Writers of Pro Football Prospectus 2008

Most Recent FO Features

SimBla15.jpg

» OFI: Don't Make Saban Angry

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

12 Oct 2004

Week 5 DVOA Ratings

by Aaron Schatz

Here are the Football Outsiders team efficiency ratings through Week 5 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 average based on situation in order to determine value over average.  (Explained further here.)

Moving Up: TEN (30 to 22), BUF (24 to 18), DET (13 to EIGHT), SD (14 to 9), HOU (21 to 16)

(How many games, Rene? EIGHT games in one day. EIGHT. 8))

Moving Down: ATL (5 to 13), OAK (20 to 28), NO (17 to 24), WAS (22 to 27)

This week for the first time this year we introduce opponent adjustments into the formula, which turns VOA into DVOA (or Defense-adjusted Value Over Average).  Yes, it is called DVOA even we're adjusting defenses for the quality of opposing offenses, we just thought it sounded better that way.  The adjustments are only at 50 percent strength, so that we don't overcompensate for one extreme performance, and to lessen the effect of the "feedback loop" that occurs early in the season.  That's the effect where good teams look good because they've played bad teams, and bad teams look bad because they've played good teams.  Until more games are played, it is difficult to tell which teams are truly good and bad, and which are just mediocre.  You'll notice that nearly every team is drawn towards 0% by adjustments; that will change over the course of the year as more games are played.  Each week we'll increase the strength of the adjustment until after Week 10, when every team has hit their bye and played the same number of games, we're at 100 percent.

One change this year is that the difference between DVOA and non-adjusted VOA on the table below does not reflect only schedule strength.  It also reflects which teams have recovered a greater or lesser share of fumbles than expected.  So when a team's DVOA is higher than its VOA, while this probably indicates a hard schedule, that might not be the only reason.  What is important is that DVOA is a better indicator of a team's quality the rest of the season than non-adjusted VOA.

This week we also add in ESTIMATED WINS.  The top five have been appearing on the side all season, but we'll do all 32 teams from now on.  This is the statistic that considers DVOA in certain situations as more important, including red zone defense, first quarter offense, and performance in the second half of close games.  For all who have been asking, yes, it is based on total games played so teams with a bye will have a lower number until Week 10 is over.  To quote my fraternity pledgemaster Cris Modesto, "Suck it up and deal."

I figured I would use this week's commentary as a notes column to note which teams and players are playing better than would be indicated by conventional statistics, and which teams and players are not playing as well as it looks.

BETTER THAN THEY LOOK

The Philadelphia Eagles defense.  How could you be better than your record when your record has no losses?  The Eagles have played a harder schedule than the other three undefeated teams, including two of the top ten offenses, Minnesota and the New York Giants.

The Kansas City Chiefs defense.  I mentioned this a couple of weeks ago -- the Kansas City secondary can't tackle or pursue runners but they're playing good pass defense.  The Chiefs even rank above average against the run, although this is mostly due to a single game where Houston's Domanick Davis inexplicably rushed for 12 yards on 10 carries.  Kansas City was not playing like an 0-3 team before they beat Baltimore and that win may have been the emotional lift they needed to get back into the playoff picture.

Kansas City's defense also has something in common with the defenses in San Francisco and Dallas.  Each team has recovered only one fumble on defense, even though San Francisco has caused seven and Kansas City and Dallas have caused six apiece.  That's just unlucky random chance.

Jeff Garcia and Carson Palmer.  Here's that feedback loop I was talking about.  Cincinnati has played three defenses in our top ten: Pittsburgh, Baltimore, and Miami.  Cleveland has played four defenses in our top ten: Pittsburgh, Baltimore, Washington, and the Giants.  So do Pittsburgh and Baltimore look like they have good defenses because they've played Garcia and Palmer, or do Garcia and Palmer look like they have worse offenses because they've played Pittsburgh and Baltimore?  When it comes to Baltimore, at least, I would guess the latter.  Garcia has -0.3 PAR but 7.8 DPAR.  Palmer has -5.8 PAR but 5.3 DPAR.  (PAR is Points Above Replacement, DPAR is PAR adjusted for quality of defenses, both explained here.)

