Writers of Pro Football Prospectus 2008

Most Recent FO Features

KhanSha1.jpg

» Futures: My Expansion Franchise

You've just been awarded an NFL expansion team and must build your personnel department. How would you do it? Matt Waldman takes on the exercise.

16 Dec 2003

Week 15 Team Efficiency Ratings, Featuring Notes on Running Backs

Plus Notes on Running Backs

by Aaron Schatz

Here are the team efficiency ratings after Week 15, 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: DAL (from #15 to #10), NOR (from #27 to #24)
Moving down: BAL (from #11 to #14)

As you can tell, not a lot of movement this week.  Week 15 was filled with a bunch of games that, for the most part, pitted good teams against bad teams, and, for the most part, they turned out as expected.  The teams atop of our rankings, Kansas City and Indianapolis, clobbered two of the worst teams in the league.  St. Louis and Seattle, ranked #4 and #5, played a very tight contest.  New England won yet another game by a moderate amount.

At this point, what the Patriots are doing is truly fascinating.  Sportswriters like to attribute the emergence of a dominant team to "the ability to win the close ones," often described as if it was less an athletic skill and more a sign of moral righteousness.  In general, however, the dominant teams in all sports get that way because they demolish the lesser competition and then win a fair share of games against more evenly matched opponents.  That's a good description of how Kansas City has played this year: They beat Buffalo by 33, Houston and Detroit by 28 each, Cleveland and Pittsburgh by 21 each.  But wins over Denver, Baltimore, and Green Bay were all very close.

As I mentioned last week, the Patriots are the exception to this rule.  They began the year with two games that we now know were colossal flukes: a loss to Buffalo and a win against a Philadelphia team that is playing far better now than it was then.  But since their Week 4 loss to Washington, the Patriots have been on a run of ridiculous consistency.  Against playoff teams, like Tennessee and Dallas, their performance is solid but unspectacular.  Against poor teams, like Houston, Cleveland, or the Giants, their performance is solid but unspectacular.  The lack of dominant games, plus the effect of the losses to Buffalo and Washington, keeps their Total VOA out of our top five, but they have to be the Super Bowl favorite because they seem to be able to handle anything at any time.

We use a statistic called VARIANCE that measures a team's consistency from game to game, with the higher ranking teams (Arizona, Buffalo) being the least consistent.  The Pats are #27, making them the sixth-most consistent team in the NFL this year.  But that includes all 15 weeks so far.  What if we only measured VARIANCE from Week 5 through the present?  As you can see, the list of the six most consistent teams since Week 5 shows the unnerving steadiness of the New England Patriots:

Most Consistent Teams by VARIANCE
Weeks 5-15 Season Total
NWE 4.2% CAR 12.5%
CHI 12.3% NYJ 12.5%
PHI 13.3% CIN 14.1%
CAR 14.3% DET 16.0%
IND 14.8% PHI 16.2%
HOU 15.6% NWE 17.4%

If the NFL teams were 32 college students, Indianapolis and Kansas City would be the kids whose transcripts featured almost all A's, but a couple of C's in classes they just couldn't get the hang of, while the Patriots would have nothing but B+'s and A-'s.  Following this metaphor, of course, Arizona is the kid who drank too much freshman year, failed everything, dropped out, took a couple dead end jobs, acquired a nasty smack habit, and currently lives in a cardboard box near the train tracks.

Seattle, I guess, would be the kid who was doing really well until he went to study for a year abroad and disappeared into the Amazon jungle.  You may be surprised to see Seattle still among our top five teams.  Part of that is because they were better early than they have been recently, but they really are a better team than people think.  Not to belittle the horrible Week 14 loss to the Vikings, but for all the talk about the Seattle road problems, the situation isn't that bad.  Losing by five points to another team in our top five is not really a failure, and the VOA system actually says Seattle barely, just barely, outplayed St. Louis (7.4% to -1.3%).  If back judge Greg Steed doesn't step in front of Bobby Engram, we might all be talking today about Seattle overcoming its road curse.  VOA also says that Seattle didn't play that badly in the losses to Cincinnati and Baltimore, and that they outplayed Washington despite losing in Week 10.  My prediction: If the Seahawks can get into the playoffs, and they face overrated Carolina in the first round, they will win.  Yes, on the road.

