Writers of Pro Football Prospectus 2008

Most Recent FO Features

HarvinPer09.jpg

» Impact of the NFL's Kickoff Rule Change

After three NFL seasons of kicking off from the 35-yard line, what has been the impact on touchbacks, returns, field position, scoring and injuries? Also, is this rule responsible for a record number of big comebacks?

06 Oct 2009

Week 4 DVOA Ratings

by Aaron Schatz

(Note 10/7: The original posting of these ratings had an error in the defensive and total DAVE ratings. Those ratings are now fixed, as are the stats pages and playoff odds. Our apologies.)

New Orleans moves back into the top spot of the Football Outsiders DVOA ratings this week as we begin to include opponent adjustments for the first time in 2009. Those adjustments are currently at 40 percent strength and will go up 10 percent each week until we get to Week 10.

The shocking thing about the Saints is not that their offense ranks second. It's that their defense ranks first in the entire league. In fact, this year's defensive DVOA ratings seem like a testament to the power of coaching changes. The Giants and Eagles are in there, sure, and we all knew they would have good defenses. But the Saints are first, the Broncos third, and the Jets fifth. I don't remember a lot of teams where a defensive coordinator or head coach had this much effect on the quality of the defense. This certainly didn't happen when Mike Nolan arrived in San Francisco, or when Gregg Williams went to the Jaguars last year.

So are these defenses for real -- particularly the Saints and Broncos, who were so bad on defense last year? I decided to go and look for the defenses with the biggest difference between DVOA one season and DVOA over the first four games of the next season. Here's that list going back to 1994:

BEST DEFENSIVE DVOA IMPROVEMENT IN
FIRST FOUR GAMES vs. PREVIOUS YEAR
YEAR TEAM DVOA AFTER
FOUR GAMES
DVOA
Y-1
DIF FINAL
DVOA
1996 GB -52.2% 4.8% -57.0% -24.3%
2009 DEN -25.4% 24.7% -50.1% ???
1996 DEN -33.3% 11.0% -44.3% -20.8%
2001 CLE -32.2% 8.2% -40.3% -17.0%
1999 JAC -40.3% -0.8% -39.6% -15.2%
2009 NO -28.7% 10.7% -39.5% ???
2008 BAL -45.5% -8.7% -36.9% -24.5%
2006 ATL -25.3% 10.7% -36.0% 3.1%
1998 MIA -27.5% 7.7% -35.2% -26.1%
1996 NE -21.7% 11.4% -33.1% -9.2%
2007 WAS -17.7% 14.9% -32.6% -6.7%
2001 STL -18.4% 13.5% -32.0% -16.3%

As you can see, only one of these defenses ended up being a fluke: the 2006 Atlanta Falcons. The Falcons had not added any particularly notable free agents or coaches. They simply had a nice four-game run, keeping Carolina, Tampa Bay, and Arizona to a combined 19 points. The other game in that string was the well-remembered "return to the Superdome" Monday night contest against the Saints, but that game was actually a lot closer than the emotion made it seem at the time. The Falcons only allowed 16 points on defense (plus a special teams touchdown). However, after their Week 5 bye, the Falcons defense disintegrated, giving up at least three touchdowns in six of the next seven games.

For the other teams on this list, however, strong defense early was an indicator of defensive improvement that would last for the season. (The 2007 Redskins were essentially average over the rest of the year, but even that was a big improvement on 2006.) There are a lot of good teams on this list. Three of them made the Super Bowl. The 1996 Broncos went 13-3, the 1999 Jaugars 14-2. The 2001 Browns went 7-9 despite ranking 30th out of 31 teams in offensive DVOA.

What if we look only at good defenses through four games, no matter how good those defenses were the year before? From 1994-2008, 60 different teams had defensive DVOA better than -20% after four games. Only eight of those teams finished the season with a defensive DVOA above 0%, and only one finished with a defensive DVOA higher than 4.0% (the 2004 Seahawks, at 5.7%).

Based on all these stats, the chances seem pretty good that the Saints and Broncos defenses are for real. They may not be among the top three defenses at the end of the year, but they should be good defenses for the rest of the season. Combined with their current offensive levels, that should be enough for Denver to make a serious playoff run and for New Orleans to be a major Super Bowl contender.

For all those curious, here's the same list for offenses with the biggest improvement in the first four weeks compared to the previous season.

BEST OFFENSIVE DVOA IMPROVEMENT IN
FIRST FOUR GAMES vs. PREVIOUS YEAR
YEAR TEAM DVOA AFTER
FOUR GAMES
DVOA
Y-1
DIF FINAL
DVOA
1999 WAS 49.0% -3.8% 52.8% 17.1%
1999 OAK 20.4% -28.1% 48.6% 15.2%
1999 STL 28.9% -17.6% 46.5% 12.4%
2005 NYG 33.9% -10.3% 44.3% 8.9%
2002 KC 49.6% 7.9% 41.7% 38.0%
2008 NYG 40.7% -0.3% 40.9% 23.7%
2006 PHI 30.2% -9.1% 39.3% 22.0%
2006 SF -3.5% -42.0% 38.5% -7.5%
2006 NYJ 16.5% -21.5% 37.9% 6.2%
2005 SEA 40.3% 4.1% 36.2% 27.4%

Surprised that the 1999 Rams aren't number one on this list? Me too. I love that the 2005 49ers were so bad that they show up on this list even though they still had a below-average offense in the first four games of the next season. There aren't any 2009 teams that come close to making this list, but there is one team on the opposite list, the teams with the biggest drop in offensive DVOA from one season to the first four games of the next season. Look at these teams, and you can see that the Carolina Panthers are in trouble unless they can somehow pull Matt Cassel out of a hat:

WORST OFFENSIVE DVOA DROP IN
FIRST FOUR GAMES vs. PREVIOUS YEAR
YEAR TEAM DVOA AFTER
FOUR GAMES
DVOA
Y-1
DIF FINAL
DVOA
2001 WAS -62.0% 3.2% -65.2% -14.4%
1999 DEN -30.5% 30.8% -61.3% -2.3%
2005 MIN -42.1% 18.4% -60.5% -15.3%
2005 NYJ -34.9% 22.1% -57.0% -21.5%
2004 MIA -60.1% -6.6% -53.6% -28.5%
2009 CAR -33.5% 18.0% -51.5% ???
1996 ATL -35.5% 15.7% -51.2% -7.8%
2008 NE -0.8% 45.2% -46.0% 16.5%
2008 CIN -27.5% 12.7% -40.2% -13.4%
1996 CHI -18.5% 21.1% -39.6% -5.1%
2006 OAK -41.3% -2.2% -39.1% -36.5%

That leaves one more list: The teams that had the biggest defensive declines between one season and the first four games of the next season. There are two 2009 teams on this list. One of them is not very surprising. The other one is extremely surprising, especially because that team is sitting on top:

WORST DEFENSIVE DVOA DROP IN
FIRST FOUR GAMES vs. PREVIOUS YEAR
YEAR TEAM DVOA AFTER
FOUR GAMES
DVOA
Y-1
DIF FINAL
DVOA
2009 PIT 14.5% -26.9% 41.4% ???
2001 BUF 26.2% -13.4% 39.5% 8.8%
1995 ARI 24.1% -14.5% 38.6% 0.4%
2004 STL 23.3% -13.9% 37.2% 15.8%
2001 DET 24.2% -12.8% 37.0% 10.2%
2001 TEN 8.7% -27.0% 35.7% 6.5%
2009 TB 28.1% -7.3% 35.4% ???
1998 PHI 31.2% -3.8% 34.9% 5.6%
1998 WAS 26.9% -6.1% 32.9% 12.7%
1997 CIN 21.7% -11.1% 32.8% 14.1%
2000 STL 12.4% -19.1% 31.5% 13.5%
2005 BUF 1.5% -28.8% 30.3% 7.8%

Tampa Bay has a new coach and a ton of new players. They got a lot worse. Not a shock. Pittsburgh, though? Thanks in large part to their complete defensive meltdown against San Diego -- with an absurd 93.8% DVOA for the fourth quarter -- the defending champions now rank 23rd in the league in defensive DVOA. They are also the first team whose defensive DVOA went up over 40 percentage points between one season and the first four games of the next season. You also might notice that none of these teams really got back their defensive mojo except for the 1995 Arizona Cardinals. Of course, none of these teams had a player like Troy Polamalu who was injured in the first game but was expected back for most of the season. (The 2005 Bills lost Takeo Spikes early, but he was gone for the year.) The Steelers should be putting a better defense on the field starting in a couple of weeks. But these numbers should still worry Steelers fans a little bit, just like those other numbers should give confidence to Broncos and Saints fans.

* * * * *

All team stats pages should be updated for Week. Here are some housekeeping notes regarding the Premium database and player pages:

  • The Head-to-Head Matchups View is now fixed and should properly show 2009 stats for matchups.
  • We've added a new view that shows how teams do in the shotgun compared to non-shotgun plays on both offense and defense. When looking at these numbers for past years, remember that until a couple years ago there still were a few teams that never used shotgun at all, so beware of small sample size.
  • Player pages now include 1998 penalty numbers, and 1998 individual defense numbers should be on the pages by this evening.
  • 2009 rookies now have player pages, although there are no stats on them as of yet.

* * * * *

These are the Football Outsiders team efficiency ratings through four weeks of 2009, measured by our proprietary Defense-adjusted Value Over Average (DVOA) 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.)

OFFENSE and DEFENSE DVOA are adjusted to consider all fumbles, kept or lost, as equal value. SPECIAL TEAMS VOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season. Because it is early in the season, opponent adjustments are currently at 40 percent strength.

As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.

DAVE is a formula which combines our preseason projection with current DVOA to get a more accurate forecast of how a team will play the rest of the season. Right now, the preseason projection makes up 40 percent of DAVE (55 percent for teams that had a bye week).

To save people some time, please use the following format for all complaints: <team> is clearly ranked <too high/too low> because <reason unrelated to DVOA>. <subjective ranking system> is way better than this. <unrelated team-supporting or -denigrating comment, preferably with poor spelling and/or chat-acceptable spelling>


TEAM TOTAL
DVOA
LAST
WEEK
TOTAL
DAVE
RANK W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
1 NO 58.5% 2 32.2% 2 4-0 32.3% 2 -28.7% 1 -2.5% 19
2 PHI 43.8% 4 22.6% 6 2-1 14.7% 12 -25.9% 2 3.1% 7
3 BAL 41.1% 1 30.9% 3 3-1 27.5% 4 -16.0% 7 -2.5% 17
4 IND 40.8% 7 38.1% 1 4-0 36.6% 1 -5.6% 9 -1.4% 14
5 NYG 36.9% 8 27.8% 5 4-0 16.6% 10 -23.8% 4 -3.5% 24
6 MIN 36.7% 5 29.4% 4 4-0 12.7% 13 -15.5% 8 8.4% 3
7 DEN 36.5% 3 12.3% 10 4-0 14.7% 11 -25.4% 3 -3.6% 25
8 GB 20.8% 6 10.2% 11 2-2 21.0% 8 -3.9% 10 -4.1% 27
9 JAC 20.4% 17 19.6% 8 2-2 28.4% 3 10.0% 20 2.0% 9
10 NE 15.9% 11 20.2% 7 3-1 25.5% 7 7.9% 18 -1.7% 15
11 NYJ 12.0% 9 0.3% 16 3-1 -15.8% 27 -20.2% 5 7.6% 4
12 DAL 9.6% 10 5.7% 14 2-2 27.0% 5 18.4% 26 1.0% 12
13 PIT 8.2% 14 13.6% 9 2-2 26.1% 6 14.5% 23 -3.4% 22
14 ATL 3.3% 12 -7.5% 21 2-1 17.8% 9 15.3% 24 0.8% 13
15 SF 1.4% 22 -7.8% 22 3-1 -13.0% 24 -16.6% 6 -2.1% 16
16 CHI 1.1% 23 9.5% 12 3-1 -14.4% 25 -3.6% 11 11.9% 1
TEAM TOTAL
DVOA
LAST
WEEK
TOTAL
DAVE
RANK W-L OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
17 MIA -1.7% 26 -8.9% 23 1-3 1.8% 17 1.0% 13 -2.5% 18
18 HOU -2.8% 21 -5.4% 18 2-2 2.7% 16 10.2% 21 4.7% 6
19 CIN -3.1% 16 -3.8% 17 3-1 7.7% 14 1.9% 16 -8.9% 31
20 WAS -6.7% 15 -6.6% 19 2-2 -6.4% 21 1.6% 15 1.4% 10
21 SEA -7.3% 13 0.5% 15 1-3 -3.6% 19 4.6% 17 1.0% 11
22 ARI -10.7% 24 -15.4% 24 1-2 -9.3% 23 -2.9% 12 -4.3% 28
23 SD -11.9% 20 8.0% 13 2-2 6.2% 15 23.1% 30 5.0% 5
24 BUF -19.7% 18 -22.9% 27 1-3 -7.6% 22 8.7% 19 -3.4% 23
25 TEN -24.6% 19 -7.2% 20 0-4 -3.4% 18 10.3% 22 -10.9% 32
26 KC -27.6% 25 -17.2% 25 0-4 -14.9% 26 15.6% 25 2.9% 8
27 CLE -30.0% 32 -21.9% 26 0-4 -18.7% 29 21.9% 29 10.5% 2
28 TB -36.3% 28 -26.2% 28 0-4 -5.4% 20 28.1% 32 -2.8% 20
29 OAK -44.6% 27 -31.4% 31 1-3 -40.5% 32 1.1% 14 -2.9% 21
30 DET -48.6% 30 -32.7% 32 1-3 -17.1% 28 26.2% 31 -5.3% 29
31 STL -49.0% 29 -31.3% 30 0-4 -24.8% 30 20.6% 28 -3.6% 26
32 CAR -60.0% 31 -26.8% 29 0-3 -33.5% 31 18.6% 27 -7.9% 30

  • 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. Teams that have had their bye week are projected as if they had played one game per week.
  • PAST SCHEDULE lists average DVOA of opponents played this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • FUTURE SCHEDULE lists average DVOA of opponents still left to play this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
  • VARIANCE measures the statistical variance of the team's weekly DVOA performance. Teams are ranked from most consistent (#1, lowest variance) to least consistent (#32, highest variance).


