Extra Points
News and commentary from around the Web

FO Mailbag: DVOA Without Fourth Quarter

OK, it's time for a quiz.

Each of these teams sees its offensive DVOA drop in the fourth quarter: Green Bay, Houston, New England, and Pittsburgh. Can you put them in order from the team with the biggest drop to the team with the smallest drop?

While you work on that, here's the mailbag question from today's DVOA discussion thread.

Paul M.: If there was a "First Three Quarters" DVOA, I would think the Packers would dominate and begin to show they are more of a historically dominant team than they are currently credited for, but then again that would discount some teams (any one in particular that comes to mind??? Hmmmm..... maybe they play in the Mountain Time Zone??) and their ability to rally late.

Nat: Aaron, could you publish the rankings/numbers for "First Three Quarters" DVOA? ... In theory, this all-but-late DVOA should avoid the prevent-defense, garbage time, hail-mary, shut-the-offense down, play-the-backups issues -- while still being a large enough sample to characterize each team pretty well. Pretty please?

OK, so, first let's answer the quiz question above. The answer, in order from biggest drop to smallest, goes: Pittsburgh, Houston, New England, Green Bay.

Didn't expect that, I bet?

An idea that came up in the DVOA discussion thread today is that the Packers take their foot off the gas in the fourth quarter, and that's the biggest reason they don't have a historically dominant DVOA that compares with teams like the 2007 Patriots and 1998 Broncos. Well, if DVOA is to believed, this is simply not true. In general, compared to this year's other top offenses, the Packers don't drop off much in the fourth quarter. This week against the Raiders was a dramatic exception, with the Packers putting up 33.4% DVOA in the first three quarters and then -155.4% DVOA in the fourth quarter (on only eight plays, compared to 53 plays in the first three quarters).

Actually, the offense which drops off the most in the fourth quarter is Miami, which is slightly above average for three quarters and then the worst offense in the league in the fourth quarter. Apparently, the Dolphins take their foot off the gas even when they are losing the race. And of course, we know which team improves the most in the fourth quarter.

MIA 7.1% 14 -38.4% 32 -45.5%
PIT 28.8% 4 -16.0% 27 -44.8%
HOU 25.7% 5 -4.2% 24 -29.9%
CHI -2.3% 21 -25.5% 29 -23.2%
NE 41.0% 1 19.1% 8 -21.9%
SD 19.7% 7 -0.7% 19 -20.5%
STL -22.1% 31 -36.9% 31 -14.9%
CIN 9.0% 10 -3.4% 23 -12.4%
KC -16.0% 29 -26.6% 30 -10.6%
GB 37.2% 2 26.9% 3 -10.4%
BAL 10.8% 8 4.5% 14 -6.3%
BUF 6.6% 15 0.6% 17 -6.0%
OAK 2.8% 18 -1.7% 20 -4.5%
DET 7.2% 13 3.3% 16 -3.8%
ATL 10.1% 9 7.0% 13 -3.1%
WAS -8.5% 24 -10.4% 26 -1.8%
SF -2.0% 20 -3.2% 22 -1.2%
CAR 19.8% 6 19.6% 7 -0.2%
MIN -2.5% 22 -1.8% 21 0.7%
JAC -22.1% 32 -19.0% 28 3.1%
PHI 6.3% 16 9.6% 11 3.3%
NO 32.4% 3 37.2% 1 4.9%
IND -15.7% 28 -9.9% 25 5.8%
CLE -9.9% 25 -0.1% 18 9.8%
DAL 8.8% 12 19.8% 6 11.0%
TEN 4.6% 17 18.9% 9 14.3%
TB -8.3% 23 7.6% 12 15.9%
NYG 8.9% 11 30.2% 2 21.3%
NYJ 0.7% 19 23.5% 4 22.8%
ARI -19.9% 30 3.8% 15 23.7%
SEA -12.6% 26 13.7% 10 26.2%
DEN -12.9% 27 22.8% 5 35.6%

Actually, Green Bay seems to take its foot off the gas more on defense; its defense would rank 17th if we didn't include the fourth quarter. But San Francisco and New England see their defensive DVOA ratings decline even more in the fourth quarter than Green Bay's.

