diminishing returns in a game, examples of which are usually better found in NCAA play.

In general, a 20-point win is better than a 3-point win, but a 40-point win might not be that much better than a 23-point win, and a 60-point win might not be any better than a 43-point win. At some point, the difference in scores has very little to do with the difference between teams; it's only about the way the leading team plays the last part of the game. One team continuing to attack at 100% and another team having its second- and third-string offense run off-tackle plays, those aren't things that a system can predict or evaluate well, so at some point, I might explain it as "this isn't as important" when what I might really mean is "I have no way of telling whether this would happen again or not, so I have to mute its effects so that it doesn't unbalance some of my other conclusions."

(As a side note, I think the biggest issue people have with DVOA is that it misses sometimes. Well, duh. Shouldn't it? If it only took a few years to figure out how to accurately evaluate teams, it wouldn't be much of a field worth studying, right?)

by B :: Tue, 11/02/2010 - 9:27pm

This is a common misconception. It's not that DVOA ranks big plays over small plays, but it ranks all plays. So the question is what did the team that got the 80 yard play do in the other 7 plays? If they didn't gain anything, that would be 2 punts, giving the opposing team two more chances to score. Just like 4 singles vs a home run and three strikeouts, that's three lost chances for the batter's team.
As for big plays, a big play is worth more than a small play, but there's a cap to the value. An 80 yard touchdown isn't worth that much more than a 40 or 50 yard TD. Once the receiver gets past the last line of defense, the only difference between the two plays is where the LOS was.

by Tracing plan (not verified) :: Tue, 11/02/2010 - 10:49pm

I feel that way too--it seems that most scoring drives have a big play.

But the question at hand (for designing DVOA) is what correlates best with future success. If you think about it, many big plays involve mistakes--blown coverage, somebody slips. Offenses that are consistently good, blocking well, keeping the qb upright, catching the ball, and gaining yards in small chunks consistently, are more likely to benefit from those mistakes.

Or look at it another way. The best teams of the last 10 years, the Pats and Colts in the AFC and the Eagles in the NFC, both efficiently gain yards in small chunks, and both do well in DVOA. The outliers are teams like the Steelers and Ravens, and they have defenses which are difficult to run against and tougher than most against the short passing game.

The outlier is the Rams, who ate up yards in huge chunks.

So maybe the best way to be very very good in the NFL is to consistently gain small chunks of yards.

by JMM* (not verified) :: Wed, 11/03/2010 - 7:04am

"So maybe the best way to be very very good in the NFL is to consistently gain small chunks of yards"

I think that is what DVOA is saying.

Yet Dick LeBeau's philosophy is to take away the big play and make them drive the length of the field. He also takes away the run first. These tenets should lead to failure by forcing teams to go with short passes. Yet his track record is one of success. There is a disconnect somewhere.

by Athelas :: Wed, 11/03/2010 - 7:31am

Belichick feels the same way, but I don't think there's a disconnect.

It's very hard to do that, so if you can, you're very successful.

by Andrew Potter :: Wed, 11/03/2010 - 7:48am

I don't see that as a disconnect; I'd say it backs up DVOA.

Dick LeBeau and others try to take away the big play and force teams to drive the length of the field precisely because that's so difficult to do. If a team is able to do that consistently and effectively, that is a hallmark of a good to very good offense. Any NFL offense can execute on one play; only the good ones will consistently execute on 10-15 play drives, hence the defensive philosophy.

To put it another way: Defensive co-ordinators don't try to get you to do easy things; they try to force you to do what is difficult. Yes, completing one short pass is usually easier than completing a bomb, however consistently completing short passes successfully all the way down the field is more difficult than pulling off one big play. Sustained success in that situation is predictive and a sign of a good team.

by JMM* (not verified) :: Wed, 11/03/2010 - 8:32am

Athelas, Big-Hairy Andy,

Thanks. Between your comments and this coffee, it all seems to make sense.

by BJR :: Wed, 11/03/2010 - 10:41am

Yes, and of course the great offences have the ability to mix in the occasional big play alongside consistently good short play execution, because of the skill of the players in taking advantage when a defensive mistake does occur.

