Writers of Pro Football Prospectus 2008

Most Recent FO Features

CoxMar15.jpg

» Varsity Numbers: Honing in

Bill Connelly again looks at which college football teams the F/+ ratings are sure about, and which teams remain a mystery (led by Appalachian State).

23 Nov 2007

Players Have Opponent Adjustments Too

by Stuart Fraser

Opponent adjustments are one of the signature concepts at Football Outsiders, and frequently one of the most important in explaining the difference between FO metrics and more conventional statistics or win-loss records. The DVOA commentary about Team X, which is 3-6 but nonetheless above 6-3 team Y, will often refer to the relative strengths of schedule.

But what about the individual players? Quick Reads concentrates on the week that's just happened, so there's rarely a chance to go back and revisit the schedule adjustments for offensive skill position players over an entire season. With that in mind, I went through the archives and looked for the quarterbacks and running backs with the strongest opponent adjustments since 1996 -- players who were nowhere near as good (or as bad) as their conventional statistics suggested.

Before we head off into examining the figures, a brief pause to explain some terminology I'll be using in the rest of the article. I've defined an opponent adjustment as negative or down if that player's DPAR is lower than their unadjusted PAR. This means the player played against a weaker-than-average (substantially weaker than average, if he's appearing in this article) schedule in terms of the ability of the defenses to stop his particular brand of offense. Conversely, an opponent adjustment is positive or up if DPAR is higher than PAR, and implies an RB playing against good run defenses or a QB throwing into the teeth of a good pass defense.

I started this article because I was surprised at the disconnect between the popular perception of Ben Roethlisberger's 2007 season and his struggling-to-crack-the-top-10 rank in DPAR. It turns out there is a current player on the list of quarterbacks with the largest penalties from opponent adjustments -- but it isn't Roethlisberger. Through Week 11, Seattle's Matt Hasselbeck has 71.5 PAR but only 53.2 DPAR. Even though the season isn't finished yet, that difference of 18.3 is good for ninth place on the list of negative opponent adjustments. The Seahawks' future schedule has an even lower average DVOA than their 32nd-ranked past schedule, so Hasselbeck may end up the most "overrated" (by unadjusted statistics) quarterback since DVOA stats begin in 1996.

Player Team Year DPAR PAR Adjustment
Kurt Warner STL 1999 102.1 129.9 -27.9
Steve Beuerlein CAR 2000 12.0 38.4 -26.4
Jeff Garcia SF 2000 97.8 119.6 -21.8
Jake Delhomme CAR 2004 49.4 71.1 -21.7
Billy Volek TEN 2004 5.0 24.8 -19.8
Peyton Manning IND 2004 170.1 189.2 -19.1
David Carr CAR 2004 27.5 46.3 -18.8
Kurt Warner STL 2000 48.0 66.4 -18.4
Matt Hasselbeck SEA 2007 53.2 71.5 -18.3
Peyton Manning IND 2005 112.5 130.0 -17.5

The collection of cupcakes faced by the 1999 Rams is fairly well known, so seeing Warner atop this list isn't particularly surprising. Warner didn't face a pass defense with a negative DVOA between Baltimore on opening Sunday and Philadelphia in Week 17, and got two divisional matchups with the 49ers, whose 43.5% DVOA pass defense eclipses even this year's Jason David-inspired Saints as the worst pass defense in DVOA history.

Warner also shares (with Peyton Manning) the dubious distinction of appearing in this list twice, which has a lot to do with the sustained ineptitude of their respective divisions. The NFC West never seems to have more than one good team in it; the 2002 division realignment, though it is hard to believe, actually strengthened the division, bringing in the Seahawks and removing the listless Falcons and Saints. The AFC South is now emerging from the doldrums but for some time featured the expansion Texans and the Titans journeying through cap purgatory. Highlights of the 2000 Carolina schedule, from quarterback Steve Beuerlein's point of view, included Minnesota (31st in pass defense DVOA), Seattle (27th), and divisional opponents San Francisco (29th), St. Louis (25th), and Atlanta (20th). The best pass defense the Panthers faced was that of the Washington Redskins, whose -20.7% DVOA was only good for 7th in a year which, like 2004, was just uneven, not bad, as Jets fans can attest:

Player Team Year DPAR PAR Adjustment
Aaron Brooks NO 2002 58.5 32.4 +26.1
Vinny Testaverde NYJ 2000 39.1 14.7 +26.4
Carson Palmer CIN 2004 39.0 15.7 +23.4
A.J. Feeley MIA 2004 -11.6 -33.0 +21.4
Brett Farve GB 2005 40.5 19.6 +20.9
Kyle Boller BAL 2004 23.1 3.1 +20.0
Tom Brady NE 2004 113.4 95.6 +17.8
Kurt Warner STL 2002 -1.5 -18.7 +17.2
Chad Pennington NYJ 2004 77.2 60.5 +16.7
Jeff Garcia CLE 2004 11.2 -4.3 +15.5

