Writers of Pro Football Prospectus 2008

Most Recent FO Features

RiceSid07.jpg

» Sidney Rice: What Could Have Been?

Sidney Rice has retired. Is he the most random single-season DYAR leader ever? One-year wonder? Injury prone? We offer a career retrospective for the second-best wide receiver named Rice in NFL history.

11 Nov 2003

Week 10 Team Efficiency Ratings

by Aaron Schatz

Here are the team efficiency ratings after Week 10, 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: CAR (from #20 to #14), TEN (from #12 to #7), PHI (from #25 to #20)
Moving down: MIA (from #14 to #22), NYG (from #9 to #13), HOU (from #24 to #28)

Did everybody enjoy Week 10?  Out of last week's top ten rated teams in DVOA, five lost -- all to teams that had been ranked #20 or lower -- and two were on bye.  As a result, Kansas City has opened a huge gap in total team efficiency between themselves and the rest of the NFL.  This week also gives me a great opportunity to discuss what stats can and cannot tell you, and what our goals are here.  Scroll down a bit for that discussion.

St. Louis moves up to the #2 slot after their strange victory over Baltimore.  The ESPN commentators spent the whole night saying, "Wow, who ever expected the St. Louis defense to so outshine the offense?"  Well, if you've been reading Football Outsiders, you expected it, since we had the Ram defense ranked #4 and the offense only #9.  After this week, the Rams actually move into the top spot among defenses.  Do you want to see what a defensive battle looks like in VOA ratings?  Here are the numbers for the Baltimore-St. Louis game -- and note that defense is better when it is negative:

St. Louis offense/Baltimore defense: -74%
Baltimore offense/St. Louis defense: -101%
Baltimore special teams: +15%
St. Louis special teams: +0%

That-101% means that Baltimore's offense averaged less than 0 "play success value" per play.  Considering that the only way to get a negative number is to lose three or more yards or turn the ball over, that's pretty amazing.  Even more amazing, that's not the worst offensive performance of the year -- Chicago in Week 1 against San Francisco was -110%.

Despite losses, Tampa Bay, Indianapolis, and Seattle all remain in the top five.  Yes, we still have Tampa Bay this high despite another loss to Carolina.  I'll be writing a lot more about these two teams in the NFC midseason report article that will be up Thursday or Friday.  Like this week's AFC midseason report, the article will look at where each team stands at the midpoint of the season, including breakdowns of offense, defense, and special teams. 

The midseason report inspired a couple of minor changes in the VOA formula.  First, the variables that adjust rushing plays in the fourth quarter of non-close games have been adjusted.  Second, the field goal numbers used to compare kickers to average have been adjusted because so many more field goals are successful this year than last.  These changes don't move any teams more than a couple percentage points either way, but the team that gets the biggest boost from the fix in fourth quarter rushing is the Tennessee Titans.

I took some flak from Titans fans for having their team ranked #23 in both overall and rush defense after Week 9.  The adjustments did raise the Titans' defensive ratings a couple points, but even after these adjustments, and even after the shellacking of Miami on Sunday, Tennessee's defense is only ranked #15, and the rush defense is down at #24.  Yet the NFL says that the Titans lead the league in rush defense?  How can our ratings be so much different from the NFL's?

The answer lies in the difference between total stats and rate stats.  Yes, the Titans have given up the fewest rushing yards.  They've also faced the fewest rushing attempts -- by a ridiculous margin compared to the rest of the NFL.  But they are one of the top five teams in pass attempts faced.  Here are the NFL's top defenses in fewest rushing plays faced and most passing plays faced, along with the ratio of pass plays to rush plays.