TEAM TOTAL
DVOA
W-L NON-ADJ
TOT VOA
ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VAR. RANK
1 NO 58.5% 4-0 62.4% 4.0 1 -3.1% 19 -15.1% 32 0.6% 1
2 PHI 43.8% 2-1 50.6% 2.3 13 -7.3% 23 2.4% 12 25.5% 28
3 BAL 41.1% 3-1 47.9% 3.2 4 -13.4% 28 1.9% 13 20.2% 26
4 IND 40.8% 4-0 44.1% 3.2 3 0.2% 16 0.3% 16 14.7% 24
5 NYG 36.9% 4-0 35.7% 2.8 6 -15.2% 30 8.2% 4 8.2% 13
6 MIN 36.7% 4-0 41.3% 3.2 2 -14.1% 29 -5.8% 26 1.0% 2
7 DEN 36.5% 4-0 46.4% 3.0 5 -17.0% 31 4.7% 8 9.4% 16
8 GB 20.8% 2-2 24.5% 2.5 9 -3.6% 20 -6.9% 27 12.3% 23
9 JAC 20.4% 2-2 19.9% 2.7 7 0.7% 15 -7.7% 28 29.6% 30
10 NE 15.9% 3-1 14.1% 2.4 11 9.1% 8 1.8% 14 4.5% 7
11 NYJ 12.0% 3-1 11.4% 2.4 10 11.7% 6 -8.9% 30 10.4% 19
12 DAL 9.6% 2-2 11.7% 2.5 8 -5.7% 21 8.5% 3 8.0% 12
13 PIT 8.2% 2-2 18.8% 2.3 12 -9.6% 26 -0.8% 20 2.2% 3
14 ATL 3.3% 2-1 10.5% 2.2 14 -11.4% 27 5.2% 7 5.0% 9
15 SF 1.4% 3-1 4.7% 1.9 17 -7.6% 24 -1.1% 21 8.0% 11
16 CHI 1.1% 3-1 12.5% 2.0 16 -6.7% 22 3.5% 10 12.0% 22
TEAM TOTAL
DVOA
W-L NON-ADJ
TOT VOA
ESTIM.
WINS
RANK PAST
SCHED
RANK FUTURE
SCHED
RANK VAR. RANK
17 MIA -1.7% 1-3 0.3% 1.5 22 3.1% 12 0.0% 18 28.9% 29
18 HOU -2.8% 2-2 5.3% 1.9 18 -9.2% 25 0.3% 17 11.3% 20
19 CIN -3.1% 3-1 1.8% 2.1 15 8.9% 10 -2.1% 23 8.7% 14
20 WAS -6.7% 2-2 4.0% 1.8 19 -24.2% 32 8.1% 5 2.2% 4
21 SEA -7.3% 1-3 -15.5% 1.5 23 -1.4% 17 -7.8% 29 5.3% 10
22 ARI -10.7% 1-2 -10.5% 1.6 21 15.6% 4 -12.6% 31 42.7% 32
23 SD -11.9% 2-2 -15.5% 1.8 20 0.8% 14 -0.1% 19 2.5% 5
24 BUF -19.7% 1-3 -26.0% 1.4 24 9.1% 9 -3.5% 24 25.0% 27
25 TEN -24.6% 0-4 -30.4% 1.3 25 9.5% 7 1.4% 15 9.8% 18
26 KC -27.6% 0-4 -31.0% 1.0 27 19.3% 3 -1.4% 22 4.9% 8
27 CLE -30.0% 0-4 -43.2% 1.1 26 27.8% 1 -4.6% 25 30.2% 31
28 TB -36.3% 0-4 -31.6% 0.6 30 5.0% 11 7.2% 6 9.3% 15
29 OAK -44.6% 1-3 -36.2% 0.7 29 -1.4% 18 9.1% 2 11.8% 21
30 DET -48.6% 1-3 -59.3% 0.9 28 22.4% 2 2.5% 11 9.4% 17
31 STL -49.0% 0-4 -52.7% 0.6 31 2.1% 13 4.5% 9 2.8% 6
32 CAR -60.0% 0-3 -63.1% 0.0 32 14.1% 5 9.2% 1 18.4% 25

Posted by: Aaron Schatz on 06 Oct 2009

185 comments, Last at 11 Oct 2009, 11:57am by SKD

Comments

1
by ammek :: Tue, 10/06/2009 - 5:20pm

There are six NFL teams worse than the Chiefs. Wow.

10
by MC2 :: Tue, 10/06/2009 - 5:56pm

Eight, according to DAVE.

28
by Theo :: Tue, 10/06/2009 - 6:37pm

But they ARE the (top of my head) Lions, Bux, Panthers, Browns, Rams and the Raiders.
Am I right?
[edit... yeah they're pretty bad]

96
by hythlodaeus :: Wed, 10/07/2009 - 5:40am

I thought about that too. There are always bad teams in the league, but this year there seems to be more truely HORRIBLE teams than usual.

103
by bravehoptoad :: Wed, 10/07/2009 - 10:28am

It's because the salary cap is so huge (compared to past years) that bad teams can't make up the gap by poaching good free agents from good teams. (IMHO)

2
by ChiTown11111 (not verified) :: Tue, 10/06/2009 - 5:28pm

Im surprised by how low the bears are looking at their opponents rankings, but I suppose when you win by tiny margins based on lucky special teams plays...

53
by Thanos (not verified) :: Tue, 10/06/2009 - 9:04pm

and Cutler's fumblitis, and their third down conversion

3
by MikeM (not verified) :: Tue, 10/06/2009 - 5:29pm

Denver's a particularly interesting case to me. Without opponent adjustments they currently rank fourth.

What I've seen from their offense the improvement has been steady week over week. Losing Corell Buckhalter will hurt but Moreno with Jordan to spell and Hillis for goal line situations should work just fine.

I think the vast majority of Denver's "they're schedule's too hard coming up" comes from people over valuing SD, Pitt, and NE. Plus one of those games is on the road to Arrowhead, which the Broncos haven't played well historically at, but this is a new regime and we can throw past performances as predictive of future performances out the window with this group.

13
by An Onimous (not verified) :: Tue, 10/06/2009 - 6:00pm

Denver's got 6 games left against teams ranked 20th or worse in DVOA (2 against KC, 1 against OAK, 1 against WAS, 2 against SD). It's a tough schedule, sure, but it's hardly the back-breaker it looked like prior to the season. Even if they only take out KC/Oak/Was and lose to everyone else, that's an 8-8 record. Given the 6.5 over/under, that's a very nice season. Heck, as bad as SD is doing, 8-8 might win the division again.

33
by Formersd (not verified) :: Tue, 10/06/2009 - 6:46pm

Coming into the season, I thought the Chargers could sleep-walk their way to a division title. That is clearly no longer the case. Denver seems well-suited to give the Chargers trouble with the strong running game and suprisingly effective defense. I still think Denver has a rough patch coming, but I suspect they'll be playing meaningful games in December and I'm saying that as a Chargers fan.

38
by TBW (not verified) :: Tue, 10/06/2009 - 7:02pm

I think "sleep-walking" exaggerates how well the Chargers are playing. Face it, Norv has that team comatose, face down in the gutter. That smells like 2nd place to me this year in the AFC West.

41
by Eddo :: Tue, 10/06/2009 - 7:30pm

Weren't people saying the same things just last year, when the Broncos got off to a hot start, and the Chargers were struggling?

58
by Tracy :: Tue, 10/06/2009 - 9:36pm

Yes, I believe they were. It's remarkable that Norv Turner has done so little with so much talent, and still has a job. He's now wasted the prime of LDT's and Gates' careers, if he doesn't get fired soon, he'll waste the best years Rivers has, too.

65
by tunesmith (not verified) :: Tue, 10/06/2009 - 10:01pm

Note that this happened with San Diego before Norv Turner, too. They seem to have a habit of underperforming their talent.

77
by RickD :: Wed, 10/07/2009 - 1:26am

Under Marty, the Chargers only underperformed in the playoffs. Norv has taken a near-great team and made them mediocre.

124
by anotherpatsfan :: Wed, 10/07/2009 - 12:01pm

Would like Aaron or Bill or someone FO to analyze/opine on why the Chargers are way underperforming the nearly unprecedented glowing predictions for the Chargers preseason. Recent Kubiak update really downgraded Rivers and LT predictions for rest of season, so they realize something isn't going according to plan. Is is Norv, bad D, LT injury, all of the above?

132
by Matt R (not verified) :: Wed, 10/07/2009 - 12:28pm

The problem is Turner thinks O-lines and D-lines are optional extravagances. The Chargers have lots of offensive toys but put little focus on the guys in the trenches. For Schottenheimer, that was their identity every play. For Turner, the long bomb three times a game is what he lives for.

133
by Matt R (not verified) :: Wed, 10/07/2009 - 12:31pm

Other than going for it on 4 and 11 in the first quarter against the 2006 Pats, Schottenheimer coached a good game. But he was trying to be agressive and anti-MartyBall. The Chargers lost that game because of five poor plays -- including the absolutely inexcusable head but my Drayton Florence. Hard to blame that on Schottenheimer.

134
by Matt R (not verified) :: Wed, 10/07/2009 - 12:40pm

I'd like to know what the best theory is for why a team comes out as flat as the Chargers did. Obviously a coach doesn't need to spit and swear to get a team going. Dungy and Caldwell do just fine with their soft-spoken prep. And clearly Ron Rivera is willing to challenge the manhood of the D on Sunday night, which did nothing. So what explains the Norvesque incompetence? Wimpy practices? Guys not practicing at all? An unwillingness to bench stars? Honestly, we all agree Turner is terrible, but what is the exact reason he can be counted on to perfect the suckiest bunch of sucks that ever sucked.... Gotta go my damn weiner kids are listening.

62
by Purds :: Tue, 10/06/2009 - 9:48pm

Agreed. Put me down on the list of "not buying it yet." I think the Pats will expose Denver this weekend.

73
by PatsFan :: Wed, 10/07/2009 - 12:16am

Hope you're right! :)

Ought to be a pretty interesting game. Belichick said on his weekly radio call-in that Denver is running essentially the same offense than McDaniels was running in NE. Of course, who knows if Belichick is telling the truth :). Still, both coaching staffs must be pretty familiar with what the other staff is going to be trying to do on both sides of the ball.

78
by RickD :: Wed, 10/07/2009 - 1:27am

...but Kyle Orton is no Tom Brady.

138
by nat :: Wed, 10/07/2009 - 1:21pm

He's faster.

87
by The Hypno-Toad :: Wed, 10/07/2009 - 3:07am

The one thing I *think* I understand about this game is that it will probably be pretty low-scoring, as both sides are presumably pretty familiar with what the other side will be trying to do. That said, in the end, I expect talent to win out, and the Patriots will win, but that it will be harder than they (and most NFL fans) expect.
And while it's certainly possible that the Bronco defense will be "exposed" as some posters have suggested, I think it's more likely that the defense is just about as good as it appears... good enough to nearly shut down disappointing/mediocre offenses. But facing a truly elite group of offensive players, I expect the Broncos to lose in the roughly 17-10, 20-16 range, while giving a good account of themselves.
Of course, I hope I am underrating the Broncos, and would love to see them get out to 5-0.