The NO/PIT game from the weekend was an excellent example. PIT took the run away, as you would expect them to, and forced the NO passing game to execute flawlessly. In the second half it did. And when there eventually was a mistake by the PIT defence (a blown coverage downfield), Brees was good enough to spot it and complete the pass for a huge gain.

by DeltaWhiskey :: Wed, 11/03/2010 - 10:10am

Every play is an opportunity for error. So if you make a team gather small chunks over and over and over, there is a better probability for making a significant error mistake. Watching the Steeler Defense can be frustrating, as they give up 3-4 and 5 yard chunks, and then all of a sudden they blow up a play for a sack and/or turnover.

by chemical burn :: Wed, 11/03/2010 - 10:59am

But again, every play is an opportunity for DEFENSIVE error as well - and a small five yard run with a missed tackle can suddenly turn into a 25 yard gain.

And for the record, FO's research showed that only the Eagles under Jim Johnson and the Pats under Belichek actually qualified for the "bend don't break" status everyone is attributing to LeBeau right now. I gotta say, I find the consensus idea that the Steelers have some elite defense this year kinda strange. Sure, their run defense is top notch, but their pass defense is pretty pedestrian and certainly not capable of shutting teams down. Their DVOA is similar to the Eagles pass def DVOA and I would describe the Eagles pass defense as "a problem."

by tuluse :: Wed, 11/03/2010 - 5:13pm

I thought Herm Edwards had a record of break but don't' bend defenses.

by chemical burn :: Wed, 11/03/2010 - 5:18pm

Maybe you are right - it's in one of the Almanacs and I'll have to dig it up. I know Johnson's Eagles and Belichek's Pats for sure. I really think there was 1 other team at most (maybe Herm)...

by Shattenjager :: Wed, 11/03/2010 - 8:00pm

I think you need to re-read tuluse's post.

by chemical burn :: Wed, 11/03/2010 - 8:33pm

ha. It's my irrational Herm love blinding me once again...

by Alternator :: Tue, 11/02/2010 - 8:44pm

Want the simplest, easiest possible solution?

Something that the Eagles do poorly, but which is not predictive league-wide, IS predictive for the Eagles. Without making a specific "screw the Eagles" adjustment, there's nothing to be done about it.

Also worth noting is that the Eagles really do tend to perform well in the playoffs compared to their record, so DVOA might not be entirely wrong.

by chemical burn :: Tue, 11/02/2010 - 8:54pm

There's also the fact that the Eagles have won the division and made the playoffs all but twice in the past ten years despite playing in one of the hardest divisions in football according to DVOA. There are only 3 teams with more wins since 2001: the Colts, NE and the Steelers - how come no ever accuses DVOA of over-rating those teams?

by Whatev :: Wed, 11/03/2010 - 4:17am

One of the things that FO has stated is not predictive is blocked kicks, for example. And this appears to generally be true--except this year, the Chargers apparently have such abominable special teams that it IS predictive and you can count on them to get kicks blocked every other game or so.

by jebmak :: Wed, 11/03/2010 - 10:22pm

I believe that blocking kicks is not predictive, but missing the field goal (due to it being blocked) or having the tipped punt go 10 yards is.

by Andrew Potter :: Wed, 11/03/2010 - 11:56pm

Same for missed field goals, I believe. Missing the field goal is predictive. Having it missed against you isn't.

by Pat (filler) (not verified) :: Tue, 11/02/2010 - 10:45pm

I'm seriously beginning to question this whole system after inexplicably seeing the awful Eagles teams of the past 5+ years consistently finish in the top five of the DVOA rankings.

Yeah, those awful Eagles teams. How dare they finish in the top 5 in each of the years when they make the playoffs!

Jeez, if you're starting to question DVOA, you must really be questioning the NFL's playoffs, too. (Also, considering only 3 times out of the past 5 years the Eagles finished in the top 5, you might want to reconsider those cutoffs).

Seriously, the Eagles are the 5th ranked team in net yards per drive, and the 3rd ranked team in net turnovers/drive. Their "horribly porous" defense currently ranks 5th in turnovers per drive and 5th in the league in adjusted sack rate.

