Writers of Pro Football Prospectus 2008

Most Recent FO Features

RoethlisbergerBen04.jpg

» Any Given Sunday: Bucs Over Steelers

For several reasons, Sunday's upset in Pittsburgh looks more like a Steelers loss than a Buccaneers win.

09 Dec 2003

Week 14 Team Efficiency Ratings

by Aaron Schatz

Here are the team efficiency ratings after Week 14, 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 up: MIN (from #10 to #14), PHI (from #16 to #12), JAC (from #22 to #16)
Moving down: DAL (from #10 to #15), MIA (from #12 to #18), OAK (from #13 to #19)

OK, I know the first thing you are asking.  You don't even need to page down to the ratings tables.  Just look over to the left where we publish the top five teams each week and you'll notice that the best team in football by acclimation -- my beloved New England Patriots -- is not on top of the VOA ratings, or even in the top five.  Have we lost our marbles?

Well, no.  The problem remains that the total DVOA number is just that -- the TOTAL DVOA number, the sum total of every single play run this season with a few exceptions (kneels, spikes, plays I haven't figured out how to count yet like onside kicks).  New England started the year with a colossal thrashing at the hands of the Buffalo Bills that we now know was a complete fluke.  In Week 4, they were completely outplayed by Washington even though they actually had a chance to win that game at the end.  Both of these games count in the total DVOA rating.

At the same time, the Patriots have yet to have a real ass-whuppin' type of win.  Those two losses are their only two games with negative DVOA ratings, but their wins are a succession of good but not great performances.  This does not mean they are "lucky" -- they've been an above-average team in all 11 wins this year.  But at +62%, the DVOA system considers this week's win against Miami to be their best performance of the season.

Compare that with two teams who rank higher than the Patriots in total DVOA: Seattle and Tampa Bay.  Yes, I'm breaking my promise to never mention Tampa Bay again, but they are very useful for this exercise.  This graph -- oh, he's getting fancy now -- shows the game-by-game DVOA for New England, Seattle, and Tampa Bay (bye weeks show up as zero).  You can see that Tampa has had three wins more impressive than any of the Patriots' victories, while Seattle has had two.  On the other hand, since Week 4, Tampa and Seattle have each had four negative games while the Patriots have enjoyed an amazingly consistent run of mild if unspectacular success.

Even when it comes to WEIGHTED DVOA, the Patriots are #5 because of these two losses.  But as the WEIGHTED DVOA weights for the early season games become smaller and smaller, the Pats' ranking in WEIGHTED DVOA will move up.  Right now, I'm weighing Week 1 at 45% strength and Week 4 at 72% strength.  At season end, Week 1 will be at 9% strength with Week 4 at 45% strength.

Of course, every game of the season counts when you are adding up season totals.  I can't imagine the NFL going to Jamal Lewis at the end of the season and saying, "Jamal, we're taking away 100 yards from your season rushing total.  That 295-yard game against Cleveland was clearly a fluke since you only topped 140 yards one other time during the season."

In the same way, we can't just remove the Buffalo and Washington losses from the New England stats.  But that doesn't mean they aren't the best team in football right now.  Stats don't tell you which team is best, they help tell you which team is best.  You still have to interpret them in context.  So this might be the bias of a diehard Patriots fan but I do think New England is the best team in football right now and the favorite for the Super Bowl.  Their last three games won't be pushovers -- especially surging Jacksonville -- but they will likely win out to get home-field advantage throughout the playoffs.  They'll likely be on top of the WEIGHTED DVOA rankings at that point, and while I haven't had a chance to fully study this, it looks like home field is worth roughly 8-10% in DVOA for one game.  Those two facts combined would make them the favorites to go to Houston.

That week-by-week graph is kind of fun, so let's do another one to show three more interesting teams.  This next graph shows the most consistent and least consistent teams in the NFL in 2003, Carolina and Buffalo, as well as the #1 rated team, Kansas City.  OK, actually Buffalo is #2 in VARIANCE but do you really want to see the graph for Arizona?