CAR 9.1% 24 42.7% 32 33.6%
SF -20.9% 2 9.9% 20 30.8%
PHI -3.0% 10 23.8% 29 26.8%
NE 8.0% 21 32.9% 31 24.9%
NYG 4.4% 16 28.0% 30 23.6%
MIA -5.6% 8 15.8% 23 21.4%
SD 8.6% 22 21.8% 28 13.2%
GB 5.7% 17 17.7% 25 12.1%
MIN 8.8% 23 20.2% 26 11.3%
DET -10.1% 5 -1.9% 13 8.3%
WAS -0.6% 12 7.2% 17 7.8%
ARI 6.7% 18 14.2% 22 7.5%
BUF 13.1% 29 20.4% 27 7.2%
DAL 1.6% 13 8.8% 19 7.1%
HOU -10.0% 6 -3.8% 11 6.2%
BAL -22.2% 1 -18.0% 2 4.2%
TB 13.0% 28 16.1% 24 3.1%
NYJ -12.9% 3 -10.5% 8 2.5%
OAK 7.0% 19 7.5% 18 0.5%
JAC -10.6% 4 -11.6% 7 -1.0%
TEN 2.4% 15 -0.4% 15 -2.8%
NO 17.6% 31 12.7% 21 -4.9%
STL 11.4% 27 5.0% 16 -6.4%
DEN 7.1% 20 -1.2% 14 -8.4%
CHI -7.1% 7 -16.9% 4 -9.8%
SEA 2.2% 14 -10.5% 9 -12.7%
PIT -0.8% 11 -17.6% 3 -16.8%
CLE 13.8% 30 -4.8% 10 -18.6%
ATL -3.1% 9 -22.6% 1 -19.5%
KC 10.7% 26 -13.9% 6 -24.6%
CIN 10.5% 25 -15.9% 5 -26.4%
IND 25.1% 32 -2.1% 12 -27.1%

Here is what the overall ratings would look like if we just included the first three quarters -- except in special teams, where frankly I'm too lazy right now to go do a whole new set of "first three quarters" special teams ratings.

1 HOU 25.7% -10.0% 1.1% 36.8%
2 NE 41.0% 8.0% 3.0% 36.0%
3 GB 37.2% 5.7% 2.9% 34.4%
4 PIT 28.8% -0.8% 2.3% 31.9%
5 BAL 10.8% -22.2% -3.5% 29.5%
6 SF -2.0% -20.9% 8.4% 27.3%
7 NYJ 0.7% -12.9% 4.7% 18.4%
8 CHI -2.3% -7.1% 10.0% 14.7%
9 NO 32.4% 17.6% -0.4% 14.4%
10 MIA 7.1% -5.6% 1.2% 13.9%
11 ATL 10.1% -3.1% 0.4% 13.6%
12 DET 7.2% -10.1% -5.9% 11.3%
13 PHI 6.3% -3.0% 0.5% 9.8%
14 SD 19.7% 8.6% -2.4% 8.8%
15 TEN 4.6% 2.4% 5.1% 7.3%
16 DAL 8.8% 1.6% -1.3% 5.8%
17 NYG 8.9% 4.4% 1.2% 5.8%
18 CAR 19.8% 9.1% -6.0% 4.7%
19 CIN 9.0% 10.5% 2.0% 0.6%
20 OAK 2.8% 7.0% -1.0% -5.3%
21 BUF 6.6% 13.1% -1.8% -8.3%
22 WAS -8.5% -0.6% -0.5% -8.5%
23 SEA -12.6% 2.2% 1.0% -13.8%
24 MIN -2.5% 8.8% -3.0% -14.3%
25 JAC -22.1% -10.6% -2.9% -14.4%
26 DEN -12.9% 7.1% 3.7% -16.3%
27 TB -8.3% 13.0% 1.0% -20.3%
28 CLE -9.9% 13.8% -0.5% -24.2%
29 ARI -19.9% 6.7% 2.3% -24.3%
30 KC -16.0% 10.7% 0.7% -26.0%
31 STL -22.1% 11.4% -2.8% -36.3%
32 IND -15.7% 25.1% -5.6% -46.3%


66 comments, Last at 20 Dec 2011, 12:41pm

47 Re: FO Mailbag: DVOA Without Fourth Quarter

I was pleasantly surprised watching the Monday Night game. While I wouldn't exactly call Jackson good, I also wouldn't exactly call him bad. He's at least settled into mediocre NFL QB territory.