Top 5 Fewest Rush Plays Faced Top 5 Most Pass Plays Faced Top 5 Ratio Passes to Runs Faced
TEN 165 NWE 372 TEN 2.12
SFO 208 GNB 354 NWE 1.69
DEN 209 KAN 352 MIN 1.58
DAL 211 TEN 349 SFO 1.56
MIN 214 MIN 339 GNB 1.53

The league average is 1.28 passes for each rush, so that 2.12 ratio against Tennessee is just ridiculous.  So Titan opponents are spending almost all their time either picking on the Titans secondary or trying to come back from large deficits. But when teams do run against the Titans, they are doing pretty well.

I actually spent the entire Miami-Tennessee game trying to watch for rushing plays against the Titans, but not because I wanted to see the Titans' defense.  I wanted to see watch all of Ricky Williams' runs to see if his troubles this year were more his fault or the fault of his offensive line.  I had also taken Ricky Williams in the loser league, our "pick the worst players" weekly draft, and it turned out to be a good idea.  He hardly ran, and when he did he went nowhere.

From watching each Ricky run, it seems to be that the offensive line is the larger problem.  They're saying that Ricky isn't hitting the holes strongly this year, but that would require actual holes.  The Miami offensive line just looked pathetic.  The play that stands out to me came in the third quarter.  On 2nd-and-3 from Miami's 44, they gave the ball to Ricky for a run behind the right guard, and was tackled for a loss of about a yard from behind.  Somehow Keith Bullock of the Titans managed to get around his blocker and move all the way behind Ricky before Ricky even hit the line of scrimmage.  I'm not talking side tackle here, he was completely in back of him.  I didn't think that was possible without one of those Star Trek matter transporters.

I don't want to say that Ricky was putting great moves on when he had the chance, however.  On receptions, without a terrible offensive line keeping him back, I don't really remember him once seriously juking out a Titans defender.  Compare that to monday night, when Ahman Green and even Duce Staley were dragging two or three defenders along for extra yardage.

The play that epitomized the Tennessee-Miami game on the other side of the ball came near the end of the first half after the Titans pounced on a Miami fumble at the Dolphin 24.  On first down, Miami brought in the dime package; after all, Eddie George blows and he'll never hurt you, certainly not with only a minute left in the half.  Yet, despite six defensive backs, Steve McNair managed to find Justin McCareins completely open in the middle of the field, around the 18-yard line.  When I say, completely open, I mean completely; there were no Miami players within 15 yards of him and he wasn't even touched by a defender until he got to about the 2-yard line.  He scored anyway.  If I were TMQ, this is where I would have written in my notebook "game over."  I mean, what the heck were those six defensive backs doing, if not covering McCareins -- playing pinochle or something?

Anyway, despite the fact their defense is still rated fourth in the league, Miami's total VOA rating collapses this week.  They were #14 last week, +5.1%, and now they are #22, -7.1%.

As I have said, this week's topsy-turvy results give me a good chance to address some issues about what the statistics of this site are trying to accomplish.  The goal here is intelligent football analysis, whether that's more subjective (Scramble for the Ball) or objective (the stats).  But when it comes to the stats, it is important to realize a few things (and I've said some of these before, so I apologize for repetition):

1) Don't use VOA ratings to pick your office pool.  I've actually gotten emails that say, "Hey, I'm going to test your site by seeing if the ratings can pick games each week."  No, please, no.  These ratings are meant to try to figure out which teams are better over the long haul, and why they are better over the long haul, but not which teams are better this week.  They don't take into account home-field advantage, or injuries, or which team is coming off an emotional win and is in for a letdown, or which teams match up particularly well with which other teams, or which teams have had fluky wins or losses in the past, or just plain luck and the weirdness that leads to Atlanta demolishing a far superior Giants team.  Trust me, whenever I try to pick games based on these ratings, it turns out I should have gone with my gut, and whenever I try to pick based on my gut, it turns out I should have gone with the ratings.  If you want pick advice, read Scramble -- Ian and Al are much better at picking specific games than I am.  Well, except this week, when we all stunk up the joint.