98
by Mr Shush :: Wed, 10/07/2009 - 7:35am

I watched the Broncos for the first time this season on Sunday. On the one hand, their defense is clearly vastly improved compared to last season. On the other hand, that was the kind of low scoring game that seems like a crapfest, not two teams playing great defense. Dallas's offense (and Romo in particular) just looked awful. Boneheaded penalties. Miscommunication between quarterback and receiver leading to a red-zone interception. Passes thrown miles off target under little pressure. Sacks where the quarterback had forever to find a receiver or throw the ball away, and didn't. The Broncos have faced the Browns (who may have the worst offense in the NFL), the Raiders (who are so bad the freakin' Texans held them to six points and picked up a safety) and the Bengals (who appear completely unfathomable). I think that come full strength opponent adjustments based on more data, the Broncos defensive performance to date will look a lot less impressive. They're probably a league average or slightly better unit on that side of the ball. That's a massive improvement both on what they were last year and what I and many other people expected. Unfortunately, I don't think their offense is anything to write home about. The Broncos are an indifferent team, not an awful one. In a division with two basket cases and the incomprehensible Chargers, that may give them an outside shot at a playoff spot, but I doubt it.

146
by C (not verified) :: Wed, 10/07/2009 - 1:53pm

Once again I agree 100% with you.

145
by C (not verified) :: Wed, 10/07/2009 - 1:51pm

New England is a 4 point road favorite...

The one that got me is that the Dallas offense is ranked higher than the Giants ...
I thought Houston's defense was ranked to high...

109
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 11:03am

When 'comatose' and 'face down in the gutter' are good enough for second place, that's a pretty crap-tastic division.

117
by B :: Wed, 10/07/2009 - 11:21am

and yet it's an accurate description of the AFC West.

126
by Matt R (not verified) :: Wed, 10/07/2009 - 12:10pm

Man, I love the quality of writing and brainpower of the writers and readers of this site. Well said.

4
by Temo :: Tue, 10/06/2009 - 5:33pm

FOMBC curse strikes PIT defense, Steelers fans recoil in terror.

20
by Jerry :: Tue, 10/06/2009 - 6:07pm

It's what the preseason DVOA formula predicted, much to everyone's surprise.

94
by bubqr :: Wed, 10/07/2009 - 4:14am

Yeah I recall a lot of discussion about the projected decline, and I don't remember really why (apart from some key players aging).

123
by anotherpatsfan :: Wed, 10/07/2009 - 11:58am

Certaintly those bitching about that "crazy" Steeler's D prediction seem quite silent now...

5
by Joe in Seattle (not verified) :: Tue, 10/06/2009 - 5:35pm

Seahawks still with a 50% chance to make the playoffs! woo hoo! Thank you NFC West.

11
by Tundrapaddy (not verified) :: Tue, 10/06/2009 - 5:57pm

(sigh) My money's on the 'under' for that 50 percent chance, Joe.

But at least we'll still have Denver's 1st-rounder in April. So we can finally get either a new franchise QB or a new franchise LT. Or both.

15
by Joe in Seattle (not verified) :: Tue, 10/06/2009 - 6:03pm

Yeah, it certainly doesn't feel like 50%. I still have a little hope, but they need to win the next two games.

17
by An Onimous (not verified) :: Tue, 10/06/2009 - 6:04pm

It's looking more and more like Denver's first rounder isn't going to be anyhwere near "franchise QB or LT" territory. You can get yourself a franchise WR, though.

19
by Joe in Seattle (not verified) :: Tue, 10/06/2009 - 6:07pm

Seattle's pick might be there.

29
by Tundrapaddy (not verified) :: Tue, 10/06/2009 - 6:37pm

Depends on how soon the O-line and Hasselbeck return to health.

I'm expecting a mid-round pick for Seattle right now.

69
by mm (not verified) :: Tue, 10/06/2009 - 10:54pm

As someone who doesn't root for anyone in the NFC West, I think the Seahawks should sign Pennington and Garcia next year. Have Pennington and Hasselbeck as co-starters (one will be injured at any one time!), and have 47 year-old Garcia ready for the inevitable stretches their both injured.

70
by mm (not verified) :: Tue, 10/06/2009 - 10:56pm

"their" should be "when they're"!

142
by Tundrapaddy (not verified) :: Wed, 10/07/2009 - 1:45pm

I was going to leave accidental typos alone; it's douchey to quibble over such.

Then I saw that you misspelled 'grammar' in your preemptive strike.

:)

24
by bravehoptoad :: Tue, 10/06/2009 - 6:22pm

SF's pick from Carolina is looking preeeeeeety good right about now, though. We might be able to afford an honest-to-god quarterback.

79
by RickD :: Wed, 10/07/2009 - 1:29am

look for Panthers to "upset" Redskins. Zorn has no clue what he's doing.

147
by Kevin from Philly :: Wed, 10/07/2009 - 2:08pm

Yeah, that game should be...how did an earlier poster say it? Oh yeah, craptastic!

30
by Jacob Stevens (not verified) :: Tue, 10/06/2009 - 6:37pm

Call me crazy, I see an upset against Jacksonville. Even with Seneca, and I am among the worst of his critics. He's atrocious and if he wins it'll be the only game this year he does. I'd like to see Hasselbeck back, but the inconsistent Jags whose defense relies on turnovers and offense is relatively feast or famine are about as favorable a matchup with a superior team as the Seahawks could face at home, when you study the dynamics.

45
by sam :: Tue, 10/06/2009 - 7:44pm

I don't think the Jaguars ARE inconsistent. They've played well in three games (one of which they lost on the road at Indy) and they lost ugly to last year's NFC champions who dismantled a rookie cornerback playing in his second game ever (I think he may have missed the whole preseason). I don't think Seneca Wallace is going to scare the Jacksonville defense. Now, I know very little about the Seahawks and have no clue how their lines stack up, but I think your analysis is off.

--
sam! or the original sam from the old FO

66
by Bobman :: Tue, 10/06/2009 - 10:35pm

Sam, Allow a Colt fan living in Seattle to edumicate you:

Seahawks OL looked a bit like the Packers OL last night--they started I 2 or 3 replacement guys and it showed. Wallace had little time all day, but he was semi-effective with the short-drop/short-yardage stuff. Colts totally stifled the run game, which I assume the Jags will be able to do.

The DL... it's hard to judge against a Manning-led Colts team. I think they got to him once or twice, but didn't succeed in sacking him. I don't think he felt much pressure, but he probably got rid of the ball before I was even aware of pressure from the comfort of my family room. Colts ran with decent success (wild success for us) so again, I imagine the Jags will do okay there. The D was tyring a new anti-Manning numerical audible scheme with everybody wearing a wrist-playbook and Tatupu calling numbers--the guys referred to the chart and seemed to be a bit in disarray.

good luck.

139
by Jacob Stevens (not verified) :: Wed, 10/07/2009 - 1:42pm

Heh. So far this week online I have yet to meet a Jag fan who doesn't question my claim that they are inconsistent. 4 short weeks, but they're back to the typical high variance when they're good. Last year they were consistent, but that wasn't a good thing as the line kept them from getting in gear.

Seneca only scares Seahawk fans.

6
by Temo :: Tue, 10/06/2009 - 5:39pm

Intersting that you didn't also point out that all these teams did regress in DVOA to the end of the season. Seems like the real point is "Expect the Broncos and Saints to have good defenses this year, but not quite as good as they've shown thus far".

To that enxtent, I think a stat for "DVOA for weeks 5-17 vs. DVOA for year Y-1" would have been appropriate here.

25
by dmb :: Tue, 10/06/2009 - 6:23pm

I had the very same thought.

37
by LarryinLA (not verified) :: Tue, 10/06/2009 - 6:56pm

Those were my thoughts as well.

42
by Dan :: Tue, 10/06/2009 - 7:32pm

You can calculate the approximate number, assuming that final DVOA is 1/4 DVOA through 4 games and 3/4 DVOA for games 5-16. On average, the 10 teams (before this season) with the largest defensive DVOA improvement after 4 games had a 7.3% DVOA the previous year, a -31.4% DVOA through 4 games, and about a -10.5% DVOA over the rest of the season (-15.7% over the full season). And Atlanta 2006 is the only team with a positive DVOA over the rest of the season.

52
by dmb :: Tue, 10/06/2009 - 9:02pm

Sure, it's an easy thing to approximate, but the point is that in this analysis, performance in weeks 5-16 is more interesting and relevant than performance for the season as a whole. After all, it would seem extremely likely that performances in year y will be better than the performance in year y-1 simply because 1/4 of the performance in y is already vastly superior to anything from y-1. (E.g., the Broncos' defense would have to play substantially worse than it did in 2008 for weeks 5-17 in order to show no improvement in whole-season play.) What's interesting is how much performance will regress for the rest of year y -- will it return to the y-1 average, stay close to the early part of year y, or fall somewhere in between? So requesting info about Weeks 5-17 is less about the need for calculation; it's really just pointing out that Aaron's focus on whole-season performance was slightly off.

7
by bsr (not verified) :: Tue, 10/06/2009 - 5:47pm

I am confused by Aaron's discussion of the defenses. Looking at his chart, it would seem to indicate that almost all the defenses listed end up with a final defensive DVOA much less than after week 4. As such wouldn't that mean that the defense regresses after week 4 and by an amount less than what is indicated by the final DVOA?

I am probably doing the math wrong but for instance in the first example, if GB had a -52% DVOA after four games but ended up with a -24% DVOA wouldnt that mean that their defense averaged around 9% DVOA after week 4?

9
by Thok :: Tue, 10/06/2009 - 5:55pm

More like a -14% DVOA (and that ignores the effect of opponent adjustments.) Remember, there's 12 games remaining to weight against the 4 already played games. You want [12*x+4*(-52)]/16=-24, where X is the average performance after week 4.

(All this is ignoring opponent adjustments, which would also eat away at the -52.)

12
by Aaron Schatz :: Tue, 10/06/2009 - 5:59pm

The funny thing is that I thought I wouldn't have to point out that all these defenses regressed a little bit. Teams that start the season hot generally do. But the fact remains that most of the defenses were still pretty good the rest of the year.

Green Bay, for example, had a -16.8% DVOA in Weeks 5-17 of 1996.

21
by BucNasty :: Tue, 10/06/2009 - 6:09pm

I think there's a problem with your math, but I can't break it down. -52% is based on a small sample size. I believe the best defense of the DVOA era was something like -32%, so it's nigh impossible to sustain -52% over the course of a season. The defense shouldn't crash and burn, but it will regress to the mean. Despite the numbers, I don't think anyone seriously believes that New Orleans has the best defense in football. The point is that they will likely continue to be above average this year, so if Sean Payton can resist the urge to call reverses in short yardage while his team is just trying to run out the clock the Saints have an excellent chance of going all the way this year.

106
by Sophandros :: Wed, 10/07/2009 - 10:41am

"The point is that they will likely continue to be above average this year, so if Sean Payton can resist the urge to call reverses in short yardage while his team is just trying to run out the clock the Saints have an excellent chance of going all the way this year."

Of course the Bucs fan would have to bring THAT up...

-------------
Sports talk radio and sports message boards are the killing fields of intellectual discourse.

8
by jimm (not verified) :: Tue, 10/06/2009 - 5:48pm

Vikings number 3 in DVOA with a punt return and blocked fg td's against. They must be incredible other than those two plays.

14
by Tundrapaddy (not verified) :: Tue, 10/06/2009 - 6:01pm

I'm not sure the extent to which the 'blocked fg TD' counts against them. The bounce off the block, and thus the TD return, might be considered 'random'.

But in general, I think the Vikes have been getting pretty good field position, and perhaps pinning their opponents back more.

18
by Will Allen (not verified) :: Tue, 10/06/2009 - 6:06pm

Their returns, especially punt returns, have been very good.

40
by reinhard (not verified) :: Tue, 10/06/2009 - 7:19pm

Blocked kicks almost always have a good chance at getting returned for touchdowns. For example Clements against the Vikings, his job there is specifically to grab the ball in case of a block. And obviously no one on the kicking team is in a good position to make a tackle.

16
by Will Allen (not verified) :: Tue, 10/06/2009 - 6:04pm

Wow, the playoff odds calculations now have the Vikings at 11.9 mean wins. I guess that is pretty consistent with the ease of their schedule. Aaron, I wonder if you were to pretend it was mid August again, and were to recalculate your win projection for the Vikings, except now assuming a qb performance in the, say, 75th percentile, what the number would be. I know that is not how the model works, but I think you get my meaning.

111
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 11:11am

Where are the playoff odds? I've seen the link in the DVOA articles in the past, but I can't ever find it on my own.

120
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 11:36am

Nevermind! I found the link on the homepage.