Do I think they're a great team? Hell no. But no one else is this year, either. I'd say the entire top 10 are completely identical, except San Diego's special teams are horrendous, and Indy/GB/NYJ have had injuries to some of their best players.

by chemical burn :: Tue, 11/02/2010 - 8:25pm

Funny how in the early going of the season, everyone was talking about how the Eagles have had a somewhat easy schedule and how after the bye, it would get a lot harder. Now the their pre-bye schedule ranks 10 and their post-bye schedule ranks 15.

by Kevin from Philly :: Wed, 11/03/2010 - 11:25am

Well, I'm guessing the Cowboys meltdown, along with the poor performances of Chicago and Minny, explains some of the weakness of the second half of their schedule.

by chemical burn :: Wed, 11/03/2010 - 11:37am

Of course. But even I myself remember thinking, "Man, they really need to be 5-2 or 6-1 going into the bye week if they're going to have any shot at making the playoffs." Suddenly, a schedule that goes the Cowboys twice, Minnesota, Washington and Chicago doesn't look harder than one that featured Atlanta, Tennessee and Green Bay. Of course, the Eagles have a knack for losing games when you least expect it and winning ones that seem hopeless, so they'll probably beat Indy and then lose to Washington again.

by Independent George :: Wed, 11/03/2010 - 12:00pm

I love how the entire NFC East fan base on these boards are absolutely convinced that their teams are always on the verge of complete and utter destruction. We maintain absolute certainty of our eventual humiliation and ridicule. This is not to say that we are irrational in our pessimism, but it still amuses me to no end.

by dmb :: Wed, 11/03/2010 - 12:25pm

I don't think that's entirely true. A few of the Cowboys' regulars were pretty adamant that the preseason projection for Dallas was way too pessimistic. IIRC, their case partially based on FO's seemingly annual habit on missing low on the Cowboys' projection, and an expectation that Dallas would maintain its abnormally low injury rates (because of a special training staff?). I think there was a third part to the argument -- perhaps the regulars believed that the offensive line wouldn't struggle as much as FO predicted? -- but I can't recall exactly what it was. Anyway, there were several quite outspoken Dallas fans who were decidedly not in any sort of "doom and gloom" mood.

As a Redskins' fan, I probably tend to be a little on the pessimistic side, but that's probably because I became a fan after the glory days of Gibbs I. When a couple of 10-6 seasons are the highlight of one's rooting experiences, it's probably better to not be swayed easily into optimism.

by chemical burn :: Wed, 11/03/2010 - 1:43pm

This is definitely true. I've been shocked by how down Giants fans seem to be on their team. To me, they seem like one of the safest bets for Superbowl winner this year. And the Redskins are at least winning games this year - they could easily double their win total from last year.

As for my Eagles? After this decade is there any reason to get my hopes until they actually hoist the Lombardi? No. No there is not.

by Dean :: Wed, 11/03/2010 - 3:41pm

I don't recall who it was, but I definitely remember debating back and forth in the offseason with a Redskins fan who was convinced that they were poised for greatness. Maybe that's an exageration. I know I was extremely down on them - but then again I usually am. For several years I've been saying the same thing - that the whole is less than the sum of the parts because Danny-boy is interested in individuals, not a team. This year I went a step further and said that aside from McNabb, they had the least talented roster in the entire league and that if he had a serious injury, they'd be on the clock. So maybe it was less that he was optimistic than it was that he thought that I was overly pessimistic.

by androlebow44 :: Tue, 11/02/2010 - 8:45pm

The Giants fell in DVOA and the Eagles rose in DVOA despite bye weeks in large part because of the Cowboys Jaguars game. The Eagles had already played the Jags and the Giants had already played the Cowboys. The Jaguars blew out the Cowboys making the Giants schedule look easier and the Eagles schedule look harder.

But the Jaguars got to play the Cowboys without Romo and with their season essentially over, making the Cowboys less hungry and possibly hoping for a top draft pick. On the other hand, the Giants played the Cowboys with Romo for a half and the stakes were higher because a win in that game could've potentially salvaged their season.

I'm aware that this probably didn't effect either the Eagles' or the Giants' DVOA THAT much, but it does bring up a flaw of DVOA. DVOA doesn't measure how strong an opponent was during the week a team played the opponent. The Falcons played the Steelers without Roethlisberger while many other teams will play the Steelers once or even twice with Big Ben. But DVOA at the end of the season will give equal credit to the Falcons and teams like the Dolphins, Saints, and Browns for facing Pittsburgh despite the fact the Steelers' opponents faced different teams.