I've gone on and on about how I think Carolina is having a fluke season but you may not realize they've been amazingly consistent in mediocrity whether winning or losing.  Buffalo has been as schizophrenic as Latka Gravas.  By putting Kansas City and Buffalo on this graph together, you can see the effects of the massive flogging the Bills received from the Chiefs in Week 8.  As good as Kansas City has been all year, that win is a bigger fluke than the Pats' loss in Week 1 because it was just so ridiculously dominant.  That win is also responsible for much of the Chiefs' lead on the total DVOA rankings.  Remove the worst loss and best win from each team, and the Chiefs lead in DVOA over #2 Indianapolis goes from 5.6% to 1.6%.

You can also see that the Chiefs have been fading over the past four weeks, and this graph doesn't show what you already know -- they are fading mostly on defense.  The one bright spot for the Chiefs is that the teams they may face on the road to the Super Bowl may not be able to take advantage of their glaring weakness stopping the run.  Three teams that will make the AFC playoffs are among the four worst rushing teams in the NFL according to our stats: New England (#31), Baltimore (#30, though admittedly much of this rating is related to their feet of stone quarterbacks, not Jamal Lewis), and Tennessee (#29).  Possible wild card Miami is #27.  Who said rushing was the key to winning football games?

Oh, yeah, Len Pasquarelli said it, yet again.  We've already taken a look at how running yards are usually the effect of wins, not the cause, both before the season and then after Week 1.  But if Pasquarelli is going to beat this into the ground I will too.  Here's the quote: "It isn't just a fluke that there were nine 100-yard rushing performances on Sunday, and all nine teams that had a runner go over the century mark won their respective contests."  Yes, that's true, but many of those backs sure did get their yards late in the game with their team ahead, not early on when their team was getting that lead:

  • 99 of Correll Buckhalter's 115 yards came in the fourth quarter, which started with the Eagles already up 19-10.
  • 63 of Jerome Bettis' 106 yards came in the second half, which started with the Steelers already up 17-7.
  • 81 of Kevan Barlow's 154 yards, plus 42 of Jamal Robertson's 46 yards, came in the second half, which started with the 49ers already up 34-0.  By the way, what was with all the San Francisco touchdown dances.  Yay, you guys ran it up on the worst team in football, on your home field.  That sure does make up for a year of massive underachievement that will end with you all sitting on the couch in January.

Many of the yards gained by backs like Travis Henry, Jamal Lewis, and Michael Bennett (not to mention backups like Onterrio Smith) also came late in games with their teams running out the lead.  This isn't to take away from the nice rushing days on Sunday -- especially by a Mr. Clinton Portis -- but rushing totals are still caused by a lead more often than they cause a lead. 

How does Portis' amazing game this week compare to Jamal Lewis' record-setting Week 2 performance when it comes to the stats we use here at Outsiders?  Well, this is where I was going to publish a list of the top ten rushing performances of the year and discuss them, but I won't have time today so I guess that gives everyone a reason (besides reading Scramble for the Ball and participating in discussion threads) to come back later in the week.  But I will note two interesting things about that game:

  • In the Five Showdowns article, I noted that Kansas City's rush defense was pretty good in the first quarter.  Portis' DVOA in the first quarter: -52%.  Portis' DVOA in the other three quarters: +95%.
  • Partly because of the first quarter, Denver's passing DVOA was higher than their rushing DVOA despite Portis' amazing performance: 145% to 58%.  15 of the 28 passes thrown by Plummer went for first downs or touchdowns, and that doesn't count successful plays like a 13-yard throw on 1st-and-15, and a 10-yard throw on 2nd-and-15 (both to Shannon Sharpe).  Plummer also had no sacks and no interceptions.