121
by Tom Gower :: Wed, 10/07/2009 - 11:37am

It's one of the links under the Statistics drop-down at the top of the site, and also at the bottom of the page under "Innovative Statistics", between Drive Stats and FEI Ratings.

22
by Tundrapaddy (not verified) :: Tue, 10/06/2009 - 6:11pm

Aaron and/or other FO staff -

I'm curious as to why, on the 'Playoff Odds' page, the Vikings have lesser odds than the Saints of making (45.9% to 53.7%) or winning (26.7% to 29.9%) the NFC Conference Championship game, yet they have slightly better odds of winning the Super Bowl (14.7% to 14.6%). Clarification por favor?

27
by Vince Verhei :: Tue, 10/06/2009 - 6:36pm

The system simulates the entire season, one game at a time, up to and including the Super Bowl. Then we run it again. And again. Then 9,997 more times. If the Giants are playing the Raiders, the Giants will win most of those, probably 9,950 of those, but every once in a while the simulation will pick the Raiders to win.

So after 10,000 simulations, the Vikings won the Super Bowl 147 times, compared to 146 Super Bowl wins for New Orleans.

If we ran the system 10,000 more times, maybe New Orleans would win 160 and Minnesota would only win 100.

Or, to answer your question in one word: Luck.

34
by Tundrapaddy (not verified) :: Tue, 10/06/2009 - 6:47pm

Thanks for the explanation.

39
by Dan :: Tue, 10/06/2009 - 7:09pm

Shouldn't that be 1470 times vs. 1460 times, rather than 147 vs. 146? Or really 1465-1474 vs. 1455-1464?

55
by Vince Verhei :: Tue, 10/06/2009 - 9:33pm

Oh, probably.

35
by JuridianSantaal... :: Tue, 10/06/2009 - 6:47pm

I'm not staff, but if I were to take a guess:

Vikings have lesser odds of making or winning the NFCCG because of division and schedule. There's some upset potential in the NFCN with GB and Chicago twice. The Saints still have 6 games against the NFCS, which features two of the worst teams in football, which has a good chance to make any NFCCG being played to be played in New Orleans and there is a homefield advantage factor.

However, it probably means it sees the Vikings as an overall better team.

EDIT: I guess I need to hit refresh before I post.

57
by Vince Verhei :: Tue, 10/06/2009 - 9:34pm

Actually, that's part of it too.

48
by Pat (filler) (not verified) :: Tue, 10/06/2009 - 8:43pm

I think Vince is actually wrong here - there probably is a non-luck reason.

The Vikings probably have lesser odds than the Saints of making the NFC Championship game because of their schedule. Look at the details: the Saints have a 42% chance of getting the #1 seed, and a 70.6% chance of getting the #1/#2 seed. Minnesota only has a 54.9% chance of getting the #1/#2 seed. That's just schedule, and when you're not the #1/#2 seed, your chance of getting to the NFCCG is much worse because of the extra game.

Once they get to the NFCCG, it thinks they have a slightly *higher* chance than the Saints of winning the game, indicating that it thinks the Vikings are probably a very slightly better 'playoff team' than the Saints. The Vikings win 58.2% of their NFCCGs; the Saints win 56% of their NFCCGs. The higher "win" percentage for the Saints is because of their higher chance to get to the NFCCG in the first place.

Once they get to the Super Bowl, it then says the Vikings would win 55% of those, and the Saints would win 49% of those. That's consistent with thinking the Vikings are a better 'playoff team.' The bigger gap is probably because since the Vikings have a lower chance of making the NFCCG, in those situations, the Saints played an easier opponent than the Vikings (so the 'average Vikings opponent' in the NFCCG is harder than the 'average Saints opponent').

60
by Vince Verhei :: Tue, 10/06/2009 - 9:41pm

Actually, we are both right and wrong.

I was right about the luck factor -- that is how the system works -- but I was wrong to ignore schedule entirely.

You're right in that the schedule is a factor, but wrong in thinking the system picks superior "playoff teams." The system picks better teams based on DVOA, and that's that.

Except ... this early in the season, the system is still using DAVE, not DVOA. And as far as DAVE is concerned, Minnesota is not a better "playoff team" than New Orleans, they're a better team, period.

Which brings us back to your major point, which is true: the Saints have an easier path of getting TO the NFC title game than Minnesota, because of their easier schedule, but Minnesota has a better chance of winning once they do get there (according to DAVE).

Now watch, they'll both lose next week and this will all have been a giant waste of time.

68
by mm (not verified) :: Tue, 10/06/2009 - 10:45pm

to all of the above: the schedule means the Saints are more likely to have home field advantage in an NFC championship game (which gives them some boost in the formula); there is no home field advantage in the Super Bowl (barring a Dolphin surge).

Now watch, they'll both lose next week and this will all have been a giant waste of time

I think the Saints will defeat their bye.

75
by DGL :: Wed, 10/07/2009 - 1:04am

The Raiders, on the other hand...

113
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 11:13am

Seems to me that the Saints lose the NFC championship game more often in the simulation.

23
by bravehoptoad :: Tue, 10/06/2009 - 6:18pm

SF 15th, ATL 14th in total DVOA. That might actually be a good game, particularly since our offense is as bad as their defense, and our defense as good as their offense.

It's lovely to be in the top half at last, maybe the first time since 2003?

If only we could play the Rams 16 times in a row.

26
by jimm (not verified) :: Tue, 10/06/2009 - 6:27pm

The change in Kluwe since game 1 has been remarkable. Last night - 5 punts 0 return yards. After week 1 the Vikings apparently told Kluwe to work on hang time as opposed to distance (really - week 2 2009 - it took that long to figure this out).

Well it's working very well.

Only thing left to add - kickoff specialist.

31
by jimm (not verified) :: Tue, 10/06/2009 - 6:38pm

bravehoptoad - I just got over my hate of the niners after 20 or so years of ridiculously good teams and here they come back into the top half of the league - I can feel the hate (jealousy?) creeping back.

Now if only Dallas could go another 10 years or so not making the playoffs - then I could lose that hate.

Funny how I hate some successful franchises but not others

List of hated franchises very good franchises

1. Dallas (America's team - how annoying and I'm Canadian for gosh sakes
2. NE

List of very good franchises I like even though they win Super Bowls and the Vikings don't

1. GB (go figure - always cheer for them once Minnesota is out)
2. Indy (maybe it's the Dungy connection)
3. Pittsburgh (Tomlin? nope always kind of liked them)
4. Philly (always liked Andy Reid - but maybe it's because they win the NFC but never the Super Bowl - just like the Vikings of the past)

Good Franchises that I have no feeling one way or the other

1. New York Giants
2. Baltimore

115
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 11:16am

Interesting.

I hate the Eagles and Giants because of the Vikings' playoff losses to them in the not-so-distant past. I mean, c'mon, 41-0? Better to miss the playoffs.

149
by Kevin from Philly :: Wed, 10/07/2009 - 2:20pm

If you were from Philly, you'd know that there's NO end to hating the Cowboys.

32
by thescatman (not verified) :: Tue, 10/06/2009 - 6:44pm

hmm....given that New Orleans and Carolina play in the same division it seems a little unusual that one should have the hardest future schedule and one the easiest

36
by John (not verified) :: Tue, 10/06/2009 - 6:51pm

Well, they have yet to play each other, so Carolina has to face New Orleans twice, and New Orleans gets to play Carolina twice. Quite a strength of schedule difference, don't you think?

44
by the silent speaker (not verified) :: Tue, 10/06/2009 - 7:37pm

On top of that, NO finished last and CAR finished first last year. That only acccounts for two games each, of course, but those two and their head-to-head are 25% of the season that's at maximum possible differential.

176
by BigCheese :: Thu, 10/08/2009 - 1:25am

Well, the Saints already played the Lions, so that 4th-place team is not in the future schedule. The actual diferences in their future schedules are:

CAR, STL, ATL, DAL, CAR for the Saints

ARI, NO, NYJ, MIN, NO for the Panthers

So yeah, that's a pretty big difference right there. Man, Carolina is cooked...

- Alvaro

43
by Bjorn Nittmo (not verified) :: Tue, 10/06/2009 - 7:33pm

Sorry to beat a familiar drum, but: is there a "Philadelphia Eagles" dummy variable buried somewhere in the DVOA formula? Is there another way to explain how a defense that's played 3 games, in one of which its defense surrendered 421 yards and 41 points, and beaten two putrid offenses is somehow ranked as the 2nd best defense in the league? (I suppose I'd also note that the New Orleans defense gave up 463 yards to the Eagles in that same game yet ranks #1.) DVOA's successful measurement of the Eagles last year was pretty remarkable, so it's earned the benefit of the doubt, but I truly can't understand how the math could work out to rank these units so high.

47
by Dan :: Tue, 10/06/2009 - 8:39pm

Just look at the drive stats - Philly's D is 2nd in fewest yards/drive & 1st in most TOs/drive. They're slightly below average in points/drive, but bad field position is a big factor there and creating double the average number of turnovers lets them overcome that. Their drive stats mean that they should have a good defensive VOA, and opponent adjustments only cost them a few points of DVOA since they're only at 40% and the Saints are about as good as the Panthers are bad.

Or you could look game-by-game. They completely destroyed Carolina (7 turnovers & a -94.3% VOA), got beat badly by the Saints (but not as badly as they beat Carolina, as we could see in the week 2 DVOA ratings), and played well against the Chiefs. That equals a good defense, on average, but also one that's extremely inconsistent (2nd to Miami in variance).

Will they keep this up? Probably not. It's only based on 3 games (and mostly just one amazing game against a horrible offense) - that's why DAVE after 3 weeks is 55% preseason projections and only 45% DVOA.

150
by Kevin from Philly :: Wed, 10/07/2009 - 2:24pm

We play Tampa Bay and Oakland next, followed by the Redskins - so the Eagles D ranking will probably stay pretty high for a while longer.

46
by Brendan Scolari :: Tue, 10/06/2009 - 8:18pm

I'll start this off by saying that i think this is a wonderful site with some great content and stats. But at what point do we start to question whether DVOA has much predictive power at all?

I had a discussion in the forums during the preseason with a commentor here (I think it was tuluse) about DVOA being descriptive rather than predictive. DVOA uses things like field goal percentage, defensive turnovers, and 3rd down conversion rate to evaluate teams but these stats are either completely or mostly luck. The FO guys readily admit that FG% has no correlation from year-to-year, so I'm not sure why it's part of the special team's rating. Defensive turnovers also have very little correlation from year to year (as shown here: http://www.advancednflstats.com/2008/01/is-3rd-down-conversion-percentag...), they are mostly dependant on the offense making mistakes, yet they are a big part of defensive DVOA. And then there's 3rd downs, which play a big part in DVOA but yet the sample size is so small that overall passing efficency is a better predictor of future 3rd down conversion rate (as shown here: http://www.advancednflstats.com/2008/01/explanation-vs-prediction.html). Obviously if you are trying to be descriptive with your stats you probably want to measure these things, but they are going to help you if you're trying to predict future performance. It seems this has been a frustration at FO after doubting the Cardinals and Giants the last two postseasons and having both teams reach the Superbowl.

And a quick look back at the preseason projections (http://www.footballoutsiders.com/dvoa-ratings/2009/2009-dvoa-projections) shows us that DVOA hasn't done a good job of predicting the season thus far. Of the top 8 projected teams (SD, IND, NE, CHI, PIT, JAC, TEN, MIN) only the Colts have played like a top 5 team, though the Pats and the Vikings have also done well. Most of the group has been pretty ordinary thus far, including "the best team in football" Chargers. The Jags and especially the Titans haven't really fared well at all.

The teams projected from 9-16 (BAL, NYG, SEA, STL, PHI, CAR, DAL, KC) don't seem to be projected well at all. The Ravens and Giants both look like top 5 teams thus far. All of DVOA's sleepers in this group have looked awful, the Seahawks to some extent but moreso the Rams and Chiefs. Those are two of the worst teams in football. The Panthers are another team that DVOA has been way off on.

The teams projected from 17-24 (CIN, GB, WAS, NO, DET, HOU, CLE, MIA) are slightly better projected than the other groups. The Saints have been one of the best teams in football and the Skins, Lions, Browns, and Dolphins are among the worst teams in the NFL. The Packers, Bengals and Texans seem to have good projections.

The bottom 8 teams (TB, OAK, ATL, NYJ, ARI, SF, DEN, BUF) should be the easiest group to project but DVOA had a few huge misses among this group. The Jets are clearly a very good team and the Falcons look good as well. The Niners being projected to have the 31st ranked defense looks downright absurd and FO's general hatred of the Broncos chances now looks misguided. The Raiders and Bucs should be ranked lower than they are if anything and while the Bills aren't very good they clearly aren't the worst team in football. Overall I'd guess only 3 of the 8 teams in this group actually finish in the bottom 8.

