Writers of Pro Football Prospectus 2008

Most Recent FO Features

manningbrees.jpg

» Scramble for the Ball vs. DYAR Fantasy Football

Mike and Tom finally get around to a candid discussion about the oft-requested and never-implemented DYAR fantasy football league.

16 Nov 2004

Week 10 DVOA Ratings

by Aaron Schatz

Here are the Football Outsiders team efficiency ratings through Week 10 of 2004, measured by our proprietary Value Over Average (VOA) system that breaks down every single play and compares a team's performance to the league averaged based on situation in order to determine value over average.  (Explained further here.)

Moving UpIND (8 to 4), ATL (17 to 13), STL (28 to 24)

Moving DownHOU (11 to 15), BUF (15 to 19), CHI (21 to 25), DAL (24 to 28)

New England once again takes over the top spot in our ratings this week, as their demolition of the Buffalo Bills comes out as slightly more impressive than Philadelphia's demolition of the Dallas Cowboys -- mostly because Buffalo has been a better team this season than Dallas has.  Pittsburgh remains behind two teams that it has beaten on the field, because its total performance this year is slightly less impressive.

I know I keep promising a VOA mailbag, and I really do hope to do one sometime this week.  I'm running obscenely behind on emails.  But I want to talk about one question this week, because it brings up some interesting aspects of the current ratings.  Chris Miraglia emailed and asked:

"Does DVOA weight plays more in the clutch?  If not, I think a good idea might be to do a clutch factor in games. For example, in games where the score is within 7 going into the two minute warning, double the effect that the play has. A 19 yard completion is much more important when trying to ice a game then on the opening drive. A kind of hyperbola effect happens also: a 19 yard completion when you're winning 42-7 with 4 minutes left means next to nothing (in fact, I don't even think you count it). It would probably have the added benefit of bringing down teams like KC and Tampa, while raising teams like Jacksonville and putting teams more in line with how they actually play."

The answer is that DVOA itself does not weight plays more in the clutch.  I guess we can ask the same question about "clutch" that baseball stats people have asked for years: does good performance in the clutch in the past signify good performance in the clutch in the future?  I'm not sure.  There is obviously one difference in football, because in "clutch" situations the rules change somewhat.  You are now concerned with saving time and running a two-minute drill and so you run different types of plays.  (It isn't like the strike zone changes when a team is down by one in the ninth inning.)

My goal with DVOA is to get a better idea of team ability, divorced from the bias of situation, so I specifically did not make it an "added value" formula that measures how much a specific play increases a team's chances of winning.  I don't want to penalize a good running back on a bad team whose third quarter runs never help his team win because they are always losing by two touchdowns -- a system like that would say that LaDainian Tomlinson has been much more valuable in 2004 than he was in 2003, which is absurd.  I also don't remove plays when the game is a blowout -- even those 19 yard completions with the score 42-7 in the fourth quarter count.  (As I've noted before, I've done a little playing around with the numbers to see if they are more accurate when plays during blowouts are removed, and they are not.)

I do, however, have a number that takes clutch performance into account.  If you look over to the left-hand column you'll notice that we run a little box each week with the top five teams listed by estimated wins.  This is our statistic that is based on a complicated formula that includes not only total DVOA and consistency from week to week, but also DVOA in the most important specific situations: red zone defense, first quarter offense, and performance in the second half when the score is close (defined as a spread of 0-8 points).

Look at the estimated wins numbers, though.  You might be surprised by the effect of including "clutch" performance.  Not only is Pittsburgh below Philadelphia and New England in estimated wins, the Steelers are also below two 6-3 teams, Indianapolis and Baltimore.

Pittsburgh, it turns out, has a poor DVOA in the second half of close games, particularly on defense.  Pittsburgh's defensive DVOA is -18.2%, second in the league, but their defensive DVOA in the second half of close games is 20.9%, 27th in the league.  Nobody has noticed, because those games were all early in the season, the Steelers ended up pulling them all out, and they all came against terrible teams: Oakland, Miami, Cincinnati, and Dallas.  When the Steelers blew out the Eagles and Patriots, this issue didn't rear its head.  When some team does give Pittsburgh a close game, will their defense wilt at the end?  Or is this simply an issue of small sample size, not to mention a trend from earlier in the season when the Steelers weren't playing as well as they are now?  I am guessing that some people feel that there is a flaw in DVOA because it ranks the Steelers below two teams that they have already beaten, but if we gave a greater weight to clutch performance, Pittsburgh would be even lower.