One last fun fact for the day.  Many of the hosannas for Jacksonville's late-season run have been for the offense, led by rookie quarterback Byron Leftwich and the shockingly uninjured Fred Taylor.  According to DVOA, however, the change has been far more substantial in the defense.  Over the past six games -- where Jacksonville has had four positive DVOA games and two that were negative but higher than -10% -- the offensive DVOA has been +3.5%.  But the defense has held opponents to a DVOA of -33%.  That's the second-best in the league over the last six weeks, only Baltimore has been better.  The Patriots have been #3 in defense over the past six weeks, by the way, so I would take that under this week. 

Due to contractual obligations, we hereby interrupt this close game with major implications where the team losing by a touchdown has the ball on the one-yard line to bring you a truck commercial.  Actually, we bring you some housekeeping of the statistics we use here at Football Outsiders.  I have some good news and some bad news.  Luckily, a lot more good news than bad news.

Bad news first.  My article on Building a Better Mousetrap discussed a new formula for determining WEIGHTED DVOA -- DVOA adjusted to make early-season games less important -- and fixed one math error but unfortunately added another.  This error resulted in the Jets being listed slightly too low, and Oakland being listed way too low.  See, and there I was getting happy about how our ratings finally were lining up with conventional wisdom.  Clearly hubris, alas.  The ratings will soon be fixed in that article, and the ratings below should be accurate.  By the way, Oakland had its worst performance of the year in Week 14. 

Now the good news.  In response to a number of requests, we've added separate pages with tables for offense and defense that separate each into rushing and passing with ranks.  The pages also give WEIGHTED DVOA for offense and defense separately.

You also may remember from the Mousetrap article that I mentioned two common complaints about the VOA ratings.  One complaint, that the ratings didn't take into account how teams have changed over the season, led to the new, stronger WEIGHTED DVOA formula.  The other complaint, that by adding together all plays equally we were missing the forest for the trees, leads to a new metric called FOREST INDEX.

(Yes, that's a stupid name and people are welcome to offer suggestions for changing it.  Heck, offer suggestions for what we can call WEIGHTED DVOA so that sounds better as well.)

What the heck is the FOREST INDEX?  Short version: FOREST INDEX uses a complicated formula that takes into account offense, defense, and special teams, as well as consistency, red zone performance, and performance in the second half when the score is close.  The result is a number that represents projected wins, adjusted for strength of schedule.  It does not take into account team improvement or decline over the course of the season.  Long version of explanation is here.  NOT for the mathematically faint of heart.

Which of these ratings is the best indication of team quality?  I've always believed in freedom of choice, so we'll provide all three and you can compare.  TOTAL DVOA correlates better with points, while the FOREST INDEX correlates better with wins.  A team's ability to score and prevent points is more stable than its ability to win games, so I think DVOA is a better indicator of a team's true quality, but FOREST INDEX is probably a better indication of why teams have won or lost games during the season.  WEIGHTED DVOA is probably the best indicator if the question is "who is likely to win next week," but season totals are probably more important in the long run.  After all, we don't pick playoff teams based on how they've played in the last two weeks, or Tennessee would be heading home while Jacksonville goes to the playoffs.

To make space, I've removed the FUTURE SCHEDULE rankings.  I figure with only three games remaining in the season, it's easy enough to look up your favorite team's last three opponents.

  • As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.
  • All numbers are adjusted for opponent quality except for NON-ADJ TOTAL VOA.

Here are the ratings through Week 14:
 