The Giants receivers, Tim Carter and Amani Toomer, who have played not just good overall defenses (Washington) but also defenses who have been better defending passes to wide receivers than defending passes to backs and tight ends (Dallas).  Unfortunately, Carter's now out for the year.

Eric Johnson.  Do you realize how much he is Tim Rattay's favorite receiver?  In Rattay's three starts, he has averaged 120 yards.  Antonio Gates has been great, but Johnson is now the top-rated tight end in DPAR.  Johnson and Gates are way ahead of the rest of the league this season, and Tony Gonzalez is middle of the pack.  (Yet, despite overpaying for him, my fantasy team is 5-0.   Just had to slip that in there.)

Curtis Martin.  Not only is he the top-rated running back, but DPAR says he's actually played an above-average schedule of opposing defenses.  As for the rest of the Jets...

WORSE THAN THEY LOOK

Seattle, Pittsburgh, and the New York Jets.  According to DVOA, they own the three easiest schedules in the league, and all three teams see their ratings drop significantly when opponent adjustments are added in.  For Seattle and the Jets, the distortion is caused by opponents who are poor on both sides of the ball.  For Pittsburgh, the offense is probably as good as it looks but the defense has feasted on, well, Jeff Garcia and Carson Palmer (see above), not to mention Miami's three quarterbacks, Ham, Shem, and Yaphet.  I picked all three of these teams for the playoffs partly for this reason -- all three looked to have easy schedules this season, and so far that looks like it is the case.

The Atlanta Falcons defense, who so far look like the St. Louis Rams of 2004.  By this I mean that they have caused eight fumbles and recovered seven of them.  That's more than their fair share of turnovers.  The Falcons also have played a relatively easy schedule of opposing offenses (Detroit, Arizona, San Francisco).

Steve McNair.  Last night doesn't really mean he's out of the woods, the Green Bay defense is just a shell of its former self.  I knew Grady Jackson was important, but I didn't think he was this important.

Edgerrin James.  He looks better than he has at any point since his injury, and his RB Success Rate of 63% (explained here) is the highest of any back with more than 40 carries.  But he's had a very easy schedule of opposing defenses.  Even Jacksonville now looks like it wasn't the great run defense we thought it was.

Kyle Boller's newfound rushing ability.  Lots of yards, but few first downs and two fumbles.

David Carr and Andre Johnson.  They're playing well, but their numbers are a bit inflated by games against poor defenses like Minnesota and Oakland.

* * * * *

We have some new stats this week on Football Outsiders, thanks to reader Jim Armstrong.  Jim has been tracking statistics like yards per drive and points per drive for a few years now, and he's provided his numbers to Football Outsiders.  You'll find them on the JUST THE STATS menu under Drive Stats, and the 2004 numbers will be updated weekly.  These are raw stats -- in other words, unlike DVOA, they have not been adjusted based on situation or opponent.  But they do separate offense from defense, so that, for example, Baltimore's offense doesn't get credit if the defense returns an interception for a touchdown.  The stats also feature the average line of scrimmage where each team begins its drives -- for example, San Diego has had the best average start on offense this season.  Drive stats also go back farther than DVOA; every year back to 1998 is up on the site right now.  Meanwhile, Jim and I in the coming months hope to work on ways to create some sort of "drive-based DVOA" that we can combine with regular ol' play-by-play DVOA to get an even better overall rating for each team.

* * * * *

Have you ever heard the statement that the NFL is a copycat league?  Remember Baltimore's flea flicker touchdown on Monday Night Football last week?  That was the only attempted flea flicker in the past two weeks.  This week, there were three flea flickers.