There were two major upsets in the NFL this week, of course, and it is probably worth mentioning each.  Our stats say that Minnesota outplayed Chicago on offense and defense, but lost the game on special teams.  The aborted punt by Eddie Johnson in the third quarter was a large part of that; turning the ball over on their own 20-yard line instead of kicking an average punt cost the Vikings three points worth of field position.  The Vikings also lost 2.6 points compared to average with a mean of zero yards on punt returns, and another 3.2 points from poor kickoffs and kickoff returns.

By the way, here's an example of why adjusting for opponent quality is important: Chicago's defense allowed the Vikings +22% VOA, while the Vikings allowed the Bears slightly less, +16% VOA.  But adjusting for the powerful Minnesota offense gives Chicago's defense +1% DVOA, while adjusting for the lame Chicago offense gives Minnesota's defense +34% DVOA.

The other shocking upset was Oakland over Baltimore.  Although Oakland outplayed Baltimore on offense and defense, special teams were a large part of this one as well.  I've been wondering all year why my ratings had Oakland so much higher than expected, but perhaps there was something to that.  I'll let you know when I find out what it is, because right now I've got no idea.

The sequence of plays that started that game point out to me the advantage of our statistics over the usual NFL numbers.  Baltimore started on its own 21-yard line, and went Jamal Lewis run for no yards, Anthony Wright incomplete, and Anthony Wright interception returned to the one-yard line.  On the next down, Zach Crockett went in for a one-yard score, 7-0 Oakland.  Now, who gets credit for that score?  The official NFL numbers will list the Oakland offense with one more touchdown but let's get real here.

Talking about Baltimore of course brings up another question: If Jamal Lewis is one of the best backs in the league, why does Baltimore rate so low in our DVOA ratings for rushing offense?  They are #27 this week, which is actually up from #30 last week.  This question came up in the Week 14 Open Thread, but I wanted to expand on the answer here.

ushing ratings include all rushing plays, from all running backs as well as quarterbacks and wide receivers.  The Baltimore quarterbacks don't have two left feet, they have about twelve left feet.  Each.  The three quarterbacks, combined, have a -99% rushing DVOA this year, and have been worth 15 points less than replacement-level quarterbacks when running with the ball.  Yes, I know, these aren't necessarily planned runs, many are scrambles, but unfortunately I have to deal with the limitations of current play-by-play logs which do not differentiate.

So, the pitiful quarterbacks are part of the explanation, but Lewis still isn't one of our highest-rated running backs even when you only judge him on his own.  Our running back ratings (only updated through Week 14 at the moment) list Lewis as the #20 running back in Points Above Replacement, with a below-average DVOA.  This means that, on average, he has been worse than an average running back, even though he leads the league in yardage and has one of the best yard-per-carry averages.  What the heck?

The answer lies not in the man's feet but in his hands.  For all the talk of Tiki Barber and his mystical fumbling powers, it is Jamal Lewis who leads the league in fumbles on rushing plays, with eight.  No other running back has seven; Barber and Ahman Green have six each.  Six of Jamal Lewis' fumbles have been recovered by the other team; each of those plays counts as a massive negative for both Lewis and the Ravens in general.

This fumbling problem is not a recent phenomenon.  Last year, Travis Henry led the league with 11 fumbles on rushing plays, but Lewis and Ricky Williams tied for second with seven fumbles each, and once again Lewis had six fumbles recovered by the opposition.  In 2000, Lewis was close to the top of the league with six fumbles in only 13 games started.

There is some indication that the ability to recover fumbles by a defense that causes fumbles is fairly random.  I have no idea if the ability to recover fumbles by the team that loses the ball is also random.  Right now, the VOA system treats a kept fumble as if it didn't even happen, while a lost fumble is a significant negative.  But, what if instead all fumbles were counted as equally negative?  On average, two of every three rushing fumbles is recovered by the defense.  That's not much different than Lewis and his ratio of six lost fumbles out of eight.

Contrast Lewis with the running backs that top our statistics.  Priest Holmes, on top of his zillions of yards, lost only one fumble last year and has lost only one fumble this year.  Clinton Portis has fumbled three times on rushing plays, but has only lost one.  LaDainian Tomlinson has fumbled once, and that one was recovered by San Diego.

Hey, speaking of Portis and Lewis, that brings me to the other issue I wanted to discuss.  In Week 2, Lewis set an NFL record by rushing for 295 yards against the Cleveland Browns.  In Week 14, Portis tied an NFL record with five touchdowns to go with 218 yards against the Kansas City Chiefs.  Which was the best running back performance in the NFL this year?