TEAM
TOTAL
DVOA
LAST
WEEK
NON-ADJ
TOTAL VOA
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
VOA
S.T.
RANK
1 KAN 31.0% 1 34.9% 11-2 26.7% 1 0.7% 18 5.0% 2
2 IND 25.4% 3 21.2% 10-3 24.1% 2 -0.6% 17 0.7% 12
3 STL 23.1% 4 28.0% 10-3 1.3% 12 -23.2% 2 -1.5% 27
4 SEA 21.4% 2 24.4% 8-5 19.4% 4 -1.3% 15 0.7% 11
5 TAM 20.8% 6 21.9% 6-7 0.1% 14 -22.7% 3 -2.0% 29
6 TEN 18.2% 5 18.9% 9-4 16.7% 5 -2.2% 13 -0.8% 23
7 DEN 16.2% 7 17.7% 8-5 6.4% 10 -9.4% 9 0.5% 15
8 NWE 16.1% 8 16.4% 11-2 -3.1% 17 -17.7% 5 1.4% 7
9 SFO 15.9% 9 18.4% 6-7 11.1% 6 -6.1% 10 -1.3% 26
10 MIN 13.4% 14 16.9% 8-5 20.7% 3 5.4% 24 -1.9% 28
11 BAL 8.5% 11 6.4% 8-5 -24.0% 31 -27.4% 1 5.1% 1
12 PHI 6.5% 16 9.4% 10-3 9.3% 8 4.6% 23 1.8% 6
13 GNB 5.5% 15 5.5% 7-6 3.0% 11 -2.0% 14 0.5% 14
14 PIT 4.5% 17 2.9% 5-8 -2.8% 16 -6.1% 11 1.3% 8
15 DAL 1.9% 10 7.8% 8-5 -8.7% 23 -11.2% 7 -0.6% 20
16 JAC 1.7% 22 1.5% 4-9 -4.7% 18 -10.8% 8 -4.4% 32
TEAM TOTAL
DVOA
LAST
WEEK
NON-ADJ
TOTAL VOA
W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
VOA
S.T.
RANK
17 BUF 0.5% 18 -2.2% 6-7 -12.0% 25 -12.0% 6 0.5% 16
18 MIA 0.2% 12 0.2% 8-5 -18.3% 28 -19.6% 4 -1.1% 25
19 OAK -4.5% 13 -9.1% 3-10 -6.4% 21 -0.8% 16 1.1% 9
20 NYJ -5.3% 20 -6.5% 5-8 10.9% 7 17.2% 31 1.0% 10
21 CAR -5.7% 19 -7.5% 8-5 -5.6% 19 2.7% 21 2.5% 3
22 CIN -6.5% 21 -0.1% 7-6 6.4% 9 10.0% 27 -3.0% 31
23 CLE -7.9% 23 -15.8% 4-9 -12.0% 24 -4.4% 12 -0.3% 19
24 WAS -8.5% 25 -6.6% 5-8 -0.5% 15 8.2% 26 0.3% 17
25 NYG -11.1% 24 -13.6% 4-9 -8.2% 22 2.3% 19 -0.6% 21
26 SDG -12.3% 26 -19.4% 3-10 1.0% 13 12.5% 29 -0.8% 24
27 NOR -17.9% 27 -17.1% 6-7 -5.8% 20 11.5% 28 -0.6% 22
28 CHI -21.6% 29 -19.1% 5-8 -19.5% 29 4.0% 22 1.8% 5
29 ATL -23.2% 30 -22.6% 3-10 -16.2% 27 6.9% 25 -0.2% 18
30 HOU -26.3% 28 -30.2% 5-8 -12.6% 26 15.7% 30 2.0% 4
31 DET -26.4% 31 -26.0% 4-9 -24.6% 32 2.3% 20 0.5% 13
32 ARI -50.2% 32 -53.8% 3-10 -21.9% 30 26.0% 32 -2.3% 30