It's tough to find a solution except for maybe factoring a team's DVOA leading up to the game as a factor in measuring DVOA. But even that wouldn't solve the Giants-Cowboys situation.

by Whatev :: Tue, 11/02/2010 - 9:46pm

Even if you did that you'd still be screwed because autocorrelation in your numbers combined with really low statistics for the start of the season would drive all your regressions nuts.

by cfn_ms :: Tue, 11/02/2010 - 9:53pm

One not so simple solution would be to have a big set of injury indicators: starting QB out, starting RB out, etc. Presuming this info is available (which I think it is, though people being out for only part of a game makes it messier), it's probably doable to solve for the value of each to a reasonably small error range. However, even that's inadequate, since some starting QB's are worth more than others, etc.

But that at least might be a reasonably starting approach, if they felt that it was worth the trouble.

by Jerry :: Wed, 11/03/2010 - 2:08am

In a perfect world, where FO had the information about who's on the field for every play, it might be possible to develop individual DVOA or DVOA shares or something, and be able to use those for opponent adjustments. It's not a perfect world, though.

by dmb :: Wed, 11/03/2010 - 2:13am

Weighted DVOA is a way to address the injury/suspension issue. It doesn't make much of a difference now, but at the end of the season, you'll know to pay more attention to the weighted figures for teams that had a significant early- or mid-season change.

by Still Alive (not verified) :: Tue, 11/02/2010 - 10:59pm

It has been obvious people and FO misuse the "rankings" all the time. You are clearly aware of this problem so why talk about them at all? Hell you could even not even list the numbers, or break the list into colors by tier. Anything to get people and yourselves thinking more correctly.

by Still Alive (not verified) :: Wed, 11/03/2010 - 12:41am

Just to be clear I am talking about the RANKings not the RATEings.

by Morton (not verified) :: Tue, 11/02/2010 - 11:17pm

So, basically, San Diego would be, far and away, #1 in DVOA if only they had even an *average* special teams unit?

Are they a historically good team being dragged down by historically bad special teams play? That's absolutely incredible and disgusting.

by chemical burn :: Tue, 11/02/2010 - 11:20pm

SD would be #1 if they had even typically BAD (like -3% DVOA) special teams...

by dmstorm22 :: Tue, 11/02/2010 - 11:27pm

Not historically good. They would have roughly 32% DVOA rating if ST were not included. That is quite good, but I'm pretty sure there are teams each year who beat that without ST factored in.

So far, I got the 2008 Eagles, 2007 Pats (obviously), 2007 Colts, 2005 Colts, 2005 Broncos, 2004 Pats, 2004 Steelers, 2001 Rams as all teams that are better than the Chargers, without ST included. So I wouldn't say historically good, but in the top-10-12 of the last 10 years.

by onetime91 :: Wed, 11/03/2010 - 12:46am

For the geniuses that somehow think that their spread sheets are biased for the Eagles, have you seen their Michael Vick analysis in the games he has played this year? They completely pooh-poohed his performances and ranked him below average.

by tuluse :: Wed, 11/03/2010 - 12:56am

He's ranked 8th in DVOA.

by chemical burn :: Wed, 11/03/2010 - 10:42am

He means their written analysis in the Quick Reads and DVOA weekly columns. In their defense, the bulk of the negativity came from one writer and Doug Farrar had many good things to say about Vick.

by DeltaWhiskey :: Wed, 11/03/2010 - 10:09am

Last week I clustered teams based on DVOA to try and simplify discussions, so here they are again. Categories are assigned based on the SD from the mean, so:

Elite +2SD
Good +1 SD
AVG w/in +/-1SD
Bad -1 SD
Horrid -2 SD

For total DVOA, Mean = 0.37%, SD =16.94%


For OFF DVOA, Mean = 1.60%, SD = 13.98%

For DEF DVOA, Mean = 1.58%, SD = 12.05%

by Independent George :: Wed, 11/03/2010 - 10:39am

That was really helpful. Thanks.

by AudacityOfHoops :: Wed, 11/03/2010 - 10:54am

As predicted, the AFC West looks extremely strong, as it contains the only two teams with GOOD/ELITE offense and defense: KC and SD.