According to our PAR system -- that's Points Above Replacement, explained here -- the answer is neither.  The best running performance of 2003, through Week 15, came from Stephen Davis in Week 4 against Atlanta, or from Priest Holmes in Week 13 against San Diego.

Let's look at the top rushing performances of the year according to the NFL side-by-side with the top rushing performances of the year according to our stats here at Football Outsiders.*  DPAR is Defense-adjusted Points Above Replacement, while PAR is the same measurement before adjusting for defensive quality (these stats are further explained here):


TOP RUNNING BACKS BY YARDS, WEEKS 1-15
Player Wk vs. DPAR PAR Runs Yards
Lewis, Jamal 2 CLE 5.2 5.2 30 295
Portis, Clinton 14 KAN 5.3 6.5 22 218
Tomlinson, LaDainian 7 CLE 6.0 6.2 25 200
Green, Ahman 10 PHI 2.8 3.0 29 192
Tomlinson, LaDainian 4 OAK 5.4 6.4 27 190
Hambrick, Troy 15 WAS 3.3 4.2 34 187
McAllister, Deuce 12 PHI 4.3 5.7 19 184
Johnson, Rudi 10 HOU 5.0 5.9 43 182
Lewis, Jamal 14 CIN 2.6 3.7 30 180
Dunn, Warrick 10 NYG 3.2 4.9 25 178
Davis, Stephen 8 NOR 5.4 4.2 31 178
TOP RUNNING BACKS BY DPAR, WEEKS 1-15
Player Wk vs. DPAR PAR Runs Yards
Davis, Stephen 4 ATL 7.3 6.6 21 153
Tomlinson, LaDainian 7 CLE 6.0 6.2 25 200
Green, Ahman 4 CHI 5.6 5.9 18 174
Taylor, Fred 10 IND 5.5 5.5 28 152
Dunn, Warrick 10 NYG 5.4 4.9 25 178
Tomlinson, LaDainian 4 OAK 5.4 6.4 27 190
Portis, Clinton 14 KAN 5.3 6.5 22 218
Lewis, Jamal 2 CLE 5.2 5.2 30 297
Holmes, Priest 13 SDG 5.2 6.9 31 162
Holmes, Priest 15 DET 5.1 4.7 18 94
Green, Ahman 5 SEA 5.1 4.8 27 121

Yes, this list goes to eleven, because of ties for tenth place on both lists.

Stephen Davis' game in Week 4 is the only rushing game this year worth more than seven points over a replacement player.  Out of those 21 runs, only four were for two yards or less... and of those four, one was a touchdown on 1st-and-goal while another was a first down on 3rd-and-1.  The two remaining short runs were among only four Davis attempts that the system grades as below average.  He averaged 5.9 yards on first down, and added a 34-yard run on 2nd-and-10, a 10-yard run on 3rd-and-2, and a 29-yard run on 3rd-and-6 when the Panthers were trapped on their own 10-yard line.  Plus, the Falcons are an above-average run-defense team when the opposing offense is running out a lead.  (Obviously, most defenses allow fewer yards per carry when the opposing offense is running out the clock, but Atlanta offers even fewer than normal.)

While Davis' game rates the best once defense is taken into account, Priest Holmes' 162-yard, two TD game is the best without adjustment.  Holmes had more carries that rated below average, though he also had more carries overall.  But what made his day dominant was his performance on second downs.  Holmes ran on 15 second downs and got the first down or touchdown on 10 of them.  But San Diego is one of the poorest-ranked run defenses in the league, and the adjustments move Holmes' performances down from 6.9 PAR to 5.2 DPAR.

Perhaps more impressive was Holmes' performance this past week.  Despite less than 100 yards, it rates as one of the season's ten best.  Half of his 18 carries resulted in a first down or touchdown, and Detroit has had a very good run defense this year.

Portis' and Lewis' record-setting days don't make the top of our list for the same reason: long runs.  The VOA system judges all runs over 40 yards as equal.  Occasionally a defensive player will catch up to a back on one of these huge breakaway runs -- as one did on Lewis' 48-yard scamper in the second quarter of Week 2 -- but it is rare. Usually the difference between a 45-yard TD and a 75-yard TD is nothing more than field position.  Lewis got 193 of his yards that day on just three runs, but didn't do very well the rest of the time.  He had 13 different runs of two yards or less on either first-and-10 or second-and-8-or-more.  With Portis, 112 yards came on just two runs, but he was pretty good on all his other runs as well.  But our statistics seriously downgrade his performance due to the poor Kansas City defense.