What about the other teams that Chris mentioned?  Yes, Kansas City does in fact have a worse DVOA in the second half of close games both on offense and on defense.  They still have a high number of estimated wins because they are second in the league in first quarter offense.  First quarter offense, it turns out, is almost as critical to winning games as playing well in "clutch" situations.  If you get out to a big lead, it is less likely you'll even face one of those clutch situations later.  The top team this year in first quarter offense is pretty obvious: Philadelphia.

Jacksonville?  Their defense in the second half of close games (-9.8%) has been better than their total defense (-1.4%).  But that is matched by the fact that their offense in the second half of close games (-9.5%) has been worse than their total offense (1.2%).  So they really aren't better or worse in the clutch.

Tampa Bay has had a terrible offense in the second half of close games.  But they have also had the second-best defense in the league in the second half of close games.  Their total defense has been -7.0%, but in the second half of close games they are -39.9%.  So "clutch" is not the problem for Tampa Bay, unless that's Matt Stitchcomb's nickname or something.

In fact, if you want to talk about a team that defied the idea that "clutch" performance is undervalued in DVOA, it is the 2003 Tampa Bay Bucs.  Remember, they are the team that stayed in the top ten all year despite losing close game after close game, just like this year's Chiefs.  So they must have been horrible in the second half of close games, right?  Nope, the opposite.  Tampa Bay had the third-highest offensive DVOA of 2003 in clutch situations, and the fourth-highest defensive DVOA.  They were 44 of the best players in football when a game was within a touchdown in the second half, with Martin Grammatica being the unfortunate exception.  No, the problem with last year's Bucs is that they were horrible in the first quarter, with the worst first quarter offensive DVOA by far.  Every week they dug themselves a hole early and spent the whole game trying to claw out of it, only to fall short in the final minutes.  It just so happened that the one exception to this pattern, the one game they blew with poor late-game play, was on national television on a Monday night, which gave everyone the idea that Tampa Bay was a team that couldn't perform late in close games.

There's one more team that DVOA seems to have a hard time getting a handle on this season, but Chris didn't mention: St. Louis.  I have to admit that I was pretty surprised that they defeated Seattle in such a sound fashion, in a game that was not as close as the final score would indicate.  DVOA ranks them 24th despite being 5-4.  The estimated wins formula likes the fact that they get out to a strong offensive start in the first quarter, but giving more weight to "clutch" performance certainly wouldn't make their DVOA more accurate.  The Rams defense has been about 10% better in the second half of close games, but their offense has been about 10% worse.

One more team that should be discussed in terms of estimated wins: New England.  They have the best total DVOA this season, but they are also the best team in all the important splits of DVOA that are considered in the estimated wins formula.  They have the best clutch defense, they are second in clutch offense behind Minnesota, they are second in red zone defense behind Baltimore, and they are in the top ten in first quarter offense.  Their only negative indicator is a high VARIANCE based almost entirely on the fact that they sucked against Pittsburgh.

* * * * *

Speaking of advanced versions of DVOA, this week we introduce WEIGHTED DVOA.  That's the formula that weights recent games more than early games to try to get a better idea of which are the best teams now rather than over the season as a whole.  Since this is the first week we're doing this statistic, it doesn't come out much different from regular total DVOA.  Only the games from Weeks 1-3 are really weighted differently, and the only team really affected is Seattle.  This will gradually change as the year moves forward.  You'll notice WEIGHTED DVOA below but also on the offense and defense pages.

Week 10 means that opponent adjustments are also now all at full strength.

* * * * *

I was listening to Dale Arnold and Bob Neumeier on local sports radio WEEI in Boston, and they were discussing the slate of games on local television last weekend.  With the Patriots on Sunday night, the networks could give us whatever they thought were the best games of the weekend.  "This doesn't happen very often," said Dale, "But I look at the schedule this weekend and there are just no good games to watch.  Is anybody excited by any of these matchups?"  Um, actually Dale, yes.  We got Jets-Ravens, Falcons-Bucs, and Packers-Vikings.  All three of those games had playoff ramifications.  One went down to the final minutes, another to overtime.  What the heck was Dale complaining about?

Now, this weekend, that's something to complain about.  Once again the Patriots don't play Sunday afternoon, so once again I took a look at what games we are getting on our local affiliates.  Browns-Jets and Cowboys-Ravens.  Yuck.  That latter game should be held in a toxic waste dump.  So I thought, ok, maybe I need to head to a bar this weekend to catch something better.  But there is nothing better.  Take a look at the NFL schedule this weekend.  Count the number of games between teams with winning records.  One.  That's it: Atlanta Falcons and New York Giants at 4 p.m.  (This is, by the way, the last week the Giants will count as a team with a winning record.)  How the hell did this happen?