Once again the list is dominated by the 2004 season, with a single interloper from 2000. 2004 had a strange distribution of pass defenses. Six teams had a passing DVOA worse than 20%, twice as many as in the previous year. It's not unreasonable to assume that the league's stricter enforcement of illegal contact rules starting in that season affected teams disproportionately. Whatever the reason, two of these six teams (Oakland, Kansas City) were in the AFC West, and another two (St. Louis, San Francisco) were in the NFC West. These divisions faced each other in interconference play, leading to some fairly lopsided schedules. On the other side of the country the AFC North and East, featuring five of the best six pass defenses in that year, were beating up on each other, which places Browns and Bengals quarterbacks on the upwardly-adjusted list.

Back in 2000, Testaverde's Jets were facing all the good teams that Beuerlein's Panthers avoided -- Miami (first with an excellent -31.8% pass defense DVOA) and Buffalo (eighth) in their division, with Tampa Bay (third), Detroit (fifth) and Baltimore (sixth) also on the schedule. The worst pass defense faced by the Jets was the 23rd-ranked Colts, who would have fallen right in the middle of a list of the bad pass defenses on Carolina's schedule. Though Testaverde couldn't do anything with the return to a more normal schedule the next year (and consequently lost his job to Chad Pennington), three of these quarterbacks showed significant "improvement" in conventional terms their next season. Brooks had the best year of his career, and Palmer and Favre had "breakout" and "comeback" years in their respective next seasons. Favre's "much-improved" first season under McCarthy had very little to do with coaching -- his DPAR stayed almost constant between '05 and '06, but the opposition changed. Palmer combined weaker opponents with improved play coming off his rookie year and ended up with a Pro Bowl berth as a sophomore.

When it comes to running backs, the adjustments are smaller because we're dealing with fewer plays, and DPAR is a counting stat. Unlike the somewhat motley crew that makes up the list of quarterbacks who benefited most from weak opponents, the comparable running back list is rather more star-studded -- probably because the more carries you get, the bigger the adjustments end up being. First, the players with easy schedules:

Player Team Year DPAR PAR Adjustment
Emmitt Smith DAL 1998 25.2 37.8 -12.7
LaDanian Tomlinson SD 2002 25.3 35.2 -9.9
Curtis Martin NYJ 2001 23.0 32.5 -9.5
Ricky Williams MIA 2002 33.1 42.1 -9.0
Clinton Portis DEN 2003 37.0 45.8 -8.8
Shaun Alexander SEA 2005 56.3 64.7 -8.4
Tim Biakabutuka CAR 2000 -8.9 -0.5 -8.4
Clinton Portis DEN 2002 45.2 53.5 -8.3
Travis Henry BUF 2002 7.7 15.6 -7.9
Marshall Faulk STL 2000 60.0 67.8 -7.8

At first glance there's nothing terribly striking about the 2002 season (from which four backs appear on this table), but the schedule pitted the AFC West (home to San Diego and Denver) and the AFC East (home to Miami and Buffalo) against each other, allowing the 18th-ranked Chargers, the 24th-ranked Bills, the 26th-ranked Patriots, the 29th-ranked Chiefs and the 31st-ranked Jets to, ahem, showcase their run defenses against each other. I'm sure LaDanian Tomlinson and Clinton Portis approved wholeheartedly.

Three of these players crossed the dreaded 370-carry barrier (Tomlinson 2002, Williams 2002, Alexander 2005), and I suppose it makes sense that a featured running back might well be overused when facing a schedule packed with weak opposition. Not only is it smart to run against a bad run defense, but teams playing a weak schedule are going to be ahead more often than usual, and everybody reading this should know by now that you run when you win.