Player Wk vs. DPAR PAR Runs Yards
Bettis, Jerome 10 ARI -6.0 -5.4 18 54
Davis, Stephen 11 WAS -5.9 -5.4 28 92
Green, Ahman 7 STL -4.9 -5.3 20 35
Williams, Ricky 8 SDG -4.8 -3.8 23 69
Barlow, Kevan 8 ARI -4.7 -4.4 9 19

For fun, here are the five worst rushing performances of the year.  Each of these includes one lost fumble.

At least Green had six catches for 62 yards that day... Bettis in Week 10 was below average on 13 of 18 runs with seven runs for two yards or less not counting the fumble.  Davis' game against his old team the 'Skins is a good example of how our system differs from conventional statistics.  28 carries for 92 yards sounds good, right?  Well, 21 of those 28 carries were below average.  12 of those runs were for two yards or less (though one of those was ok since it was 3rd-and-1).  Nearly every run left Carolina in a 2nd-and-long or 3rd-and-long situation.

  • 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 15:
 


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 32.8% 1 38.1% 12-2 31.6% 1 3.5% 19 4.7% 1
2 IND 28.2% 2 26.1% 11-3 25.0% 2 -2.6% 14 0.7% 12
3 TAM 23.4% 5 25.8% 7-7 2.8% 11 -22.3% 2 -1.7% 28
4 STL 21.6% 3 25.5% 11-3 1.4% 13 -21.2% 3 -1.0% 25
5 SEA 21.5% 4 23.4% 8-6 20.0% 3 -0.8% 16 0.8% 11
6 TEN 17.9% 6 18.8% 10-4 17.6% 5 -1.1% 15 -0.8% 23
7 DEN 17.4% 7 19.3% 9-5 7.6% 10 -9.5% 8 0.4% 14
8 NWE 16.3% 8 16.0% 12-2 -1.0% 14 -15.8% 5 1.5% 8
9 SFO 13.7% 9 16.6% 6-8 12.1% 7 -2.9% 13 -1.3% 26
10 DAL 9.1% 15 16.3% 9-5 -8.9% 23 -18.4% 4 -0.5% 21
11 MIN 8.6% 10 15.0% 8-6 19.0% 4 7.5% 23 -2.9% 31
12 PHI 8.5% 12 10.9% 11-3 13.5% 6 6.9% 22 1.8% 5
13 GNB 6.4% 13 8.2% 8-6 1.5% 12 -4.0% 11 0.9% 10
14 BAL 6.2% 11 4.9% 8-6 -24.3% 32 -26.3% 1 4.3% 2
15 PIT 2.5% 14 2.4% 5-9 -5.4% 20 -6.9% 10 1.0% 9
16 BUF 1.1% 17 -3.9% 6-8 -10.7% 24 -11.6% 7 0.2% 16
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 JAC 0.7% 16 0.0% 4-10 -2.9% 17 -8.6% 9 -4.9% 32
18 MIA -1.0% 18 -0.8% 8-6 -15.6% 27 -15.1% 6 -0.5% 22
19 OAK -3.2% 19 -7.4% 4-10 -5.0% 18 -0.1% 17 1.7% 7
20 CIN -5.0% 22 0.1% 8-6 9.8% 9 12.1% 29 -2.7% 30
21 NYJ -6.2% 20 -6.8% 6-8 10.5% 8 17.0% 30 0.3% 15
22 CAR -7.4% 21 -7.2% 9-5 -5.2% 19 4.6% 20 2.3% 4
23 CLE -8.9% 23 -18.0% 4-10 -11.6% 25 -3.1% 12 -0.3% 20
24 NOR -12.0% 27 -9.9% 7-7 -1.1% 15 11.3% 27 0.4% 13
25 SDG -13.4% 26 -21.1% 3-11 -1.2% 16 11.5% 28 -0.8% 24
26 WAS -15.1% 24 -14.5% 5-9 -6.6% 21 8.5% 26 0.0% 18
27 NYG -17.0% 25 -18.5% 4-10 -8.1% 22 7.5% 24 -1.4% 27
28 CHI -18.2% 28 -18.1% 6-8 -17.7% 29 2.9% 18 2.4% 3
29 ATL -25.2% 29 -27.5% 3-11 -17.1% 28 7.9% 25 -0.2% 19
30 DET -26.7% 31 -30.2% 4-10 -20.5% 30 6.3% 21 0.1% 17
31 HOU -27.8% 30 -33.1% 5-9 -12.4% 26 17.2% 31 1.8% 6
32 ARI -47.7% 32 -50.2% 3-11 -20.8% 31 24.3% 32 -2.6% 29