2) Statistics can tell you what happened in the past, and what is likely to happen in the future, but not what will happen in the future.  If any weekend showed this, it is this past weekend.  Or did your power ratings have Jacksonville above Indianapolis, Atlanta above the Giants, and San Diego ahead of Minnesota?  Once again, I'll paraphrase Keith Woolner from Baseball Prospectus: our statistics aren't meant to be predictive, but are instead evaluating the value or quality of a player or team's performance in the past.  Our goal is to remove the bias inherent in standard NFL statistics, but I cannot tell you when a backup quarterback will suddenly turn into Tom Brady or a major player will get injured or a team will just plain give up in midseason.  To give another example, we know now that the 31-0 Buffalo victory over New England in Week 1 was a massive fluke but for a long time it caused Buffalo to be ranked higher than the Patriots in VOA.  This doesn't mean VOA doesn't work, even though since that time the Patriots have been the far superior team.  Anyway, part of the goal of the site is to determine which statistics are meaningful and which ones aren't, which trends are likely to continue and which are statistical noise created by small sample size, and which teams are likely to continue their success and which are enjoying success based on luck and will become the 2002 Chicago Bears next year.

I also plan on gradually honing the VOA formula based on more and more data to make it as true to real football performance as possible.  Speaking of which, to digress, Oakland is still rated a lot higher than either you, the readers, or I, the stat geek, would expect.  I'm not sure what's going on there, but I know that looks strange and I'm planning to figure out where the formula thinks they do not completely stink.  When I figure it out, I'll let you know.  They are the only team whose rating seems completely out of place.

3) Statistics tell the story, they aren't the story: Once again, let me quote from a Prospectus guy, this time John Hollander of Pro Basketball Prospectus.  In the Introduction to this year's book (which you must go buy right now if you like the NBA), John writes, "In the end, this book isn't about statistics... It's about analyzing the game to produce a better understanding of how teams are winning and losing, and what they can do to improve."  The same goes for the statistics and writing on Football Outsiders.  Bill James used to say that he saw himself as a writer who used stats, not a statistician who wrote a little, and the same goes for me.  Our goal here isn't just to churn out DVOA numbers every week; it's to figure out what those numbers mean, which ones are important, and what they say about the NFL -- and then to explain it to you the readers in as much math-free English as possible.  I'm always open to suggestions about how to make the site more math-o-phobe friendly.  Just remember that sometimes its easier to run tables of numbers when dealing with breaking down 16 games a week, doing my actual job, and spending time with my wife and newborn daughter.  Unlimited time (and roto baseball) is why God made the offseason.

4) Plays don't necessarily suffer from small sample size, but wins do.  It's important to know why teams win games, but remember that over an entire season there are only 16 of them.  All it takes is one random big play to turn a win into a loss or vice versa no matter how well the team played the rest of the game.  But there are hundreds of runs, passes, and kicks over an entire season that can be analyzed.  Statistics that depend on team performance on a play-by-play basis are based on much more significant data than a team's simple winning percentage.  Obviously, sometimes teams that play very well are losing for a reason (like Tampa, who seem to play amazing for about 55 minutes a game), but sometimes there is a lot of luck involved as well.

5) We're new at this.  People have been doing statistical analysis of baseball for something like 25 years.  John Hollander has been doing his basketball analysis for six years.  Most NFL power ratings systems, like Jeff Sagarin's or Ryan Early's, have been developed over the last few years.  On the other hand, I started playing with numbers in December, developed the first run at the VOA system in February, and launched this site in July.  So believe me, this is the very early stages of perfecting all the statistics we do.  As I say often, we're in the Bill James mimeographing Abstracts in his garage in 1978 stage right now, but thanks to the Internet we can share the work with many more readers.

OK, let's get to this week's VOA ratings.  Individual position stats will be updated Wednesday or Thursday.