Which while not amazing, is still probably 99.95th percentile performance. =)

2 Re: FO Mailbag: DVOA Without Fourth Quarter

You know I knew there'd be a huge 4th quarter improvement for Denver, but somehow I still managed to be surprised by just how ridiculous it is.

What's worse is, there are 5 Kyle Orton games in there where they didn't improve in the 4th quarter dragging that number down.

9 Re: FO Mailbag: DVOA Without Fourth Quarter

I was wondering about this too since the Week 12 DVOA article, noted an improvement in the run offense after switching from BT to AT. What I really wanted to see, though was their 1-3/4 Defensive DVOA split.

Excluding special teams, Denver appears to improve in the fourth quarter (+44.0%) just as much as New England tanks (-46.8%). This could be a fun game to watch.

22 Re: FO Mailbag: DVOA Without Fourth Quarter

The offensive improvement is insane. I'm trying to decide if it's a sign of a good team waiting to break out or the sign of a fluky team waiting to crash back down to Earth.

As others have said, the Pats/Broncos game could be all sorts of fun if it's still close in the fourth quarter.

26 Re: FO Mailbag: DVOA Without Fourth Quarter

This Bears fan's anecdotal impression is that the Tebowfence wears teams out really badly (well it did the Bears but part of that has to be a function of the Hanie and Co not moving the ball at all). Constantly having to work to maintain gap integrity on every single down and then chase Tebow around the field on third downs - and then the bugger runs over a crowd to pick up another third down - seems to really take it out of a team's legs. The variety in the running game combined with the spread looks and Tebow constantly trying to roll out to his left means that on every play defenders have extra responsibilities; extra stuff to counter against means more effort expended to ensure you are responding to the correct play and haven't missed your keys and gotten lost. Playing at Mile High can't hurt either (and not because it brings Tebow that bit closer to God).

40 Re: FO Mailbag: DVOA Without Fourth Quarter

My response was to a Broncos fan that thinks the reason for the Broncos doing better in the fourth quarter last week was down to the use of the spread. My point is that if Denver had played the spread all game I don't think they would have scored any more points. The issue was fatigue, I have watched pretty much every snap of Urlacher's career and rarely seen him looking so gassed.

3 Re: FO Mailbag: DVOA Without Fourth Quarter

Who was in the NYT's Fifth Down blog who showed that, adjusting for score and time remaining, GB was the pass-happiest team in the league? The data was only for the first half of the season, but he certainly didn't find any conservative late game pattern.

6 Re: FO Mailbag: DVOA Without Fourth Quarter

Lazy or not, it really seems like FO is making an effort at addressing these sorts of nagging questions more this season. We appreciate it.

41 Re: FO Mailbag: DVOA Without Fourth Quarter

I came in to post a similar sentiment. It may just be me (or just short-term memory), but it seems like this year stands out in FO's overall willingness to come out and address these types of complaints when they're voiced over several weeks and from several ... vocal commenters. I also appreciate this effort, and have found these posts very enlightening. Thanks to all behind the curtain.

7 Re: FO Mailbag: DVOA Without Fourth Quarter

That the 49ers decline in 4th quarter defence doesn't surprise me, they don't substitute at all on defence, partly because they don't have great depth and partly because the starting ends move to tackle to pretty good effect. I would also point out that the one pass rusher that they do substitute in, Aldon Smith, produces at a pretty high level. Increasing their defensive depth should be an offseason priority.

12 Re: FO Mailbag: DVOA Without Fourth Quarter

Very, very cool stuff.

I tend to trust this more than full game DVOA as a measure of the teams' strength. But for some of these teams, the difference between the fourth quarter and the rest of the game is the story of the year.