  • 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 32.8% 12-2 11.1 1 31.2% 1 -3.6% 26 19.4% 21
2 IND 28.2% 11-3 10.9 2 25.5% 2 -0.8% 20 19.0% 22
3 TAM 23.4% 7-7 10.4 3 18.8% 6 -4.1% 29 26.6% 12
4 STL 21.6% 11-3 9.2 7 22.9% 4 -5.0% 31 26.8% 11
5 SEA 21.5% 8-6 10.1 4 15.8% 8 -3.8% 27 22.2% 17
6 TEN 17.9% 10-4 9.1 8 23.4% 3 -0.4% 19 20.7% 19
7 DEN 17.4% 9-5 8.8 10 18.0% 7 0.5% 15 22.3% 16
8 NWE 16.3% 12-2 10.0 5 21.8% 5 0.4% 16 17.4% 27
9 SFO 13.7% 6-8 6.8 16 11.2% 10 -3.0% 25 39.4% 3
10 DAL 9.1% 9-5 6.6 18 7.9% 13 -7.4% 32 36.1% 4
11 MIN 8.6% 8-6 6.5 20 1.9% 15 -4.4% 30 30.3% 8
12 PHI 8.5% 11-3 10.0 6 14.4% 9 -2.5% 24 16.2% 28
13 GNB 6.4% 8-6 7.8 12 7.9% 12 -0.9% 21 18.2% 26
14 BAL 6.2% 8-6 8.9 9 9.9% 11 1.9% 9 19.8% 20
15 PIT 2.5% 5-9 7.4 13 -0.6% 17 3.3% 7 24.0% 14
16 BUF 1.1% 6-8 6.6 19 -2.2% 19 2.5% 8 40.8% 2
TEAM
TOTAL
DVOA
W-L
FOREST
INDEX
RANK WEIGHTED
DVOA
RANK PAST
SCHEDULE
RANK VARIANCE RANK
17 JAC 0.7% 4-10 5.0 25 2.3% 14 4.0% 5 23.1% 15
18 MIA -1.0% 8-6 7.9 11 -0.1% 16 1.8% 10 29.6% 9
19 OAK -3.2% 4-10 6.3 22 -4.9% 21 4.1% 4 18.3% 25
20 CIN -5.0% 8-6 5.5 24 -1.8% 18 0.2% 17 14.1% 30
21 NYJ -6.2% 6-8 7.1 15 -3.5% 20 1.5% 13 12.5% 31
22 CAR -7.4% 9-5 6.7 17 -6.2% 22 -3.9% 28 12.5% 32
23 CLE -8.9% 4-10 5.6 23 -15.5% 26 6.7% 1 31.6% 6
24 NOR -12.0% 7-7 7.2 14 -9.9% 23 -1.5% 22 21.8% 18
25 SDG -13.4% 3-11 4.8 27 -13.3% 25 4.3% 3 18.8% 23
26 WAS -15.1% 5-9 4.4 29 -18.4% 27 0.2% 18 30.3% 7
27 NYG -17.0% 4-10 4.1 31 -21.9% 28 1.6% 12 32.7% 5
28 CHI -18.2% 6-8 6.4 21 -11.2% 24 -1.8% 23 18.5% 24
29 ATL -25.2% 3-11 4.2 30 -25.1% 29 1.3% 14 27.4% 10
30 DET -26.7% 4-10 4.9 26 -25.9% 30 1.7% 11 16.0% 29
31 HOU -27.8% 5-9 4.6 28 -33.8% 31 4.6% 2 25.5% 13
32 ARI -47.7% 3-11 2.2 32 -47.8% 32 3.6% 6 46.2% 1


*Note: Our yardage totals might be slightly off from official NFL totals, due to small play-by-play log errors that get fixed in the offseason.

PREVIOUS WEEKS:

Posted by: Aaron Schatz on 16 Dec 2003

comments