Wait, what?

by jebmak :: Thu, 11/04/2010 - 8:20am


by Arkaein :: Wed, 11/03/2010 - 12:19pm

I have to say that I don't care for this partitioning. It seems like an exagerrated version of the problem of looking at teams rankings instead of their ratings.

The problem is that their are no bright lines between the clusters, and that the AVG cluster is so large (with twice the DVOA range of the other clusters). The idea that KC should be in a category above ATL while ATL belongs in the same category as SF is not enlightening, it's downright misleading. The difference between ATL and SF is larger than the difference between ATL and any of the teams ahead of them in DVOA.

I think any attempt to fit teams into easy to understand categories is going to fail to add clarity unless divisions are made at large gaps in DVOA. The first obvious gap now is between NYG (25.3%) and SD (17.0%), an 8.3% gap. After that the largest gaps are 3.8% between ATL and BAL, 3.7% between NO and HOU, 6.2% between HOU and CLE, 7.1% between STL and DEN, and 8.7% between BUF and CAR.

Maybe throw out all gaps below 5% and group teams into GOOD (NYG or better), DECENT (HOU or better), POOR (STL or better), BAD (BUF or better) and HORRID (ACAR and ARI). Of course this sort of clustering will work better some weeks than others, and would need new descriptors based on the results of the clustering. However I think it would be less misleading that using any predetermined group cutoffs.

by dmb :: Wed, 11/03/2010 - 12:41pm

Agreed -- when I mentally categorize teams by DVOA, I'm usually taking a look at the largest differences in DVOA, and the ranges between those "breaks." Earlier in the season, there were really pronounced tiers, but as expected, those are starting to smooth out a bit. As recently as Week 6, you could divide teams up as:

GOOD (~10% - 15%): IND, NO, ATL, BAL, GB, SD, SEA
BAD: (-24% - -11%): CHI, CLE, TB, JAC, STL

The "BAD" category was a bit spread out, but there were "bright-line" divisions between "ARIZONA" and "REALLY DREADFUL," "REALLY DREADFUL" and "BAD," "MEDIOCRE" and "GOOD," and "GOOD" and "TOP."


by DeltaWhiskey :: Wed, 11/03/2010 - 1:31pm

"The problem is that their are no bright lines between the clusters,"

I think the cut between NYG and SD is pretty clean, and for arguments sake, I'd be comfortable moving STL out of the BAD group into the AVG group. With the exception of HOU and CLE, in the average group there is no real difference from one team to the next that exceeds 5.0%. The cut from BAD to HORRID is pretty damn clear as well. I would argue that at least this week, SD had done a nice job of cutting the cake.

"and that the AVG cluster is so large (with twice the DVOA range of the other clusters)."

SD will do that to you.

"The difference between ATL and SF is larger than the difference between ATL and any of the teams ahead of them in DVOA"

Using my wildly unpopular DVOA converted to WINS descriptor from last week, SF is playing like a 7.4 (7) win team (avg), ATL is playing like a 9.6 (9-10) win team. If (big if) these descriptors were predictive, it suggests SF is a solid AVG team with little likelihood of making the playoffs, and ATL is at the upper end of average, with potential to make the playoffs (no, I haven't gone and figured out who would/would not make it due to AFC having more in the top, etc, but in general 9-10 wins lets you at least knock on the playoff door). The NYG are playing like an 11.4 (11) win team, which again, if an accurate prediction, suggests a pretty high likelihood of getting in the playoffs.

"I think any attempt to fit teams into easy to understand categories is going to fail to add clarity unless divisions are made at large gaps in DVOA."

You can use subjective criteria or objective statistically based criteria. I don't know if this breakdown means a damn thing, but I know it is based on reasonably sound statistical principles and from that standpoint I have more confidence saying that TEN is playing better than IND than if I say SD is playing better than NO.

by AudacityOfHoops :: Wed, 11/03/2010 - 1:56pm

"I don't know if this breakdown means a damn thing, but I know it is based on reasonably sound statistical principles and from that standpoint I have more confidence saying that TEN is playing better than IND than if I say SD is playing better than NO."