So where does that leave us? I don't think the preseason projections are really accurate enough to have merit, you'd probably have been just as accurate predicting every team to be average or just assigning rankings by picking names out of a hat. Each group only has 2-3 teams that look accurately pegged at this point. I think perhaps taking out some of the non-predictive factors would help the accuracy of the projections (As well as possibly coaching changes. I know those factor into DVOA but there is too much variation involved and it just seems like a bad idea in general.) Just going based of efficiency stats from the previous year and accounting for big time players added through free agency and the draft seems like it might produce the best results.

Anyway, just wanted to give some thoughts on that, keep up the good work, hopefully DVOA will be more enlightening in the future.

49
by Jimmy :: Tue, 10/06/2009 - 8:44pm

My first thought would be that however good FO thinks a team is all it can ever hand out is paper championships and I don't think anyone who writes for FO has ever pretended otherwise. You will always have to play the games.

I would also add that you need to give anything like this time, four games isn't all that many.

50
by MC2 :: Tue, 10/06/2009 - 8:56pm

In fairness to FO, in some cases where you say they're way off (e.g. Titans, Panthers), the conventional wisdom was even further off, so I don't think you can really hold these against DVOA.

56
by Brendan Scolari :: Tue, 10/06/2009 - 9:34pm

Oh no doubt I'm not trying to say that conventional expert picks or something like that were better. But when you project a team to be #14 in the league and they are dead last you are way off as well.

Obviously there's still time for things to change, it's only been 4 games. But so far the projections look really poor.

61
by Pat (filler) (not verified) :: Tue, 10/06/2009 - 9:45pm

But when you project a team to be #14 in the league and they are dead last you are way off as well.

That's *ordinal*. You can't do that. That just generates a massively correlated prediction.

But regardless, you can just say "You said they'd be #14, they're #32, you suck." You have to ask "well, how accurate did you think that #14 prediction was?" or "what's the likelihood that they'd end up at #32?"

Projections should be allowed to say "I have no clue." It's a game. How the heck would you predict a team like Denver that jettisoned half its personnel?

Any team that has a coaching change or acquires a new QB from the draft should just be assumed to have a huge error bar on the projection.

But so far the projections look really poor.

Compared to what? Poor is a relative term. What's it relative to?

80
by Brendan Scolari :: Wed, 10/07/2009 - 1:33am

"That's *ordinal*. You can't do that. That just generates a massively correlated prediction."

Huh??? Are you honestly going to argue that projecting a team to be middle of the pack and then having them be dead last (so far) is way off?

"But regardless, you can just say "You said they'd be #14, they're #32, you suck." You have to ask "well, how accurate did you think that #14 prediction was?" or "what's the likelihood that they'd end up at #32?""

I didn't say they suck , I aid I really like FO. All I said was I don't think the projections gave been very good.

As for how accurate I thought it would be, I'm biased. I thought it would be accurate because I trusted the DVOA projections, but that trust is dwindling because of how poorly the projections have fared so far.

"Projections should be allowed to say "I have no clue." It's a game. How the heck would you predict a team like Denver that jettisoned half its personnel?"

Sure, that's fair. Did FO do that? Hell no. The Broncos were ridiculed here for how bad they were going to be. Did they say that about the Panthers? The Rams? The Niners? All I remember are evaluations as though the projections were almost certain to be true.

"Compared to what? Poor is a relative term. What's it relative to?"

Well I'd expect them to fare quite a bit better than the talking heads. FO outclasses them in analysis, they should fare better than them here as well. I'd expect them to fare better than any crude projections system, say just projecting every team to be 8-8 or projecting them based solely on points scored and allowed from the previous season. How about as well as some of the more popular baseball projections (PECOTA, Zips, etc.)? Is it not fair to hold them to any standard?

88
by MC2 :: Wed, 10/07/2009 - 3:17am

I don't think it's realistic to expect NFL predictions to be as accurate as MLB predictions. The baseball season is 10 times as long as the football season, which means that you have 10 times as much data from the previous season to work with, and you have 10 times as many chances for variance to even out in the upcoming season.

101
by BMF (not verified) :: Wed, 10/07/2009 - 9:57am

Not only is there 10 times as much data, the data is cosiderably more separable and less situationally dependent. It's a pitcher vs. a batter, with defense being the only complex factor, and with a few exceptions, the batter is trying to do the same thing (or one of a couple minor variations on the same thing) all the time. It would take a lot more than 162 games per season to make PECOTA-like predictions on a given RB, WR, or QB because of how they influence each others' stats and how the defense, OL, coordinators, and score greatly change how they play the game and well they do.

93
by Pat (filler) (not verified) :: Wed, 10/07/2009 - 3:57am

"Huh??? Are you honestly going to argue that projecting a team to be middle of the pack and then having them be dead last (so far) is way off?"

Let me restate that: you're comparing predictions of teams based on ordinal placement. That is, saying that team X was predicted to be 9th, they're actually 3rd, etc.

You can't do that. If you do that, then the accuracy of team X depends on the accuracy of all other 31 teams. Since certain teams can be predicted relatively accurately (those depending on offensive success with a stable QB and coaching staff) and others can't (those who changed a ton) you'll demolish accurate predictions due to predictions that were *known* to be based on incomplete information.

You really need to be comparing the actual DVOA values, but you can't quite do that yet because the numbers are still waaay early.

"Sure, that's fair. Did FO do that? Hell no. The Broncos were ridiculed here"

You're really convolving subjective articles with a (mostly) objective projection here. The FO writers are allowed to be snarky and ridicule a team whose decisions they don't agree with. The projection system isn't making fun of the Broncos. It's just pointing out that on average, teams that change the way the Broncos change fail miserably.

I'd expect them to fare better than any crude projections system

And they do!

How about as well as some of the more popular baseball projections (PECOTA, Zips, etc.)?

Yeah, football isn't baseball. Non predictable events play a far smaller role in baseball than in football. You can work it out from looking at the distribution of team wins in both sports.

71
by MC2 :: Tue, 10/06/2009 - 11:59pm

Pat makes a lot of good points below. I'll add a couple more.

First, predictions are really hard. Being slightly better than a coin flip would allow you to take Vegas for a large sum. Even small edges are hard to come by.

Second, in regard to your claim that it's easiest to predict which teams will be really bad, I totally disagree. If anything, it's harder, since being really bad is often the result of having lots of injuries, most of which can't realistically be included in a projection system.

148
by C (not verified) :: Wed, 10/07/2009 - 2:09pm

Slightly beter than a coin flip? You have to hit 62% before you take anything from Vegas.

Statistical analysis maybe, but there are people who do make very solid predictions. How can you project a team full of new starters? Well you can start by looking at those players histories.

Just because scouting and predicting is "hard", doesn't mean people can't do it. There ARE people who actually make money off of their picks and predictions...

162
by Vandal :: Wed, 10/07/2009 - 5:03pm

You have to be 52% before you can start taking from Vegas.

http://www.pinnaclesports.com/League/Football/NFL/1/Lines.aspx

I use this site because of the lowest VIG in the business... Average payout is 1.95 on a 1 unit bet.. so if you bet 100 times, and win 52 of them, you'll come away with 101 bets in return...

I'm not for gambling, but you certainly don't need to win 62%, unless you're using scratch-and-win football tickets...

164
by C (not verified) :: Wed, 10/07/2009 - 5:22pm

Pinny was the best but they tried to cut off US customers.

Most books have 10% juice.
100 bets, 50 dollars per pop
52 wins = 2340
48 loss = -2400
net = -60 bucks

Sure you could "make money" by hitting 54%, but
1) you probably didn't get all your lines -110
2) you didn't make much at all (130 bucks) for the risk you took

Sure you could bet larger units, but you could also go big on the wrong sucker bets and you won't have the ability to chase.

"Pros" hit 60%, and there aren't many "Pros".

182
by jebmak :: Sun, 10/11/2009 - 9:17am

I'm pretty sure that pros hit about 55-57%. If someone is hitting 60%, then they are missing out on some money by not playing enough games.

I prefer Matchbook (theoretically). It is a betting stock exchange. The bets go to the point where there is just 1% separating them, like -101 +100. The bet accepter pays 1% vig, but the placer actually gets a little money.

169
by MC2 :: Wed, 10/07/2009 - 8:04pm

Well, it depends on how much juice you have to pay, which varies, but a typical amount would be 10% (e.g. you have to lay 11-to-10). In that case, anything above 52% shows a small profit. Anything above 60% would be classified as making a killing.

In spite of that, the number of people who are able to make a living as sports bettors is very small.

51
by Dan :: Tue, 10/06/2009 - 9:00pm

DVOA is different from the preseason projections. The projections already account for the fact that some stats (like turnovers & 3rd down conversions) are inconsistent. And there's evidence that the projections do provide information that you don't get from obvious sources like the previous year's record or the Vegas lines. I believe that there's also evidence out there that they do better than other people's projections at predicting the final standings, but I don't have that link handy. It's looking like they have some significant misses this year, like the Denver defense, but it's still early in the season & you have to compare them to other projections to see the value added.

54
by Dan :: Tue, 10/06/2009 - 9:20pm

I just looked at the correlations between the projections and the actual DVOA through 4 weeks. It's surprisingly good for offense (r = .53) and special teams (r = .49), but nonexistent for defense (r = -.06) and not that impressive overall (r = .24), although with the noise in only having 3 or 4 games per team it's hard to know what counts as impressive. They thought that teams like New Orleans, Denver, and San Francisco would have terrible defenses and San Diego, Carolina, Tampa, Jacksonville, and Tennessee would have good defenses. So this year, so far, they seem to have done badly projecting defenses. Are there other projections to compare them with, to see if it's just a surprising year for defenses?

59
by Pat (filler) (not verified) :: Tue, 10/06/2009 - 9:39pm

First point: You're not comparing anything with a baseline. You have some arbitrary idea of what "good" projections are. There are two baselines you could imagine:

1: Coin flip. Project every team 8-8.
2: Static. Project every team exactly the same.

The teams projected from 9-16 (BAL, NYG, SEA, STL, PHI, CAR, DAL, KC) don't seem to be projected well at all. The Ravens and Giants both look like top 5 teams thus far.

You're correlating the predictions at this point. The Giants were predicted to be a good team - they are a good team. The Ravens were predicted to be a good team - they are a good team. The ordinal rankings of those teams ("top 5") depends entirely on the behavior of the teams around them.

The FO guys readily admit that FG% has no correlation from year-to-year, so I'm not sure why it's part of the special team's rating.

Look at the Special Teams Manifesto. It's not FG% that's in there.

But at what point do we start to question whether DVOA has much predictive power at all?

When it stops having predictive power? Over the long term DVOA has a very high correlation to points and a strong correlation between seasons.

I don't think the preseason projections are really accurate enough to have merit

The correlations for previous seasons are found in the FAQ: the projections account for about 50% of the scatter in the distribution of the league (better than just assuming their DVOA stays constant). The other 50% is because it's a game!

Incidentally if you're going back to the original preseason projection article, that was based on the situation in the league at the time. It's not the same as now. It's a statistical projection, not an attempt to divine if Tom Brady is going to get hurt in Week 1.

Injury risk is in there, but that's a systematic error; when it happens, that team's going to be worse, *fast*, and stay worse.

The bottom 8 teams (TB, OAK, ATL, NYJ, ARI, SF, DEN, BUF) should be the easiest group to project

Why? I don't see any reason for this. Decline and improvement should be equally likely over the long term. The league is full of teams that make sharp improvements both upwards and downwards from year-to-year.

FO's general hatred of the Broncos chances now looks misguided.

That's because it didn't know anything about McDaniels or the new coaching staff, and so it expected them to be "an average new coaching staff." An average new coaching staff tends to be worse.

That sort of error is going to be systematic: on *average*, over many teams, it's going to be right, but for one specific team if it's wrong, it's going to affect everything.

It happened last year for Atlanta, too.

Just going based of efficiency stats from the previous year and accounting for big time players added through free agency and the draft seems like it might produce the best results.

Brilliant! Wait.. that's exactly what they do.

You would be *much* better off right now going back and looking at the *old* projections (which are all available) for previous seasons and seeing how those worked out. They do better than a coin flip and a static projection.

But really, c'mon. It's a game. You can't be that accurate when single events (coaching changes, quarterback changes) can shift a team from awful to good.

It also raises the question for how you actually rank a projection system; that's another question. It's really not as simple as looking at the mean projection and comparing it to the actual value, because the projection *has an expected error*.

76
by Brendan Scolari :: Wed, 10/07/2009 - 1:21am

Wow, I didn't expect a point-by-point response. But I suppose it's only fair that I respond in turn:

"First point: You're not comparing anything with a baseline. You have some arbitrary idea of what "good" projections are."