  • Opponent adjustments are now at full strength and will be for the rest of the season.
  • 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 10:


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 41.9% 1 51.3% 9-0 19.4% 3 -14.9% 6 7.5% 1
2 STL 30.9% 4 31.8% 6-3 4.0% 12 -28.2% 1 -1.4% 27
3 TAM 27.7% 3 29.0% 4-5 6.2% 10 -23.0% 3 -1.5% 29
4 IND 25.7% 2 24.0% 7-2 19.1% 4 -5.1% 11 1.5% 12
5 SEA 23.9% 5 27.6% 6-3 18.7% 5 -4.5% 12 0.6% 19
6 SFO 22.7% 7 20.6% 4-5 13.0% 6 -10.4% 7 -0.7% 22
7 TEN 22.2% 12 24.6% 7-2 21.4% 2 -1.6% 15 -0.7% 23
8 DAL 18.3% 8 28.3% 7-2 -6.5% 21 -23.9% 2 0.9% 16
9 MIN 16.2% 6 25.8% 6-3 25.7% 1 8.1% 24 -1.5% 28
10 NWE 8.6% 10 13.3% 7-2 -0.4% 15 -8.1% 8 0.9% 15
11 DEN 7.7% 13 6.3% 5-4 2.1% 13 -3.8% 13 1.8% 8
12 GNB 4.7% 11 -0.9% 4-5 13.0% 7 8.4% 25 0.1% 20
13 NYG 3.3% 9 1.0% 4-5 -2.5% 16 -7.1% 10 -1.3% 26
14 CAR 0.4% 20 -5.1% 7-2 -1.5% 18 3.1% 19 5.0% 2
15 BAL -1.5% 19 -0.9% 5-4 -25.6% 31 -19.5% 5 4.6% 3
16 PIT -2.2% 16 -8.0% 3-6 -4.8% 20 -1.4% 16 1.2% 13
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 OAK -3.2% 17 -4.0% 2-7 -1.0% 17 4.0% 20 1.8% 7
18 CLE -5.2% 15 -16.6% 3-6 -7.9% 23 -3.2% 14 -0.5% 21
19 NYJ -6.5% 21 -3.3% 3-6 10.0% 8 18.8% 32 2.3% 6
20 PHI -6.7% 25 -6.5% 6-3 -10.3% 25 -0.9% 17 2.7% 5
21 BUF -6.9% 18 -13.2% 4-5 -14.9% 28 -7.3% 9 0.8% 18
22 MIA -7.1% 14 -4.8% 5-4 -25.8% 29 -19.8% 4 -1.1% 24
23 WAS -7.6% 22 -9.2% 4-5 5.1% 11 13.5% 27 0.8% 17
24 NOR -11.3% 23 -11.2% 4-5 -6.4% 22 6.4% 22 1.5% 11
25 CIN -12.9% 26 -1.2% 4-5 0.0% 14 8.9% 26 -4.0% 32
26 JAC -13.1% 28 -11.8% 2-7 -8.3% 24 1.0% 18 -3.8% 31
27 SDG -13.6% 27 -17.5% 2-7 6.3% 9 18.7% 31 -1.2% 25
28 HOU -15.7% 24 -22.9% 3-6 -3.0% 19 15.9% 28 3.2% 4
29 ATL -27.3% 32 -34.7% 2-7 -11.6% 26 16.6% 29 1.0% 14
30 DET -30.2% 30 -27.2% 3-6 -25.7% 32 6.1% 21 1.6% 10
31 CHI -30.9% 29 -28.0% 3-6 -24.9% 30 7.8% 23 1.7% 9
32 ARI -33.0% 31 -37.9% 3-6 -12.8% 27 16.8% 30 -3.5% 30