* * * * *

OFFENSE and DEFENSE DVOA are adjusted based on strength of opponent as well as to consider all fumbles, kept or lost, as equal value.  SPECIAL DVOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season.  NON-ADJ TOTAL VOA does not include these adjustments.  DVOA is a better indicator of team quality.  VOA is a better indicator of wins so far.

As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.


TEAM TOTAL
DVOA
LAST
WEEK
NON-ADJ
TOTAL VOA
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
1 NE 40.9% 2 31.5% 8-1 26.3% 6 -13.6% 3 1.0% 14
2 PHI 39.8% 1 36.0% 8-1 34.8% 2 5.7% 21 10.6% 2
3 PIT 33.3% 3 44.3% 8-1 14.3% 10 -18.2% 2 0.8% 16
4 IND 27.0% 8 23.3% 6-3 38.5% 1 6.4% 22 -5.1% 30
5 BAL 22.7% 4 19.7% 6-3 -4.6% 21 -19.9% 1 7.4% 4
6 SD 19.4% 6 32.2% 6-3 21.3% 8 3.5% 20 1.6% 13
7 KC 17.2% 7 9.1% 3-6 30.6% 3 7.3% 23 -6.2% 31
8 NYJ 14.2% 5 30.1% 6-3 27.5% 5 14.3% 25 1.0% 15
9 DEN 10.8% 9 13.2% 6-3 9.6% 11 -2.4% 10 -1.1% 19
10 SEA 5.4% 10 24.2% 5-4 0.0% 15 -7.7% 6 -2.2% 25
11 NYG 4.4% 13 19.1% 5-4 0.4% 14 -5.6% 8 -1.6% 21
12 MIN 2.6% 14 0.4% 5-4 30.2% 4 23.7% 28 -3.9% 28
13 ATL 2.2% 17 4.6% 7-2 -4.3% 20 -0.8% 16 5.7% 5
14 DET 1.2% 12 2.6% 4-5 -1.1% 18 0.8% 17 3.0% 7
15 HOU 0.6% 11 -8.6% 4-5 -0.5% 16 -1.1% 14 0.0% 17
16 CIN 0.5% 16 -3.4% 4-5 -0.9% 17 1.2% 19 2.7% 9
TEAM TOTAL
DVOA
LAST
WEEK
NON-ADJ
TOTAL VOA
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
17 JAC -0.5% 19 -5.8% 6-3 1.2% 13 -1.4% 13 -3.1% 26
18 TB -2.6% 18 -3.1% 3-6 -7.0% 24 -6.4% 7 -1.9% 23
19 BUF -3.5% 15 -10.7% 3-6 -15.1% 29 -3.9% 9 7.7% 3
20 GB -5.2% 22 -9.9% 5-4 16.5% 9 24.1% 29 2.3% 10
21 WAS -9.1% 20 -5.3% 3-6 -20.4% 30 -12.9% 4 -1.6% 22
22 CAR -9.4% 25 -9.3% 2-7 -6.9% 23 -1.0% 15 -3.5% 27
23 CLE -9.4% 23 -21.9% 3-6 -3.9% 19 10.6% 24 5.1% 6
24 STL -11.1% 28 -10.5% 5-4 24.9% 7 29.2% 32 -6.7% 32
25 CHI -11.3% 21 -6.2% 4-5 -25.9% 31 -11.8% 5 2.9% 8
26 NO -13.5% 27 -4.3% 4-5 -4.9% 22 19.7% 26 11.1% 1
27 TEN -15.0% 29 -11.2% 3-6 -12.5% 26 -2.3% 11 -4.8% 29
28 DAL -17.2% 24 -22.9% 3-6 6.5% 12 25.8% 30 2.1% 11
29 ARI -17.4% 26 -6.8% 4-5 -14.5% 28 0.8% 18 -2.1% 24
30 MIA -28.6% 30 -39.4% 1-8 -32.0% 32 -1.8% 12 1.7% 12
31 OAK -29.7% 31 -33.8% 3-6 -7.8% 25 20.4% 27 -1.6% 20
32 SF -41.0% 32 -46.3% 1-8 -13.5% 27 27.2% 31 -0.3% 18

  • 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.  It is based on the number of games a team has played so far, so teams which have not had their bye will appear higher.
  • 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.
  • PAST SCHEDULE lists average DVOA of opponents played so far, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative).
  • FUTURE SCHEDULE lists average DVOA of opponents remaining on the schedule, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative).
  • VARIANCE measures the statistical variance of the team's weekly DVOA performance.  Teams are ranked from least consistent (#1, highest variance) to most consistent (#32, smallest variance).