Well, you shouldn't, if you're basing it on the fact that your arbitrary division lies between TEN and IND, and not between SD and NO. Your divisions are saying that you should feel reasonably confident that TEN is a "better than average" team, not a "better than IND" team. To feel confident that a team is better than IND, you need to draw your +/- 1 or 2 SD boundaries centered around IND:

+2 SD: none
+1 SD: none

by DeltaWhiskey :: Wed, 11/03/2010 - 2:37pm

You're absolutely correct.

by DeltaWhiskey :: Wed, 11/03/2010 - 2:57pm

Sorry for the reply to myself, but to expand:

With a degree of confidence that is around 95%:

KC is better than SF and everyone down.
PHI is better than CIN and everyone down.
NE is better than WAS and everyone down.
TEN is better than WAS and everyone down.
PIT is better than WAS and everyone down.
NYG is better than WAS and everyone down.
SD is better than STL and everyone down.
IND is better than DEN and everyone down.
NYJ is better than DEN and everyone down.
GB is better than DEN and everyone down.
MIA is better than DEN and everyone down.
ATL is better than DEN and everyone down.
BAL is better than BUF and everyone down.
NO is better than CAR and everyone down.
HOU is better than CAR and everyone down.
CLE is better than ARI.
SF is better than CLE.

by AudacityOfHoops :: Wed, 11/03/2010 - 3:19pm

SF is better than CLE.

Subjective opinion? :)

by DeltaWhiskey :: Wed, 11/03/2010 - 4:42pm

Crap, got in a hurry, should be:

SF is better than ARI.

Thanks for reading through it all.

by tuluse :: Wed, 11/03/2010 - 5:07pm

BTW, you can edit one of your previous posts if no one has responded to it.

by DeltaWhiskey :: Thu, 11/04/2010 - 8:41am

Aware, but someone had responded, and besides I never miss an opportunity to be self-deprecating.

by Arkaein :: Wed, 11/03/2010 - 3:18pm

I should have gone back and revised my comment after I actually did my partitioning. You're right that there are some bright line divisions such as NYG/SD that appeared, I unfortunately wrote that out first, at the time looking more at your divisions based on Standard Deviation.

However, I still think my clustering approach has its merits. Standard Deviation simply lumps too many teams in the average category. Green Bay hosts Dallas this weekend and is going to be one of the heaviest favorites in the NFL (Vegas had them at -8.5 last I checked). The idea that these two teams are in essentially the same category is fairly absurd. If the AVG category was at least divided into HIGH AVG and LOW AVG around 0.0% DVOA so that all categories covered the same range it would be a little more realistic.

Also, the method I used for splitting teams into groups may not have a strong statistical grounding, but is quite similar to computational clustering methods in heavy use for many applications. A formal algorithm would go like this:

1. Place each team into it's own cluster

2. Find the clusters with the smallest separation (determined by minimum difference between any members in either cluster) and merge into a single cluster

3. Repeat until no more clusters can be merged (e.g., because no clusters are within a certain preset distance of each other, or a minimum number of clusters have been formed)

This method has its downsides, such as not working well with perfectly smooth distributions, but is still quite useful and does a better job of defining similar groups than simple standard deviations. Other variants can create one big cluster with subclusters, such that the largest gap divides the two main clusters, and then smaller gaps divide these clusters into sub-clusters, etc.

This Wikipedia link gives more info and few illustrations:


by DeltaWhiskey :: Wed, 11/03/2010 - 4:51pm

I agree that how you cluster things has merits, if you can tell me what the hell DVOA "means." By that I mean, if I use your system (or mine), what sort of confidence do I have that I can/should bet one way or another on the GB/DAL game. I can say that given that GB and DAL are almost 2 SD apart, the -8.5 may/may not make sense. SOunds like an opportunity to due a study, that I don't have time for tonight.

by tuluse :: Wed, 11/03/2010 - 5:33pm

This is cool and interesting. I'm not sure what it means, but I think it's a good start towards something.

Would it be appropriate to use some calculus and find the points of inflection and use those as the category boundaries?