The table of backs who found themselves faced with distinctly above-average run defenses also contains a 370-carry season, which is somewhat harder to rationalize. It probably has more to do with the team than the opponents though; Eddie George's Titans didn't have much in the way of a passing game that season, but their excellent defense made it worth running George into the ground (at least from Jeff Fisher's point of view). This list is also headed by a back who had to come back from overuse:

Player Team Year DPAR PAR Adjustment
Jerome Bettis PIT 2000 38.9 25.8 +13.1
Marshall Faulk IND 1998 22.5 11.5 +11.0
Deuce McAllister NO 2002 24.7 16.0 +8.7
Curtis Martin NYJ 2003 27.3 18.7 +8.6
Travis Prentice CLE 2000 3.5 -4.8 +8.3
Lee Suggs CLE 2002 1.8 -6.5 +8.3
Jamal Lewis CLE 2007 2.6 -5.7 +8.3
Eddie George TEN 1998 5.2 -3.0 +8.2
Eddie George TEN 2000 20.9 12.7 +8.2
Warwick Dunn ATL 2006 12.5 5.0 +7.5

Here's another player from 2007, but don't expect Jamal Lewis on the list at season's end, since the Browns have one of the easiest schedules going forward. (If we had written this article a week ago, Rudi Johnson would have made the list instead.)

Astute readers will have noticed that the 2000 season shows up in numbers yet again -- lacking the concentrated wackiness of 2004's passing statistics perhaps, but more consistent across the board. The main culprit here is the six-team AFC Central, containing two great defenses in Baltimore and Tennessee, a pretty good one in Pittsburgh, and a solid run defense in Jacksonville, all wailing on each other (and on the hapless Browns and Bengals). To make matters worse for Bettis in particular, he was pounding his head against not only the Ravens (all-time leaders in run defense DVOA, at -41.6%), Titans (2nd in run defense), and Jaguars (8th), but also the Chargers (3rd), Giants (4th), Raiders (5th), Eagles (9th), and the Jets (10th). At least the Steelers (themselves ranked sixth against the run), gave the Bus plenty of chance to practice the skills needed to move the ball on a top-10 run defense. The only team Bettis was missing from the complete top 10 was the Bills. Another reason this adjustment blows the competition away is that the Steelers showed no inclination to stop driving the Bus into these defenses. Bettis carried the ball 354 times, although, given that Kordell Stewart was only completing 52.2 percent of his passes at the time, this might have been a wise decision. Eddie George had 403 carries, but his adjustment is lower because of the arcane 31-team scheduling mechanic. The Titans didn't face San Diego, Oakland or the Jets, though they did encounter Buffalo.

Under the circumstances, the pedestrian-looking 3.8 yards per carry achieved was probably one of Bettis's best seasons -- something those who will consider his Hall of Fame candidacy should consider. Opposition matters.

Posted by: Stuart Fraser on 23 Nov 2007

17 comments, Last at 27 Nov 2007, 4:42pm by Herm?

Comments

1
by Sophandros (not verified) :: Fri, 11/23/2007 - 6:17pm

It's very interesting that the play-caller for Favre's "comeback" season and Aaron Brooks' best season is the same person: Mike McCarthy.

Maybe coaching has just a little to do with it, since Brooks has pretty much been a stinking pile of dung since he's been separated from McCarthy.

2
by Mungo (not verified) :: Fri, 11/23/2007 - 6:56pm

Maybe it is McCarthy's coaching. But then he was Favre's QB coach in 1999, which was a pretty crappy year for him, and in his prime (57% complete, 23 picks....er, gunslings).

3
by Kevo (not verified) :: Fri, 11/23/2007 - 7:03pm

Maybe it's just me, but it seems like a percentage increase/decrease would give a more effective look, especially because, as you said, the more carries you get, the bigger the adjustments end up being.

4
by V (not verified) :: Fri, 11/23/2007 - 10:17pm

The '99 49ers had the worst pass defense in dvoa history? I wonder how the '95 Falcons would fare, they allowed the most yards and passing first downs of any team in NFL history. When would the 1995 DVOA stats be completed?

5
by MarkV (not verified) :: Sat, 11/24/2007 - 2:36am

not to nit-pick, but the 2004 AFC West did not play the NFC West. They played in 2002 and 2006. In 2004 AFC West played NFC South and NFC West played AFC North

6
by Ashley Tate (not verified) :: Sat, 11/24/2007 - 2:53am

Warwick Dunn ATL 2006 12.5 5.0 +7.5

Ah, yes. The estimable "Dionne" Warwick Dunn. Not just the sweetest feet, but the smoothest voice in the NFL in 2006!

7
by Terry (not verified) :: Sat, 11/24/2007 - 7:48am

Can someone quickly tell me if DPAR and DVOA factor out the player/team themself/ves? For example, is the DVOA of New England this year, or Kurt Warner circa 2000, decreased because of how much they dominate(d) the competition and thus rendering their opponents seeming weaker? Or does it only rate the performance of the competition against opponents who are not they themselves?

8
by MarkV (not verified) :: Sat, 11/24/2007 - 12:32pm

RE 7:
I think it includes all games, because especially with divisional opponents, opponents play is important against the actual team.