As bad as the Patriots defense is, their prevent defense is far worse. Denver really does take off the training wheels in the fourth quarter.

18 Re: FO Mailbag: DVOA Without Fourth Quarter

A game doesn't have to be out of hand. It just needs a large enough margin that teams no longer play for what DVOA thinks they play for - the best next score (on average) regardless of time consumed.

I wouldn't suggest cutting the fourth quarter out of DYAR - the plays really happen after all. But DVOA is an average of a measurement with known issues in the fourth quarter. Why not use a sample that doesn't suffer from those issues? Do you think teams are trying to play badly in the first three quarters? Do you think that the rules of the game change?

By the way, drives with two score leads or deficits are very common in the fourth quarter. But these are precisely the drives where coaches are tempted to alter their schemes to save or burn time. It varies from coach to coach. Some coaches go full-prevent way too early. Others wait too long. On the other side of the ball, some teams get desperate early and others aren't desperate soon enough. But it's only the fourth quarter that has this problem in a meaningful way.

21 Re: FO Mailbag: DVOA Without Fourth Quarter

If those scenarios are so common, and if coaches are so tempted to alter their schemes, wouldn't you expect the 4th quarter to have somewhat different outcomes than the others and wouldn't that imply that it's important to consider? Frequent and different seems like a bad reason to discount a situation.

31 Re: FO Mailbag: DVOA Without Fourth Quarter

True. But that argues for tracking two separate DVOAs: most-of-the-game and fourth-quarter.

But here's the rub. We don't know if fourth quarter DVOA even works. There are a number of known problems with DVOA that are either specific to the fourth quarter or much worse in the fourth quarter.

(1) DVOA's success formula doesn't model correct fourth quarter strategic goals in many games
(2) The baselines for different down-distance-time-margin situations suffer from selection bias. (that is, teams that are desperate are more likely to be bad - polluting the baseline used for comparison)
(3) Fourth quarter plays often include scrubs put in to "try out" for a higher spot in the depth chart
(4) Fourth quarter plays include many meaningless plays, where neither team has incentive to do more than just practice something, much like preseason games

To simplify and overstate the case about the fourth quarter:
DVOA values the wrong things, compares to the wrong mix of teams, both grades and compares to the wrong players, and mixes in meaningless noise.

It doesn't do this all the time, because many fourth quarters are competitive. But it's often enough to make fourth quarters look as different from the first three as one season looks compared to the next.

32 Re: FO Mailbag: DVOA Without Fourth Quarter

Aaron has mentioned a number of times that he ran the numbers without the garbage time possessions, and the accuracy got worse. Certainly there can be improvements - he frequently admits to this as well - but it's not like the difference in the way the game is played late is being ignored.

35 Re: FO Mailbag: DVOA Without Fourth Quarter

Sure. Eliminating just garbage time by whatever criteria introduces more selection bias, and makes things worse. That's why it's better to just look at the quarters. That way you eliminate the known issues while maintaining an unbiased sample of plays.

The tests that Aaron needs to do are these:

(1) To what extent is DVOA in each quarter "predictive" of other quarters?
(2) To what extent is DVOA in each quarter predictive of DVOA in the same quarter in the next season?

I suspect that quarters 1-3 are predictive of each other pretty well, and reasonably predictive of their next seasons, but that fourth quarter DVOA is not predictive of other quarters, and not as good at predicting its next season.

I believe this because while the first three quarters DVOA is determined primarily by your team's skill, fourth quarter DVOA is dictated by both skill and the mix of game situations you face. I think the mix of fourth quarter situations is determined largely by factors that have nothing to do with skill (luck, small sample size, opponent strength to name three).

17 Re: FO Mailbag: DVOA Without Fourth Quarter

Well, thanks Aaron. I appreciate the digging. I turned out to be half-right-- Packers are clearly a better 1-3 Quarter team and more dominant viewed that way, but they're not alone. So I am duly chastised. And as for the Broncos, Holy Crunch Time, Batman!! And the Giants get all silly at both ends of the field, don't they??

34 Re: FO Mailbag: DVOA Without Fourth Quarter

Eli being insane in the 4th quarter this year is largely due to the fact that the defense usually collapsing at the same time.