  • PAST SCHEDULE lists average DVOA of past opponents, while FUTURE SCHEDULE lists average DVOA of upcoming opponents.  Teams are ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative).
  • The PAST SCHEDULE 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 the situations faced within each specific game.
  • 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 5% lower, beginning with Week 1 at 70%.
  • 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
PAST
SCHEDULE
RANK FUTURE
SCHEDULE
RANK WEIGHTED
DVOA
RANK VARIANCE RANK
1 KAN 41.9% 9-0 -4.0% 28 -9.6% 30 44.8% 1 21.7% 15
2 STL 30.9% 6-3 1.5% 16 -10.3% 31 34.5% 2 21.4% 17
3 TAM 27.7% 4-5 1.6% 14 -5.3% 28 28.3% 3 34.8% 4
4 IND 25.7% 7-2 -1.7% 22 -2.6% 22 26.9% 4 23.2% 14
5 SEA 23.9% 6-3 -4.4% 30 0.0% 16 22.1% 7 18.6% 23
6 SFO 22.7% 4-5 3.4% 7 -4.0% 26 22.7% 6 44.2% 2
7 TEN 22.2% 7-2 -2.0% 24 -2.3% 21 26.3% 5 29.4% 9
8 DAL 18.3% 7-2 -9.7% 32 -2.9% 25 20.5% 8 29.7% 8
9 MIN 16.2% 6-3 -6.6% 31 -0.1% 17 15.3% 9 26.9% 10
10 NWE 8.6% 7-2 -0.8% 21 -0.8% 20 11.0% 10 23.8% 13
11 DEN 7.7% 5-4 0.3% 19 2.8% 10 6.6% 12 14.5% 27
12 GNB 4.7% 4-5 3.1% 9 -2.8% 23 7.8% 11 19.7% 22
13 NYG 3.3% 4-5 2.1% 12 2.0% 13 2.2% 14 32.6% 5
14 CAR 0.4% 7-2 2.7% 11 -11.9% 32 0.2% 15 13.0% 29
15 BAL -1.5% 5-4 0.0% 20 2.3% 12 3.5% 13 21.0% 19
16 PIT -2.2% 3-6 8.2% 1 -2.9% 24 -3.2% 16 24.7% 12
TEAM
TOTAL
DVOA
W-L
PAST
SCHEDULE
RANK FUTURE
SCHEDULE
RANK WEIGHTED
DVOA
RANK VARIANCE RANK
17 OAK -3.2% 2-7 -3.1% 26 7.6% 5 -4.2% 17 10.3% 31
18 CLE -5.2% 3-6 7.3% 3 1.8% 14 -6.7% 21 20.7% 20
19 NYJ -6.5% 3-6 -1.9% 23 3.9% 9 -6.0% 20 14.3% 28
20 PHI -6.7% 6-3 1.6% 15 2.7% 11 -5.8% 19 9.0% 32
21 BUF -6.9% 4-5 1.7% 13 4.4% 8 -9.7% 23 61.5% 1
22 MIA -7.1% 5-4 0.4% 18 -0.3% 19 -5.2% 18 31.6% 7
23 WAS -7.6% 4-5 3.8% 6 -4.9% 27 -9.1% 22 32.6% 6
24 NOR -11.3% 4-5 2.9% 10 -7.7% 29 -11.3% 24 21.1% 18
25 CIN -12.9% 4-5 -4.0% 29 10.4% 2 -12.6% 26 14.6% 26
26 JAC -13.1% 2-7 3.2% 8 -0.3% 18 -14.1% 27 18.3% 24
27 SDG -13.6% 2-7 0.5% 17 0.8% 15 -12.3% 25 11.0% 30
28 HOU -15.7% 3-6 4.4% 5 5.3% 6 -18.3% 28 21.6% 16
29 ATL -27.3% 2-7 7.9% 2 5.1% 7 -30.0% 29 37.2% 3
30 DET -30.2% 3-6 -3.2% 27 14.9% 1 -32.0% 31 19.8% 21
31 CHI -30.9% 3-6 -2.3% 25 8.7% 3 -30.5% 30 15.2% 25
32 ARI -33.0% 3-6 6.0% 4 8.3% 4 -34.4% 32 24.8% 11

PREVIOUS WEEKS:

Posted by: Aaron Schatz on 11 Nov 2003

comments