No, I actually said near the end of my post that I thought picking names out of a hat or projecting them all to be 8-8 would be just as accurate.

"You're correlating the predictions at this point. The Giants were predicted to be a good team - they are a good team. The Ravens were predicted to be a good team - they are a good team. The ordinal rankings of those teams ("top 5") depends entirely on the behavior of the teams around them."

Well, predicting a team to be the 10th best and having them be the 3rd best in not a really accuate prediction. But fair enough, projection systems obviously aren't going to nail teams spot on. But I like how you talk about the Giants and Ravens while completely ignoring that 3 teams from that group (Rams, Chiefs, Panthers) don't look like they will be anywhere close to where they are predicted. If Baseball Prospectus predicted the Rays to finish with 86 wins and they finished with 66 they'd get critcized for it, and I think that's pretty fair. Being nowhere close on 3 teams out of 8 is not very good.

"Look at the Special Teams Manifesto. It's not FG% that's in there."

Um, I'm not sure what your reading. This is what it says in the special teams section:

Field goal rating compares each field goal to the league-average percentage of field goals from that distance."

The question is why? Wouldn't Aaron or Bill tell you that field goals (when adjusted for distance) are luck? Why include this in the formula? The sample size for field goals is too small to measure anything, you'd need hundreds if not thousands of field goals to show any kind of measurable skill. So recording something that comes down to complete luck doesn't make your statistic more predictive.

"When it stops having predictive power? Over the long term DVOA has a very high correlation to points and a strong correlation between seasons"

Well obviously actual DVOA has a strong correlation with points. I'm talking about preseason DVOA projections here. How strong is their correlation? Dan (#51 in this thread) said so far there is a .24 correlation, that's pretty small. Is it any stronger than just predicting every team to be 8-8? If not what's the use of it?

"Why? I don't see any reason for this. Decline and improvement should be equally likely over the long term. The league is full of teams that make sharp improvements both upwards and downwards from year-to-year."

Maybe your right. I figure there is a bigger tail off at the end, in other words the worst teams have a bigger and more obvious drop off in talent lavel than the best teams have an upgrade in talent. Basically, how hard is it to predict that the Raiders or the Lions or the Browns would be bad this year? That should be pretty obvious. But saying a team like the Eagles will be really good is harder to do, a key injury (like to McNabb) could derail that projection.

"That's because it didn't know anything about McDaniels or the new coaching staff, and so it expected them to be "an average new coaching staff." An average new coaching staff tends to be worse.

That sort of error is going to be systematic: on *average*, over many teams, it's going to be right, but for one specific team if it's wrong, it's going to affect everything.

It happened last year for Atlanta, too."

Okay, so what? Either way it's a bad prediction. If the FO guys weren't sure then why laugh at how the Broncos were so bad all offseason and talk about how they just gave the Seahawks the #1 overall pick? I saw this written both on the site and in magazine interviews. It sure sounded like they were damn confident the Broncos would be terrible. If you not sure, why not just admit that?

Also, once again you completely ignore other teams that made my point. Sure the Broncos made a lot of changes, what about the Niners? They have the same coaches as last year and only 1 new defensive starter (Dashon Goldson at FS). And yet their defensive projection so far looks about as far off as it could possibly be.

What about the Rams? Convential wisdom said they were terrible. They made some upgrades (notably on the offensive line) but so far they still look absolutely incompetent as a team. Is it wrong to say that FO missed big time on the Rams?

"Brilliant! Wait.. that's exactly what they do."

Ummm... no, it's not. They are including the stats that have no correlation that I said above. They try to project teams based on coaching changes and switching to different scemes. My point is that these things have too many variables to account for in a projection. If they just simply projected based on efficiency stats and added in something for top free agents or very high draft picks DVOA would be a better predictive tool.

"You would be *much* better off right now going back and looking at the *old* projections (which are all available) for previous seasons and seeing how those worked out. They do better than a coin flip and a static projection."

So just because I think the projections are bad this year that means I must be new at FO?? This is my 4th year reading the site, I know what the old projections are like.

Do they fair better than a static projection? If so I'd like to see that. What about projecting based solely on each teams points scored and points allowed from the previous seaon? Is DVOA any better than using that method? There's nowhere I can see on the site where the FO guys look critically at their own methods. If they do fair better (which I'd surely hope they do) then that's great. But still I think they hold way too much confidence in their projections. Like I said earlier all the Broncos hate looks awful silly in light of how the season has played out so far. And I'm not even a Bronco fan, but ridiculing their team as the worst in the NFL before the season started was foolish. If your projections are going to miss on as many teams as FO's have so far you might want to speak with a little less certainty when weighing teams preseason chances.

"But really, c'mon. It's a game. You can't be that accurate when single events (coaching changes, quarterback changes) can shift a team from awful to good."

Yep, it is a game, there's lots of things that are tough/impossible to measure. But I'd still expect a projection system that's supposed to be pretty insightful to fair better than this.

83
by Eddo :: Wed, 10/07/2009 - 1:56am

"Well, predicting a team to be the 10th best and having them be the 3rd best in not a really accuate prediction. But fair enough, projection systems obviously aren't going to nail teams spot on. But I like how you talk about the Giants and Ravens while completely ignoring that 3 teams from that group (Rams, Chiefs, Panthers) don't look like they will be anywhere close to where they are predicted. If Baseball Prospectus predicted the Rays to finish with 86 wins and they finished with 66 they'd get critcized for it, and I think that's pretty fair. Being nowhere close on 3 teams out of 8 is not very good."

You're mixing up some issues here. Pat is saying that instead of saying, "the Giants were predicted to be the tenth best team, but they're the third best team", you should instead say, "the Giants were projected at 12% DVOA (or whatever), and they're actually at 16% DVOA". You highlight this with your Tampa Bay Rays example. BP could very well have projected them for 94 wins, and been quite wrong. However, that 94 wins could have gone along with a third-place prediction (considering the Yankees and Red Sox also have high projections), which would have been correct on the surface.

All Pat's saying is that you can't look at ordinal rankings when determining the correctness of projections. Now, your points about inaccuracies are probably still valid, but not because the 10th place projected team is actually in 3rd.

140
by Brendan Scolari :: Wed, 10/07/2009 - 1:43pm

I see your point, and that makes sense. I should have looked at the actual DVOA's instead of just the rankings. Still , obviously many of the teams I mentioned are pretty far off even from their projected DVOA so far.

89
by Jerry :: Wed, 10/07/2009 - 3:19am

"Look at the Special Teams Manifesto. It's not FG% that's in there."

Um, I'm not sure what your reading. This is what it says in the special teams section:

Field goal rating compares each field goal to the league-average percentage of field goals from that distance."

It means that Fred Kicker hitting 24 of his 30 FGA for 80% doesn't matter. Each attempt is compared to the league-wide percentage for that distance (adjusted for site), and points are credited or debited to the kicker; if this attempt is made 67% of the time, a make is worth 1 point and a miss is worth -2. (I don't remember if the value of ensuing field position is also added.)

The question is why? Wouldn't Aaron or Bill tell you that field goals (when adjusted for distance) are luck? Why include this in the formula? The sample size for field goals is too small to measure anything, you'd need hundreds if not thousands of field goals to show any kind of measurable skill. So recording something that comes down to complete luck doesn't make your statistic more predictive.

It's luck as far as the defense goes, but there's a difference in kickers - some have the leg to hit 55-yarders and others don't.

95
by bubqr :: Wed, 10/07/2009 - 4:47am

My take on the predictive power of DVOA :

It seems like DVOA/FO can't really predict the effect of drastic changes. It has troubles, at least recently, accounting for the changes in the coaching staff (ATL/MIA last year, NYJ, DEN, STL, NO Defense this year). And you can somewhat explain that. Numbers might tell you a lot of things, but what can they tell you about such a huge OFF THE FIELD (non mesurable) and mostly human-related change ? Not much IMO.

I'd like to see any kind of study about DVOA's predictive power on teams without any coaching change, where I expect them to be well above average.

105
by Eddo :: Wed, 10/07/2009 - 10:36am

"It seems like DVOA/FO can't really predict the effect of drastic changes."

Not disagreeing with you, but that's why they're called drastic changes; so much changes that it becomes nearly impossible for a model to project.

97
by DeltaWhiskey :: Wed, 10/07/2009 - 5:40am

Another great and civil discussion on FO with many valid points by all.

I'll add that I doubt the pre-season predictions were aimed at Week 4 and to judge them this early is hardly fair.

100
by ammek :: Wed, 10/07/2009 - 8:11am

Like I said earlier all the Broncos hate looks awful silly in light of how the season has played out so far.

Whereas all the experts who predicted a Chargers-Packers Superbowl are looking like geniuses.

I don't know enough about Denver to comment, but as far as the Packers chapter in FOA is concerned, there is more to commend than call out. While pundits worked themselves into raptures over a handful of preseason drives, FOA cautioned of a likely regression on offense owing to pass protection issues (bullseye!) and Rodgers' tendency to take off running under pressure (spotted by the game charters). It forecast Rodgers' completion percentage to fall (and it has), and regression in the red zone (correct again) and third-down offense (debatable). It identified Donald Lee's unsustainable red-zone performance (four weeks in, Lee has -57% DVOA) and expected injuries on offense (two left tackles and a backup RB so far). The chapter predicted an eventual improvement on defense in the later part of the season (we'll see), and projected the number of defensive TDs to drop from seven to 1.8 (they have scored two already). It identified the pass rush as the defense's primary weakness (not so in weeks 1-3, very much so against Minnesota).

In summary, I'd second the suggestion made in this thread that the projection system has difficulty when there is a significant offseason change in sideline personnel (new coaches, coordinators, scheme). It doesn't pick up many of the out-of-nowhere turnarounds (how could it?) but it is a better judge of a team's prospects than preseason games or "Shaun Hill just wins", and if you read it in detail (rather than concentrating on the rather blunt win-projection number) you'll find it a very useful guide to understanding what teams have and have not been able to correct from one year to another.

104
by Dan :: Wed, 10/07/2009 - 10:30am

Last season there was a .52 correlation between projected DVOA and actual DVOA at the end of the regular season (Atlanta included). This year after 1 week there was a .07 correlation. After 4 weeks there's a .24 correlation. Notice a pattern?

116
by Dan :: Wed, 10/07/2009 - 11:18am

Last season there was a .30 correlation between DVOA after week 4 and projected DVOA - slightly better than this year's .24, but still a long way from the .52 correlation at the end of the season.

122
by Kulko :: Wed, 10/07/2009 - 11:54am

Well there is definitely abehavioral pattern from the readers of this site.

Each year at week 3-5 somebody comes and claims that DVOA Projections are not worth their spittle because stupid prediction do just as well. Of course without ever showing any proof for this fact.

OTOH these are facts for the 2008 Season:

Predicting everybody at last years record has a corerelation of 0,21.
DVOA projected wins have a correlation of 0,31.
DVOA win projections were better then Vegas Over/Under, so you should be able to earn money on them. (there was an article linked in the offseason where somebody evaluated that)

And Wins is actually a very weak measure of success, which a huge variance due to Strength of Schedule. For better measures of team Strength this Correlation goes up to 0.5, which I think is a pretty good number for a Sport like Football, where you have loads of random events, a small sample size and team coherence as one of the major influence for personal Success.

143
by Brendan Scolari :: Wed, 10/07/2009 - 1:46pm

Cool, thanks. It's quite possible (maybe likely) I'm just overreacting to the first few weeks of the season and the projections will move closer to the truth as the season goes on. It just seems like there are an awful lot of teams poorly projected so far.

63
by Aussie Bengal (not verified) :: Tue, 10/06/2009 - 9:55pm

The Bengals' special teams are 31st? Waaaay too high. Should be about 51st, behind the Titans, the Trojans, the Buckeyes, the Gators, the Sooners, etc.

Seriously, though. All the numbers and rankings look about right to me. 2.1 projected wins. Mediocre offense and defense. Still, I can't help but take a qualitative look at the way they played against the Packers, Steelers and Broncos and feel like this is a top half of the league team right now. If teams don't improve (or injuries/arrests strike), 7-9. If any facet of the team improves (likely the young O-Line bucks up and Carson goes deep a little more often), 9-7.

153
by Kellerman :: Wed, 10/07/2009 - 2:36pm

Well, I was a little surprised to see their seasonal ST DVOA total drop so low after only one horrendous game. Other than this week's block Graham's only missed one FG , I think, which was a 50+ yarder against the Steelers. The snapping has been very bad, but I don't KNOW that it's all St. Louis. It MIGHT be partially because of a new holder, but I just don't know. Huber did seem to get off to a very good start with the punting but has seemed to have a bad half in each of the last two games. Until this week the return game seemed very good and the coverage has been passable. Is a big game by Cribbs worth getting worried about? Maybe it is. Has Hebert regressed in coverage? Maybe he has. Would the return of Roy Williams to the actives remedy this? Should we have kept Lynch over Nelson? It's obviously too late to remedy THAT.