Best line I heard all weekend came from the announcers in the St. Louis-Seattle game.  "I think Marc Bulger's down... no, wait, that's Orlando Pace that is down."  Yes, Orlando Pace and Marc Bulger certainly do look alike.

* * * * *

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.  Opponent adjustments are currently at 50 percent strength and will gradually increase each week until Week 10.

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 PHI 51.3% 3 39.4% 4-0 39.0% 1 -5.6% 12 6.7% 3
2 NE 39.0% 2 42.4% 4-0 27.8% 4 -16.7% 4 -5.6% 29
3 SEA 35.0% 1 56.0% 3-1 4.3% 13 -33.6% 1 -3.0% 23
4 NYJ 23.6% 4 43.8% 4-0 24.5% 6 3.4% 17 2.5% 9
5 NYG 21.4% 6 36.0% 4-1 15.1% 9 -11.5% 7 -5.1% 28
6 IND 21.1% 8 20.4% 4-1 35.1% 2 12.8% 25 -1.2% 18
7 PIT 19.6% 7 29.1% 4-1 4.0% 14 -11.2% 8 4.4% 8
8 DET 18.3% 13 19.5% 3-1 -1.0% 17 -18.0% 3 1.3% 12
9 SD 17.3% 14 24.1% 3-2 20.0% 7 8.4% 20 5.6% 5
10 DEN 16.3% 11 14.7% 4-1 7.3% 12 -7.0% 10 1.9% 10
11 KC 15.3% 9 8.4% 1-3 17.7% 8 -3.8% 13 -6.2% 32
12 BAL 14.4% 12 22.1% 3-2 -10.4% 27 -11.0% 9 13.8% 1
13 ATL 9.8% 5 25.1% 4-1 -8.4% 21 -12.1% 6 6.0% 4
14 DAL 3.3% 10 -2.6% 2-2 8.9% 11 4.3% 19 -1.3% 19
15 MIN -3.2% 18 -1.5% 3-1 33.9% 3 31.1% 31 -6.0% 30
16 HOU -4.6% 21 -8.6% 2-3 9.8% 10 15.3% 26 0.8% 14

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 STL -5.1% 15 -11.9% 3-2 26.3% 5 30.1% 30 -1.4% 20
18 BUF -7.8% 24 -13.1% 0-4 -10.0% 26 3.5% 18 5.6% 7
19 TB -8.5% 23 -8.8% 1-4 -13.2% 30 -5.7% 11 -1.0% 17
20 JAC -9.2% 16 -7.3% 3-2 -8.4% 22 1.9% 16 1.1% 13
21 ARI -12.1% 19 -9.5% 1-4 -10.5% 28 -1.6% 14 -3.3% 24
22 TEN -13.2% 30 -2.5% 2-3 0.2% 15 9.6% 21 -3.8% 26
23 CAR -13.9% 25 -14.8% 1-3 -2.8% 18 10.5% 23 -0.6% 16
24 NO -14.0% 17 -7.2% 2-3 -3.2% 19 22.0% 29 11.3% 2
25 CHI -14.4% 26 -14.0% 1-3 -9.9% 25 0.7% 15 -3.7% 25
26 CLE -15.7% 27 -21.4% 2-3 -9.2% 23 12.2% 24 5.6% 6
27 WAS -17.5% 22 -24.6% 1-4 -26.0% 31 -14.6% 5 -6.1% 31
28 OAK -22.5% 20 -23.3% 2-3 -9.8% 24 10.1% 22 -2.7% 22
29 CIN -24.6% 29 -34.0% 1-3 -4.5% 20 16.0% 27 -4.0% 27
30 SF -27.9% 32 -41.2% 1-4 -11.1% 29 16.2% 28 -0.6% 15
31 MIA -29.1% 31 -41.2% 0-5 -50.8% 32 -23.4% 2 -1.8% 21
32 GB -30.9% 28 -38.9% 1-4 0.0% 16 32.7% 32 1.8% 11

  • 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).
  • 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.