  • FOREST INDEX takes into account offense, defense, and special teams, as well as consistency, red zone performance, and performance in the second half when the score is close, and then projects a number of wins adjusted to a league-average schedule.
  • WEIGHTED DVOA combines the team's DVOA performance from each game.  The past four weeks are each weighted at 100%, while each week before that is weighted progressively lower, beginning with Week 1 at 45%.
  • PAST SCHEDULE lists average DVOA of past opponents, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative).  This number will differ from the difference between DVOA and (non-adjusted) VOA because schedule strength is based on the opponent's total efficiency rating, while opponent adjustments to VOA take into account situations faced within each specific game.
  • 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
FOREST
INDEX
RANK WEIGHTED
DVOA
RANK PAST
SCHEDULE
RANK VARIANCE RANK
1 KAN 31.0% 11-2 10.0 2 29.3% 1 -1.8% 24 20.3% 21
2 IND 25.4% 10-3 10.1 1 22.4% 4 1.1% 14 19.0% 24
3 STL 23.1% 10-3 8.7 9 24.5% 2 -6.1% 31 29.0% 9
4 SEA 21.4% 8-5 9.7 3 15.9% 8 -5.2% 30 24.0% 17
5 TAM 20.8% 6-7 9.3 5 16.8% 6 -2.6% 25 28.6% 10
6 TEN 18.2% 9-4 8.8 8 23.2% 3 -0.6% 20 22.3% 18
7 DEN 16.2% 8-5 7.6 10 16.5% 7 1.3% 13 24.2% 16
8 NWE 16.1% 11-2 9.1 6 20.8% 5 0.8% 15 18.8% 25
9 SFO 15.9% 6-7 6.7 15 13.8% 9 -2.8% 26 42.4% 3
10 MIN 13.4% 8-5 6.7 14 7.7% 12 -2.9% 29 31.0% 7
11 BAL 8.5% 8-5 8.8 7 12.1% 10 2.3% 9 20.8% 19
12 PHI 6.5% 10-3 9.3 4 10.9% 11 -2.8% 27 17.2% 28
13 GNB 5.5% 7-6 6.6 17 6.7% 13 -0.1% 16 19.6% 22
14 PIT 4.5% 5-8 6.8 12 1.6% 15 3.9% 4 25.6% 13
15 DAL 1.9% 8-5 5.8 19 -1.1% 17 -6.1% 32 30.4% 8
16 JAC 1.7% 4-9 4.8 24 3.8% 14 3.2% 7 25.0% 14
TEAM TOTAL
DVOA
W-L FOREST
INDEX
RANK WEIGHTED
DVOA
RANK PAST
SCHEDULE
RANK VARIANCE RANK
17 BUF 0.5% 6-7 6.0 18 -3.6% 19 1.5% 12 44.1% 2
18 MIA 0.2% 8-5 7.4 11 0.5% 16 1.9% 10 31.8% 5
19 OAK -4.5% 3-10 5.3 21 -6.5% 22 3.9% 5 19.4% 23
20 NYJ -5.3% 5-8 6.7 13 -3.6% 18 1.6% 11 13.5% 31
21 CAR -5.7% 8-5 6.6 16 -4.0% 21 -1.7% 23 12.7% 32
22 CIN -6.5% 7-6 4.5 26 -4.0% 20 -0.5% 18 15.0% 30
23 CLE -7.9% 4-9 5.2 22 -13.7% 25 5.9% 1 34.1% 4
24 WAS -8.5% 5-8 4.6 25 -10.1% 23 -0.5% 19 24.8% 15
25 NYG -11.1% 4-9 4.4 28 -14.0% 26 3.3% 6 31.3% 6
26 SDG -12.3% 3-10 4.5 27 -12.1% 24 4.0% 3 20.3% 20
27 NOR -17.9% 6-7 5.6 20 -17.2% 28 -0.1% 17 18.5% 27
28 CHI -21.6% 5-8 5.1 23 -16.0% 27 -2.8% 28 18.6% 26
29 ATL -23.2% 3-10 3.9 31 -22.8% 29 -0.7% 21 28.5% 11
30 HOU -26.3% 5-8 4.4 29 -31.2% 31 2.8% 8 27.2% 12
31 DET -26.4% 4-9 4.4 30 -25.3% 30 -1.3% 22 16.8% 29
32 ARI -50.2% 3-10 2.5 32 -50.7% 32 4.4% 2 48.9% 1

PREVIOUS WEEKS:

Posted by: Aaron Schatz on 09 Dec 2003

comments