Just looking at the tried-and-true method of picking each game the rest of the way according to what we think we know now it looks like the last game at NY will be the difference between 9-7 and 10-6. But I've been wrong before!

Incidentally, these DVOA rankings suck because Marv Lewis and Chad Ochocinco have new resolve and Carson can turn on the jets as needed. Just looking at total number of 4th down conversions is way better than this! Blood Pressure Bengals Rule!!!!!!

159
by mm (not verified) :: Wed, 10/07/2009 - 3:48pm

Well, I was a little surprised to see their seasonal ST DVOA total drop so low after only one horrendous game.

-one game is 1/4 of the season right now
-there aren't many plays on special teams to begin with, so a few bad plays there affect that ranking more than a few bad plays on offense or defense affects those rankings.

64
by Purds :: Tue, 10/06/2009 - 9:56pm

Let's all take a moment of silence in recognition of the Colts special teams being ranked 14th right now. ...

I doubt you'll ever see that again.

108
by Uno Ocho (not verified) :: Wed, 10/07/2009 - 10:59am

McAfee does look like a keeper. I'd say this ranking primarily comes from the fact that kickoffs and punts have been pretty excellent this year. Vinatieri still looks shaky and as always they have the worst kickoff returns in the league. Combine all this together and middle of the pack seems about right.

67
by Dan :: Tue, 10/06/2009 - 10:38pm

Is there a mistake in the DAVE numbers? Miami has a -1.7% DVOA and was projected for a -10.8%, but their DAVE is -23.8%. That shouldn't be possible, if DAVE is a weighted average of DVOA & the preseason projections. Some of the other numbers also look a little screwy (with DAVE too far from DVOA), like Houston, Pittsburgh, & SF. I've also contacted Aaron via the contact form about this.

118
by Aaron Schatz :: Wed, 10/07/2009 - 11:21am

Sigh, unfortunately there *was* a error in the defensive DAVE ratings... they were counting the current rating 40 percent instead of 60 percent. I've fixed that now. Most teams are only slightly affected -- Miami happened to be the team most affected. My apologies.

167
by Jacob Stevens (not verified) :: Wed, 10/07/2009 - 7:04pm

I thought it was the other way around? I thought current rating was counted 40% now, working up to 100% by week 10? 60% now would get us to 100% at week 8, I think.

183
by jebmak :: Sun, 10/11/2009 - 9:19am

I think that you are thinking of the opponent adjustments which are at 40% and go up to 100% at week 10.

72
by B-Rick (not verified) :: Wed, 10/07/2009 - 12:13am

It seems every year for the last 10 years there is a worst(or close) to first story in the NFL. This year Denver, Cincinati and San Fran look like solid playoff contenders.

Are these teams elite team like the NYG and Indy? NO, but they are all better than the prognosticators thought and I think at least two of them make the playoffs. When you look at the the turnover in playoff teams from year to year it is amazing how much credence is given to pre-season projections, even statistically unbiased ones like FO.

In the early part of the season I have always done well playing underdogs the first 3 or 4 weeks of the season as most teams are either significantly over-rated or under-rated early season.

90
by The Hypno-Toad :: Wed, 10/07/2009 - 3:23am

I used to bet on sports a few years ago... Like 2003 and 2004, primarily. The first few weeks of those seasons were just money farms, because the betting public seemed to be so sure that the Raiders were on their way back to their Super Bowl ways. It was simply an amazing confluence of my needs (to win money) and my wants (for the Raiders to not be very good).

74
by Matthew Smith (not verified) :: Wed, 10/07/2009 - 12:53am

One thing I'm not sure of is why Washington is being given a decent ranking. Three of their four games are against bottom five teams according to your ranking set, and the other was against the Giants, and they're only 2-2. Is this simply because opponent rankings are only at 40%? I'd assume that even at that strength it would be enough to blow up their position, but maybe I'm missing something important.

82
by RickD :: Wed, 10/07/2009 - 1:38am

The relatively high rating of the Redskins exposes a problem with DVOA. DVOA primarily measures the ability to move the ball, as opposed to the ability to score. The Skins offense has become noted for its ability to move between the 20s. And then they don't score.

(My opinion is that the Pats' defense has the opposite problem - it lets teams move the ball but doesn't let them score very much.)

If I could buy stock options in teams based on DVOA, I would definitely be selling the Redskins short. There's nothing there.

91
by MC2 :: Wed, 10/07/2009 - 3:27am

If you're right about the Pats' defense, then Pats' fans should be worried. The old notion of a "bend but don't break" defense usually turns out to be a mirage. I remember a few years ago, the Broncos got off to a great start, on the strength of a defense that was giving up a ton of yards, but very few points. The talking heads on ESPN were raving about their "bend but don't break" philosophy. Well, the defense continued to bend, but unfortunately, it also began to break, and they ended up with a well below average defense. I can't remember the exact year, but some of the Broncos' fans here probably can.

112
by Rich Conley (not verified) :: Wed, 10/07/2009 - 11:13am

MC2,

IIRC, the Patriots defenses during the superbowl years were always much better in the Red Zone than elsewhere on the field (IE, much higher DVOA than outside the 20s).

Bend-But-Not-Break defenses DO exist, and its why Red Zone defense is a huge percentage of the "Secret Sauce" formula that predicts playoff performance.

The Patriots defense should be interesting, and WILL get better as the year goes on. Theres a ton of good young talent that is really just learning to play, and Mayo will be back.

170
by MC2 :: Wed, 10/07/2009 - 8:11pm

Maybe you're right, although it doesn't really make sense, unless there's some reason why a particular team would be much better at red zone defense than regular defense, e.g. really good at stuffing runs in short yardage situations, really tall DBs who are great at defending fades, etc.

172
by Eddo :: Wed, 10/07/2009 - 8:47pm

Actually, it's not that hard to imagine how a team could be better in the red zone.

Let's say a team has an awful secondary. It's going to be easy to move the ball on them, as receivers will always be open.

However, in the red zone, everything is condensed. Fewer players can cover more ground. A good linebacking corps can more easily help cover receivers, masking the secondary's ineptitude. Therefore, the red zone defense for this team will be better than their defense between the 20s.

I believe the Eagles were an example of this a few years ago, except their linebackers were their weakness, and their secondary was good. I don't know enough about this year's Patriots to tell you if they have one really weak unit, though.

173
by MC2 :: Wed, 10/07/2009 - 9:04pm

That's a good point, although teams that have awful secondaries aren't usually "bend but don't break" teams. Teams with awful secondaries usually give up lots of points, but more by giving up long TDs, as opposed to red zone TDs.

178
by Mr Shush :: Thu, 10/08/2009 - 6:44am

As a Texans fan, I can attest that having a perennially awful secondary leads to year after year of a defense that most definitely breaks on a regular basis. This year, the more aggressive approach of new defensive co-ordinator Frank Bush appears to have shifted the defense from the "bend, break, fall over own shoelaces, whatever" philosophy of the Richard Smith years to "don't bend in the slightest, but break so often it almost doesn't matter".

I give you the following play-by-play from the Week 2 TEN@HOU game by way of an encapsulation of the 2009 Texans defense:

(1-10 TEN 1) K. Collins up the middle to TEN 2 for 1 yard (M. Williams)
(2-9 TEN 2) L. White right guard to TEN 4 for 2 yards (D. Ryans). Penalty on HST-66-De.Robinson, illegal use of hands, 5 yards, enforced at TEN 4
(1-10 TEN 9) K. Collins incomplete short right to A. Hall
(2-10 TEN 9) L. White right tackle to TEN 9 for no gain (A. Smith)
(3-10 TEN 9) (Shotgun) C. Johnson up the middle for 91 yards, TOUCHDOWN.

If the Texans defense didn't give up so many big plays, they'd be almost good. Regrettably, that problem isn't going away any time soon.

184
by SKD (not verified) :: Sun, 10/11/2009 - 11:54am

The 49ers D of the past two years has been like that as well. Except is was a combination of a very below average pass rush and terrible secondary. Patrick Willis was probably the main reason we were able to keep teams out of the end zone and we finished with a 7-9 record despite Just-TurnOvers (JT-Osullivan) leading the team for half the season.

130
by tuluse :: Wed, 10/07/2009 - 12:21pm

There was an article about bend-but-don't-break defenses in pfp 2008 I think. I believe the conclusion was that it's a real phenomenon and that they can last for a whole year, but it leads to regression the next year.

92
by The Hypno-Toad :: Wed, 10/07/2009 - 3:30am

Well, it also exposes an explicitly stated limitation/corrective measure of DVOA. At this point, opponent adjustments are less than half as strong as they will be. If I had to guess, if one was to look back at Washington's DVOA over the first four games once the opponent adjustments fully kick in, thier rating would no longer be decent.
Frankly, I'd rather have the odd "Why does DVOA think the Redskins aren't bad?" moment than the completely uninformative model of applying full-strength opponent adjustments starting in week one.

81
by RickD :: Wed, 10/07/2009 - 1:35am

As a former resident of NJ, I'm going to take exception to the description of a possible Jets-Giants Superbowl as a "Subway Bowl". There is no subway from East Rutherford to East Rutherford.

Call it the Hoffa Bowl.

86
by Red Hedgehog :: Wed, 10/07/2009 - 2:53am

I commented on this in an earlier DVOA article. My suggestion was to call it the New Jersey Transit Bowl.

156
by TGT2 (not verified) :: Wed, 10/07/2009 - 3:11pm

You mised the point. It's the same stadium, so mentioning transit is silly.

175
by Red Hedgehog :: Wed, 10/07/2009 - 10:22pm

Well sure, but it's not like there's multiple games or the teams in the Super Bowl influence where it's held anyway. And New Jersey Transit is how anyone not driving would get to the game.

131
by Anon (not verified) :: Wed, 10/07/2009 - 12:26pm

Having learned the hard way after a long cab ride back to Hoboken, there's no rail anywhere near East Rutherford.

151
by Kevin from Philly :: Wed, 10/07/2009 - 2:32pm

I always wondered why they never put in trains between Manhatten and the Meadowlands. Seems like a no-brainer, based on all the complaints about traffic around the stadiums. Not like they don't spend on public transit up there.

155
by JasonK :: Wed, 10/07/2009 - 2:39pm

Rail service from Hoboken to the Meadowlands opened this year. (Hoboken itself can be reached from Manhattan via ferry or the PATH train.)

http://www.nj.com/hobokennow/index.ssf/2009/07/inaugural_train_ride_from...

84
by Rick A. (not verified) :: Wed, 10/07/2009 - 2:26am

To Vince and all those commenting on the DVOA analysis of the Vikings vs New Orleans : Minnesota is clearly a superior force to the Saints. Your concerns should be about the Giants if you are a Minny rooter. DVOA, I'm sure, will soon agree. In the meantime, it's only the end of week 4 and this quantification system that DVOA is doesn't really become relevatory, in my opinion, until after another couple of weeks or so...

152
by Kevin from Philly :: Wed, 10/07/2009 - 2:35pm

A little early to be crowing - remember, your season hangs by the thread of what's left of Bretts bicep tendons. Just recall how good the Jets looked last year at this time.

158
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 3:48pm

I agree that it's early to crow, but I'll also point out that the 2008 Vikings were good enough to go 10-6 with Frerotte/Jackson. I'd hate to see Favre get hurt, but Minnesota's defense plus Adrian Peterson equals a scary good team. The offensive weapons and their abilities in the open field would help Jackson/Rosenfels have good games, and the mobility of those two would mask the o-line's poor pass protection (last Monday not withstanding).

However, a Favre-less Minnesota team would have significant difficulty keeping up with New Orleans.

168
by Tundrapaddy (not verified) :: Wed, 10/07/2009 - 7:29pm

Rick A. and peterslappadabyass -

Please stop trying to hex my Vikings into a late-season slump and/or another putrescent playoff bow-out.

For shame, you two, for shame...

Actually, the part I'm worried about is regression along the defensive front. I think Pat Williams' tank is about empty, and opponents' inability to run (along with a strong pass-rush) is hat keys that defense. If teams don't have to worry about Pat Williams, they'll be able to focus more on containing Allen and Kevin Williams. That'd be a bad, bad thing for the secondary.

85
by Rick A. (not verified) :: Wed, 10/07/2009 - 2:28am

Make that "revelatory"...