TEAM

TOTAL
VOA

W-L

ESTIM.
WINS

RANK

PAST
SCHED

RANK

FUTURE
SCHED

RANK

VARIANCE

RANK

1 PHI 51.3% 4-0 3.9 1 5.6% 9 -5.3% 30 9.8% 19
2 NE 39.0% 4-0 3.7 2 -4.7% 25 0.8% 18 15.5% 10
3 SEA 35.0% 3-1 2.8 11 -13.9% 31 -3.0% 27 13.4% 13
4 NYJ 23.6% 4-0 3.1 7 -11.1% 30 3.7% 11 4.8% 29
5 NYG 21.4% 4-1 3.6 3 -1.9% 22 4.1% 9 11.4% 16
6 IND 21.1% 4-1 3.5 4 -7.4% 28 2.9% 13 6.0% 22
7 PIT 19.6% 4-1 3.1 8 -15.5% 32 7.1% 4 21.3% 6
8 DET 18.3% 3-1 2.7 12 10.5% 1 -7.4% 32 5.5% 24
9 SD 17.3% 3-2 3.4 5 2.6% 16 -1.8% 25 32.7% 2
10 DEN 16.3% 4-1 3.1 6 0.2% 19 -6.1% 31 5.1% 28
11 KC 15.3% 1-3 2.6 13 3.0% 12 2.6% 14 4.4% 30
12 BAL 14.4% 3-2 2.9 10 -4.6% 24 9.3% 1 25.9% 3
13 ATL 9.8% 4-1 3.0 9 -8.1% 29 2.1% 15 23.6% 4
14 DAL 3.3% 2-2 2.4 14 -3.7% 23 9.2% 2 15.8% 9
15 MIN -3.2% 3-1 2.1 18 8.9% 2 -1.3% 23 10.0% 18
16 HOU -4.6% 2-3 2.3 15 5.0% 10 -2.2% 26 10.9% 17

TEAM

TOTAL
VOA

W-L

ESTIM.
WINS

RANK

PAST
SCHED

RANK

FUTURE
SCHED

RANK

VARIANCE

RANK

17 STL -5.1% 3-2 2.2 16 -1.8% 21 1.7% 17 14.3% 12
18 BUF -7.8% 0-4 1.9 24 7.7% 6 -1.0% 21 5.8% 23
19 TB -8.5% 1-4 2.1 19 -0.6% 20 -4.5% 29 13.2% 14
20 JAC -9.2% 3-2 2.2 17 6.7% 8 -1.7% 24 6.1% 21
21 ARI -12.1% 1-4 2.0 22 0.4% 17 3.7% 10 15.0% 11
22 TEN -13.2% 2-3 1.6 27 -6.2% 27 -1.1% 22 41.0% 1
23 CAR -13.9% 1-3 2.0 21 2.6% 15 0.1% 19 5.3% 26
24 NO -14.0% 2-3 2.0 20 -5.6% 26 1.9% 16 5.4% 25
25 CHI -14.4% 1-3 1.5 29 8.9% 3 -4.2% 28 0.6% 32
26 CLE -15.7% 2-3 2.0 23 8.3% 5 6.8% 5 22.5% 5
27 WAS -17.5% 1-4 1.4 31 3.0% 13 5.9% 6 13.1% 15
28 OAK -22.5% 2-3 1.7 25 3.9% 11 5.2% 7 17.1% 8
29 CIN -24.6% 1-3 1.5 30 7.1% 7 7.9% 3 4.1% 31
30 SF -27.9% 1-4 1.5 28 2.7% 14 -1.0% 20 18.2% 7
31 MIA -29.1% 0-5 1.0 32 8.9% 4 4.7% 8 5.3% 27
32 GB -30.9% 1-4 1.7 26 0.2% 18 3.1% 12 7.9% 20

Posted by: Aaron Schatz on 12 Oct 2004

comments