I believe that Aaron uses a few passes of the defensive-offensive adjustments, to make sure that the factor you listed is minimized as much as possible.

9
by Archimedes Owl (not verified) :: Sat, 11/24/2007 - 2:51pm

7: Well, if this did happen, it wouldn't affect comparisons so much because every good team would affect their opposition in the same way. It wouldn't affect some teams more than others really.

But, I remember that they decided to do adjustments of the opposition of the opposition. I don't know how far they ended up taking this, but I'd imagine that any further would show only slight changes.

10
by Bobman (not verified) :: Sat, 11/24/2007 - 3:22pm

Stuart, FYI, the first table has David Car on Car in 2004. Should be Hou. I liked this piece quite a bit.

Archimedes, I am just re-reading Once and Future King and your name popped off the page. It's been 15 years since I last read it, so I giuess I can be forgiven for not getting the reference before. Say hi to the Wart for me.

11
by Mr. Marcus (not verified) :: Sun, 11/25/2007 - 1:27am

>>
Maybe it is McCarthy’s coaching. But then he was Favre’s QB coach in 1999, which was a pretty crappy year for him, and in his prime (57% complete, 23 picks….er, gunslings).

12
by Mr. Marcus (not verified) :: Sun, 11/25/2007 - 1:29am

--------------------------------
Maybe it is McCarthy’s coaching. But then he was Favre’s QB coach in 1999, which was a pretty crappy year for him, and in his prime (57% complete, 23 picks….er, gunslings).
---------------------------------

I could be mistaken but I think that was the year Favre had the broken thumb on his throwing hand.

13
by Rich Conley (not verified) :: Sun, 11/25/2007 - 2:51am

"7: Well, if this did happen, it wouldn’t affect comparisons so much because every good team would affect their opposition in the same way. It wouldn’t affect some teams more than others really. "

Yes, but the patriots have roughly TWICE the dvoa of the next best team in the last 10 years, so saying every good team would affect it the same way is untrue.

They do multiple runs to minimize it, but it doesn't totally go away.

Pretty much every team the Pats have played has had their DVOA go down after that game.

14
by JimboJones (not verified) :: Sun, 11/25/2007 - 11:02am

As a Baltimore fan, I was very impressed with what Jamal did vs. the Ravens last week. He looked the best he's looked since his 2000-yard season.

Based on how tough his schedule has been to date, I wouldn't be at all surprised to see him start posting big numbers, starting today with Houston.

15
by Temo (not verified) :: Sun, 11/25/2007 - 3:21pm

13: "Pretty much every team the Pats have played has had their DVOA go down after that game."

I don't know if this is true or not, but it has interesting complications. If were to take DVOA as theory says we should (that it quantifies a team's performance against the quality of performance against it), then really it should not vary from game to game (minus random error- I'm talking about the statistical "true mean" DVOA).

But if, say, a team consistently (and to a statistical significance) causes opponents to perform worse against them than DVOA says they should (IE Buffalo plays worse against NE than it theoretically should) than can it be said that teams are intimidated by NE (or some other likely explanation) and that is the reason behind this?

16
by Jacob Stevens (not verified) :: Mon, 11/26/2007 - 4:35am

15 - I wanted to comment on that factoid as well. Perhaps Aaron can confirm for us if it is true. I think that, and the implications in this article, together sort of build an indictment on the accuracy of opponent adjustments in DVOA/DPAR.

Which is something that emotions from rooting for a good NFC West team over recent years has had me frequently look for, but never able to find, but I've vaguely thought before, that opponent adjustments may be more accurate if they were measured more logarithmic, than a straight grade.

Guts n Stomps seems to support this idea. The really good, superbowl-contending teams blow out weaker opponents because the opponents are weaker, but it's also because they are that dominant. Just-good teams don't blow out the same weaker opponents. So we get tables like these, with a scattering of truly superb seasonal performances by truly superb players in their prime.

If each of the players in each table had an additional column, for their rank that season at their position, that might help indicate whether we're on to something, if we're seeing more negative opponent adjustments on the top third or so of the league each year, rather than the bottom 2/3rds.

Hasselbeck may benefit the most from weak opponents, again, this year, but even with the adjustments it's a pretty good season, and it feels better watching him than the DPAR numbers reflect.

17
by Herm? (not verified) :: Tue, 11/27/2007 - 4:42pm

My Favorite:
Tim Biakabatuka in 2000 goes from a below replacement -0.5 down to -8.9
Nothing like a bunch of smart guys inadvertently telling you they have supporting data to show that you sucked worse than everyone thought.