99
by Mr Shush :: Wed, 10/07/2009 - 7:59am

I'm thinking that once opponent adjustments are fully in place, the Jets' defensive performance to date is going to look even more impressive than it already does. The Saints are probably the best offense in football, the Patriots in the top five and the Texans in the top ten. Those three units combined for nineteen points in their games against the Jets, and the Texans and Saints were at home. Their rookie QB may keep them from contending for a superbowl, and an extremely strong divisional rival could knock them down to a wild card spot, but the Jets are a very, very good team.

114
by Rich Conley (not verified) :: Wed, 10/07/2009 - 11:15am

"the Patriots in the top five"

The patriots may become a top 5 offense, and I think they will. They most certainly aren't one yet though, and weren't one when they played the Jets and Bills.

DVOA has no way to know that though.

102
by ebinary (not verified) :: Wed, 10/07/2009 - 10:08am

Aaron,

It would be really helpful if you labeled your tables.

Thanks!

P.S. Great stuff, BTW. This is probably the only system I believe is really well thought out.

107
by Aaron Schatz :: Wed, 10/07/2009 - 10:44am

Sorry about that. Done now.

110
by jimm (not verified) :: Wed, 10/07/2009 - 11:08am

Here's the Vikings season as I see it

Vikings season seems to break down schedule wise as

First 5 – very winnable
Next 3 – balt, @pitt, @gb
Next 6 – very winnable – det, sea, chic, @arz, cinn, @ car
Tough finish - @ chi, nyg

My prediction as follows

@ StL – W
Balt – L
@ Ptt – L
@GB – L
Det – W
Sea – W
Chi – W
@Arz – W
Cinn – W
@Car – W
@Chi – L
NYG – W

12-4….the 3 losses in a row will be depressing but a six game run will put them in the hunt for the number 2 seed. Beat NY to clinch the #2 seed and the Vikes are Super Bowl bound because we always beat the Saints and we don’t care that they are plus 225 pts and 15-1.

2nd Round – Beat the Eagles….all is good in the world – I knew it all along
NFC Championship – Minn 37 – NO 21 (Brees throws 2 ints, NO fumbles 3 times and Vikes get 2 return TD’s)

Super Bowl

Balt 34 Minn 7

Lets get real we're talking about the Minnesota Vikings

119
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 11:31am

Hosting Cincy and the Giants (if they have anything to play for) could be very difficult. I could see those games being losses, but even so that's 10-6 and likely one homegame in the playoffs. That said, hosting Seattle or going to Arizona seem like dangerous, if very winnable, games. Vikings need to be as healthy and effective at the end of the season as they've been so far, or it could all unravel quickly.

I don't want to have Minnesota be 0-5 on the big stage.

125
by Julio (not verified) :: Wed, 10/07/2009 - 12:02pm

The ranking of the Pats defense at 18th doesn't make much sense.
Right now they are 9th in the league in points per game allowed
at 17.8. If Brady hadn't fumbled the ball into the end zone against
the Ravens, it would be 16 pts/gm, tied for 6th with the Giants.

127
by anotherpatsfan :: Wed, 10/07/2009 - 12:15pm

Could be a lot of bending-but-not-breaking...

129
by Julio (not verified) :: Wed, 10/07/2009 - 12:18pm

Re #127:
Could be, but usually when that happens the yds/gm ranking is a lot
lower than the pts/gm ranking. Right now the Pats are ranked higher
(6th) in yds/gm than pts/gm (9th). That's partly because of those extra
7 points on the Brady fumble but even without it, they would be ranked
the same at 6th.

135
by Rich Conley (not verified) :: Wed, 10/07/2009 - 12:47pm

Don't think so.

I think a lot of it is weak opponent adjustments. DVOA seems to think the Patriots Rush defense is good, and Pass defense is TERRIBLE. (+20%)

Which seems strange to me. They held Flacco to 27/47 with 2 sacks and a pick, and their pass defense actually got worse? I think as the opponent adjustments get to full strenght, we're going to see that BAL's offense is really good.

The lack of turnovers at this point probably hurts too.

I just know, watching the games, that the defense is significantly better than it was last year.

144
by Dan :: Wed, 10/07/2009 - 1:49pm

Their per game stats are misleading because their opponents haven't had many possessions (fewest in the league, in fact), and they aren't getting turnovers. Look at the drive stats (which haven't been updated yet this week). Through 3 weeks, they were 13th in yards/drive, 13th in points/drive, 13th in DSR, and 25th in TOs/drive (they were 18th in VOA). The Baltimore game isn't included yet, but it barely changed their defensive VOA. They gave up a ton of yards and 14 points in 8 possessions but got 1 pick, which will improve their turnover numbers, worsen their yardage numbers, and barely change their scoring numbers.

128
by tuluse :: Wed, 10/07/2009 - 12:18pm

Has anyone noticed the huge range in rushing offense?

Dallas is #1 with 42.8% and Arizona is last with -42.1%

137
by JasonK :: Wed, 10/07/2009 - 1:09pm

A large range is to be expected with early-season results. More data will pull the outliers back toward the mean.

The oddity that struck me is how good the passing offenses are across the league-- only 9 teams have a negative passing offense DVOA.

136
by Rick A. (not verified) :: Wed, 10/07/2009 - 1:03pm

Hey Jimm, that is quite a crystal ball. Wow ! I'm going to make a zillion dollars playing all those games knowing ahead of time what is going to happen. I guess you will too. But what happened to the Colts and Giants ? Also, in case you and DVOA hadn't noticed, the Ravens had played nobody going into last week. Then they were promptly dispatched by the Pats. I mean, K.C., Cleveland, S.D. with no L.T., give me a break. But since you are convinced they will win the S.B. with their 2nd year QB I will regard that they are not really overrated so far, and take out a big futures play on them. It's just a shame that the Vikes will have to run into them in the SB. Too bad for the sentimental favorite, no 4, and all...

141
by jimm (not verified) :: Wed, 10/07/2009 - 1:44pm

Rick A - you do know I was being facetious. But feel free to bet heavily based on my little post - no charge for the advice.

154
by Rick A. (not verified) :: Wed, 10/07/2009 - 2:37pm

Yes, thanks for the free advice. Do you have anything on the baseball playoffs ?

157
by jimm (not verified) :: Wed, 10/07/2009 - 3:18pm

Rick A. - take Minnesota at 14/1 to win the American League pennant and 25/1 to win the series. Not a twins fan - but that seems like the best bet to me.

160
by peterplaysbass (not verified) :: Wed, 10/07/2009 - 3:52pm

They'd only have to have one set of confetti at the Super Bowl if it's Baltimore vs. Minnesota.

161
by Rick A. (not verified) :: Wed, 10/07/2009 - 4:40pm

To tell you the truth Jimm, if I were a futures player I would have taken a ticket on the Vikes to win the NFC and the SB after last week's, last second, miraculous throw and catch and I certainly would've taken a ticket out on the Twinkies this morning to rep the American League and to win it all. The idea that they should be such a long odds prop is ridiculous and one of the best plays to come along in a long, long time...

163
by KilsonFlob :: Wed, 10/07/2009 - 5:12pm

Are the drive stats going to be updated for this week?

165
by Paul A (not verified) :: Wed, 10/07/2009 - 5:52pm

Is anyone else bothered by the correlation between strentgh of schedule and ranking? It's hard for me to believe that no crappy teams have played an easy schedule and no good teams have played a tough schedule.

171
by Bowl Game Anomaly :: Wed, 10/07/2009 - 8:21pm

Because opponent adjustments are still small and each game counts for 1/4 of the total, the best and worse teams make their schedules look easier and harder, respectively. A top team will hammer its opponents making the schedule look weak. A lousy team will inflate its opponents' performances making them look strong. Once opponent adjustments get higher and more games are counted, this effect will diminish.

166
by jimm (not verified) :: Wed, 10/07/2009 - 6:14pm

Rick A - I agree. I'm not a better but two times in my life I saw odds I thought were ridiculous and made bets. Once right - once wrong

1 - 1987 Vikings were I think 12.5 underdogs to the Niners...I thought Minnesota was a better team...Anthony Carter bless his heart - my favourite Viking win ever

1994 - Chargers were a 18 point underdog or something nutty like that...there was no way I could lose.........niners won by 23

174
by whodat :: Wed, 10/07/2009 - 10:10pm

The introductory text and a number of posts make a good point about not confusing descriptive statistics with predictive statistics. I forget what-all goes into the mix, but just to pick on two discussed teams:

1) Saints: doesn't G Williams have a history of his defenses starting out strong then degrading? Do the models therefore plug in some sort of exponential decay function -- or even account for coaching?

Also, it seems like the defense's performance thus far has been more dependent on a single player -- Sharper -- than you would see in many / most elite units. What were the with-vs-without Bob Sanders DVOA nums for Colts D 2006? Not saying Sharper's going down to injury or at risk of performance degradation, but -- to raise another question -- where does risk show up in the DVOA? Are st devs, t-stats, etc posted somewhere?

2) Vikings: similarly, does anyone think the Vikings' offensive performance thus far represents their expected performance going forward -- specifically: that the probability of Favre remaining healthy (not just IR stuff but "walking wounded" problems like dead arm) is meaningfully close to 1?

And thinking along the lines of a probabilistic injury function for Favre, can this even be modeled given Favre's age? Doesn't that fact render such efforts if not nonparametric -- I guess there's DeBerg, Vinnie, and Moon, plus Favre 2008 -- then at least statistically unreliable owing to such a small sample size?

I'm sure the answers to these sorts of questions have been answered more than once, but how much of this stuff gets factored?

177
by Jerry :: Thu, 10/08/2009 - 2:32am

The preseason numbers are derived from many factors, but the in-season numbers are just based on what's in the play-by-play.

179
by ammek :: Thu, 10/08/2009 - 8:18am

Saints: doesn't G Williams have a history of his defenses starting out strong then degrading? Do the models therefore plug in some sort of exponential decay function -- or even account for coaching?

No, and for good reason. Your assertion about Williams' worsening defenses is subjective — even if it's true (and off the top of my head Tennessee 1999 and Washington 2007 are counter-examples). Can you be sure the declines were directly attributable to Williams, rather than to roster depth, or injuries, or schedule, or luck? And do you have any proof that, simply because Williams' defenses have declined in the past, they will continue to do so? Remember "Dungy / Manning / Madden have a history of choking in the playoffs" — until the season they didn't.

Vikings: similarly, does anyone think the Vikings' offensive performance thus far represents their expected performance going forward

Of course not. Even if Favre stays perfectly healthy, he won't continue at a rate of one turnover every four games, with a 70% completion percentage.

The question should be: is Minnesota equipped to win games when its QB (be it Favre or Jackson) isn't playing spotlessly? The FO projection is tepid.

Also, the likelihood of Favre getting hurt may be higher than average because of his age, but you could argue that it's higher still for NFC North rivals Aaron Rodgers (he of the 20 sacks and counting) or Matt Stafford (already hurt). Jackson is one of the better backup QBs in the league — if all four starting QBs in the division got hurt, who would you prefer to rely on: Jackson, Culpepper, Matt Flynn or Caleb Hanie?

180
by whodat :: Thu, 10/08/2009 - 7:10pm

Good comments. I should preface this by saying I don't have a degree in statistics nor have I studied the FO models too closely. So take this post more in the form of a question rather than a declarative.

Yes on my Williams comment being subjective but as with Favre it sounds like the potential issue in both cases is nonparametric -- well, in Williams' case it could be a case of specification error: unless Williams or def coords are specified as variables their impact, such as it is, will show up in other variables. Should DefCoord be in there? I have no idea.

In the case of Favre, I wasn't trying to get at the probability of his getting injured so much as I was trying to point out there just aren't that many 40+ y.o. QBs upon which one could model age as a factor.

Honestly I have no idea if things like this belong in the model -- when you lard a model with too many variables you run into a whole nuther set of issues.

181
by DaveRichters (not verified) :: Sat, 10/10/2009 - 2:05pm

You say:

"No, and for good reason. Your assertion about Williams' worsening defenses is subjective — even if it's true (and off the top of my head Tennessee 1999 and Washington 2007 are counter-examples). Can you be sure the declines were directly attributable to Williams, rather than to roster depth, or injuries, or schedule, or luck? And do you have any proof that, simply because Williams' defenses have declined in the past, they will continue to do so? Remember "Dungy / Manning / Madden have a history of choking in the playoffs" — until the season they didn't."

I say:

That's a great criticism, but I think it is actually a criticism of DVOA. Variables that are given weight in this model are chosen simply because of a high correlation, without being theoretically motivated.

185
by SKD (not verified) :: Sun, 10/11/2009 - 11:57am

You can use the sample size of 19 years in the L to know that Brett Favre aint going down anytime soon. Ask Doug Peterson how many starts he got in GB. Injury can not be predicted at the QB spot. Unless you are Matt Schuab.