I think this is the year that Giants/Eli most closely resemble the past couple years of Peyton/Colts. Both have shitty defenses and no running game. QB and WR play getting it done almost entirely by themselves.

48 Re: FO Mailbag: DVOA Without Fourth Quarter

I think part of it is that the Colts became so infamous for pulling starters in the last two weeks and losing games to inferior competition that everyone started discounting them as a potential 16-0 team.

The Manning Colts could lose week 17 games to the Little Sisters of the Blind.
The Painter Colts could lose that game in weeks 1-16, too.

36 Re: FO Mailbag: DVOA Without Fourth Quarter

Maybe it's the lack of depth on defense that's killing the Packers (I think the Pats have the same problem). Injuries mean the d-line rotation is short; the last game was played with two middle linebackers with a total of one full and two partial seasons between them; Nick Collins means Charlie Peprah is playing all the time, and Pat Lee is missing, too; Matthews and Woodson don't practice a lot because they're dinged up.

The 4th quarter drop-off would not be unexpected in this case.

19 Re: FO Mailbag: DVOA Without Fourth Quarter

Well, based on this, the Saints & Packers are going to have a 4Q shootout if they meet in the NFCCG. If the Saints can keep it close in the 1st 3 Q's, their D gets better in the 4th, & GB's goes down--enough that they pass each other in the rankings.

37 Re: FO Mailbag: DVOA Without Fourth Quarter

This is good stuff, but I have to say it leaves me wanting more.

What would really be enlightening is separating out garbage time vs. non-garbage time in the 4th quarter. This is the big question concerning the Packers because they have had more garbage time than probably every other team. Yet they have had some close 4th quarters as well, so this data doesn't show if the offense drops off uniformly across all 4th quarters, or if it drops off a lot in blowouts and is offset by strong play in the fewer close games.

50 Re: FO Mailbag: DVOA Without Fourth Quarter

You already know the Packers are undefeated-- that means they've done well (enough) in the 4th quarter of all games close at the end. I'm not sure there's much more to learn.

Including selected 4th quarter plays would certainly boost the Packers' DVOA; but it won't give them a dominant first place DVOA ranking, if that's what you're looking for, since the Q1-3 DVOA puts them right in the midst of BAL, HOU, NE, and PIT.

44 Re: FO Mailbag: DVOA Without Fourth Quarter

Here's a cool finding...

The correlation of this year's 3Quarter Offensive DVOA to last years Offensive DVOA: 0.51

Same check of 4thQuarter Offensive DVOA to last year's OffDVOA: 0.14

It looks like 4th Quarter DVOA 'predicts' very little about a team's overall quality in an adjacent season, even when that quality factors in the fourth quarter.

Same check on the defense...
3Quarter correlation: 0.11
4thQuarter correlation: -0.04

Once again, 4th Quarter DVOA is worse at 'predicting' performance. In fact, it's essentially useless as a predictor. (3Quarter Def DVOA isn't that great a predictor either)

What's it mean? Either fourth quarter football requires a separate kind of football skill, or something other than football skill is being measured by fourth quarter DVOA. Either way, they should be reported and judged separately.

45 Re: FO Mailbag: DVOA Without Fourth Quarter

I would suggest fatigue on defense. Defensive coordinators seem to realise (and have been paraphrased as having said so) that they can only expect their defensive players to last 55 snaps or so (their number not mine - although FO reader Kal studied the Oregon offense and found the 55 number there as well so maybe there is something to it). It might be that some schemes suffer more than others either in terms of wearing down faster or the scheme not working as players slow down.

I would love to see if either offensive or defensive DVOA vary with the number of snaps played.

54 Re: FO Mailbag: DVOA Without Fourth Quarter

DVOA has no concept of game time, which is what most of the issue comes from.

DVOA thinks a team, down by 30 points, going on a 10 minute, 15 play TD drive is a very strong drive. Everyone watching the game knows that it made that team less likely to win.

There are situations where points are only valuable if they can be had quickly, and DVOA has no way of understanding that. Thats the primary reason the 4th quarter numbers are a bit silly.