TEAM TOTAL
DVOA
W-L ESTIM.
WINS
RANK WEIGHTED
DVOA
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VARIANCE RANK
1 NE 40.9% 8-1 8.0 1 39.6% 1 1.8% 12 -3.5% 23 20.7% 8
2 PHI 39.8% 8-1 7.6 2 37.5% 2 1.9% 11 -6.7% 27 8.4% 28
3 PIT 33.3% 8-1 6.1 5 35.5% 3 1.1% 15 4.1% 10 20.3% 11
4 IND 27.0% 6-3 6.3 3 28.2% 4 1.2% 13 4.1% 11 4.2% 31
5 BAL 22.7% 6-3 6.1 4 25.3% 5 8.2% 3 8.6% 4 20.3% 10
6 SD 19.4% 6-3 6.0 6 20.5% 6 -4.5% 28 4.3% 9 14.6% 20
7 KC 17.2% 3-6 5.5 8 17.7% 7 4.1% 8 2.3% 15 21.7% 6
8 NYJ 14.2% 6-3 5.7 7 13.1% 8 -2.4% 21 6.0% 7 13.5% 21
9 DEN 10.8% 6-3 5.2 9 9.6% 9 -0.3% 19 -3.3% 21 17.4% 14
10 SEA 5.4% 5-4 4.3 21 1.4% 13 -11.8% 32 -6.8% 28 9.6% 25
11 NYG 4.4% 5-4 4.7 15 3.0% 10 -2.9% 23 10.3% 2 21.7% 5
12 MIN 2.6% 5-4 4.9 10 3.0% 11 1.1% 14 -2.6% 19 15.1% 18
13 ATL 2.2% 7-2 4.8 13 2.7% 12 -3.7% 25 -8.4% 31 27.5% 3
14 DET 1.2% 4-5 4.9 11 0.6% 16 0.4% 17 -2.4% 18 30.4% 1
15 HOU 0.6% 4-5 4.7 14 0.8% 15 3.7% 9 0.0% 17 9.3% 26
16 CIN 0.5% 4-5 4.3 20 1.2% 14 0.2% 18 18.3% 1 15.1% 17
TEAM TOTAL
DVOA
W-L ESTIM.
WINS
RANK WEIGHTED
DVOA
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VARIANCE RANK
17 JAC -0.5% 6-3 4.6 17 -0.2% 18 9.4% 2 -3.5% 24 7.1% 29
18 TB -2.6% 3-6 4.6 16 0.0% 17 -4.3% 27 -9.9% 32 11.6% 23
19 BUF -3.5% 3-6 4.9 12 -3.1% 19 6.3% 6 -7.3% 29 20.9% 7
20 GB -5.2% 5-4 4.4 18 -3.7% 20 -3.0% 24 3.0% 14 22.1% 4
21 WAS -9.1% 3-6 3.3 28 -9.6% 22 -1.9% 20 8.8% 3 11.4% 24
22 CAR -9.4% 2-7 4.4 19 -9.5% 21 2.1% 10 -8.4% 30 8.7% 27
23 CLE -9.4% 3-6 4.0 24 -10.3% 23 14.5% 1 6.2% 6 20.7% 9
24 STL -11.1% 5-4 4.1 23 -10.6% 24 -5.5% 30 -3.2% 20 15.6% 16
25 CHI -11.3% 4-5 3.6 26 -11.2% 25 -2.8% 22 1.2% 16 6.4% 30
26 NO -13.5% 4-5 3.7 25 -13.2% 26 -6.4% 31 -3.3% 22 4.1% 32
27 TEN -15.0% 3-6 3.1 29 -14.9% 27 0.5% 16 3.8% 12 28.7% 2
28 DAL -17.2% 3-6 4.2 22 -17.7% 29 6.5% 5 5.5% 8 14.8% 19
29 ARI -17.4% 4-5 3.3 27 -17.1% 28 -5.0% 29 -6.2% 26 17.2% 15
30 MIA -28.6% 1-8 2.1 31 -28.2% 30 6.2% 7 3.7% 13 19.2% 12
31 OAK -29.7% 3-6 2.5 30 -32.2% 31 6.9% 4 7.3% 5 19.1% 13
32 SF -41.0% 1-8 1.8 32 -42.1% 32 -3.9% 26 -4.5% 25 13.4% 22

Posted by: Aaron Schatz on 16 Nov 2004

comments