Edit: I'm not sure inflection points is what I mean, it's been a long time since I've had a math class.

by Kisersosay (not verified) :: Wed, 11/03/2010 - 9:03pm

So this validates the article as KC is the only team above average on all three

by AudacityOfHoops :: Wed, 11/03/2010 - 9:30pm

So this validates the article as KC is the only team above average on all three

This word ... "validates" ... I do not think it means what you think it means. Try one of these: restates, parrots, "presents in a different format."

by jmaron :: Wed, 11/03/2010 - 11:19am

I like to look at some of the various rating systems to see how they differ. The three I look at are DVOA, Advanced Football Stats and Sagarin Pure Points.

Of interest:

DVOA has KC number 1st - Adv Foot - 6th, Sagarin - 13th

Philly DVOA - 2nd, Adv Foot - 10th, Sagarin 7th

SD - DVOA - 7th, AdvFoot - 1st, Sagarin - 11th

The only teams that have top 5 ratings by all three systems

Tenn, Pitt

by AudacityOfHoops :: Wed, 11/03/2010 - 11:40am

To throw in Simple Rating System (from PFR):

KC 9th
Phi 11th
SD 7th
Ten 1st
Pit 4th

by DeltaWhiskey :: Wed, 11/03/2010 - 11:22am


SD - their S.T. DVOA is over 4.5 SD's from the historical mean. Removing them from the above calculations changes the Mean to 0.90% and the SD to 4.18%. Now, NE suddenly has good S.T., unfortunately, for STL and DEN, they now have bad S.T.

Looking further at the whole special teams thing, and yesterday (or this morning?) I heard P. King talking about what he had to write for S/I at the mid-year, and he was going to (or did) write about S.T. being a surprisingly big factor so far. So, looking at the NYG, they are the only good team with bad S.T. and it's pretty obvious what SD's special teams have done to them. Also of interest are SEA, JAC, and CHI at the low end of the average range for TOTDVOA (they're not on the DVOA short bus, but their teachers are wondering if they should be) with another area of their game that is either bad or even horrid, and yet, all three have won 4 games. I haven't looked at the historical data (nor will I) to know whether there's something funny going on around here, but is S.T. DVOA getting proper credit in the calculation of TOTDOVA?

by johonny (not verified) :: Wed, 11/03/2010 - 11:40am

I can't believe the Dolphins are 17 in ST. I know their kicker is good but man their kick coverage and return teams suck, suck, suck.

by andrew :: Wed, 11/03/2010 - 11:49am

As underwhelming as the Vikings have been, I can take consolation that they do rank #1 in something (Variance).

Which means.... that while they may have been not good, they have at least been consistantly not good.

(actually the plus side is their schedule strength, which has their schedule thus far as one of the hardest, and their schedule going forward as one of the easiest).

by ammek :: Wed, 11/03/2010 - 1:46pm

What's with the rushing defense numbers this season? There are fully 24 teams with above average (ie, negative) run defense. The mean is -6.3%; the median is -7.3%.

Is it just an early season thing?

by chemical burn :: Wed, 11/03/2010 - 1:58pm

Maybe it has something to do with the league-wide increase in the use of the shotgun formation? It seems like runs out of the shotgun are very "boom... AND BUST. AND BUST. AND BUST."

Also, it seems like there are very few elite running backs producing at a high level this year. Highly touted guys like CJ2000, Ray Rice, MJD and the Cowboys backs are not really setting the world on fire. Stalwarts like Portis, Steven Jackson and Ronnie Brown/Ricky Williams have been underwhelming. It seems like the defenses haven't stepped up so much as the offenses have stepped down...

A lot of weird with limited carries names in DVOA/DYAR's Top 10 backs: LeSean McCoy, BenJArvis Green-Ellis, Arian Foster, KC's Charles, Tolbert, Peyton Hillis...

by ammek :: Wed, 11/03/2010 - 2:58pm

It's the same for WRs: Lloyd, Britt, Williams > Fitz, Steve Smith (Sr), Welker.

Offensive rush DVOA averages quite close to zero. The discrepancy is only on defense.

by Chip Paint (not verified) :: Fri, 11/05/2010 - 10:01pm

Apropos of nothing. In what other division but the mighty NFC West can a 2-6 team have a 15% chance of winning the division?