Writers of Pro Football Prospectus 2008

Most Recent FO Features

HarrisChr11.jpg

» Film Room: Chris Harris

Is Harris one of the league's top cover corners, or a product of the system in which he plays? Cian Fahey says the answer lies somewhere in the middle.

03 Sep 2008

2008 DVOA Projections

by Aaron Schatz

Before I get into the 2008 DVOA projections, I thought this was a good place to talk about the upcoming season and some of the changes you will be seeing here at Football Outsiders.

We're happy to announce that ESPN will be our exclusive (almost) online partners in 2008. We will providing five or six columns per week to ESPN during the season. Changes may still take place, but as of now those columns are:

  • Quick Reads
  • Any Given Sunday
  • an unnamed fantasy column by Bill Barnwell
  • Numbers Crunching
  • a feature related to each week's Monday Night Football matchup

We're hoping things work out to put Mike Tanier's Walkthrough on ESPN as well. Various FO writers will be doing weekly chats on ESPN.com, and we have the prime hour of 8 p.m. Eastern on Monday nights, right before Monday Night Football (well, except this week, when our chat is in the middle of the first MNF game). I'll also be appearing regularly on the Bill Simmons "B.S. Report" podcast.

The word "almost" above refers to the fact that ESPN is our exclusive partner among large sports sites. However, FO will be providing content to the websites of various newspapers, including the New York Sun, Washington Post, Seattle Times, and Dallas Morning News.

With all these changes and all the growth we've had in our first five years, it is getting harder and harder to run things myself. Therefore, I am proud to announce the promotion of Bill Barnwell to Managing Editor of Football Outsiders. He has left his full-time job at IGN and will become the second person to work on Football Outsiders full-time. Bill will deal with many of the business aspects of FO (like advertising) so that I can devote more time to research. He will also work on expanding the content in FO Premium. We're doing a lot in FO Premium now and in the upcoming months. Some of the additions will include:

  • Weekly picks against the spread, derived from a formula that combines DVOA splits, weather forecast, and injury reports. We tested this formula out in the second half of last year. Picks in the "Green" and "Yellow" categories were correct 60 percent of the time. (Picks in the "Red" category were correct a less-exciting 50 percent of the time.) The first week's picks are already up in Premium.
  • Additional team views of DVOA splits, including the oft-requested "what was Team X's DVOA through Week Y in Year Whatever" table, and the offensive and defensive ALY/ASR tables.
  • Player pages for offensive players (with DVOA and DYAR for each season all in one place) and defensive players (with all our defensive stats like Stop Rate and Defeats going back to 2000).
  • A fantasy football question-answering service, where Bill will respond to any fantasy football lineup and trade questions from Premium subscribers within 24 hours.

FO's plan for growth also includes expanding our coverage of college football, and we're proud to welcome Bill Connelly to the FO team. Some of you may know his "Beyond the Box Score" columns from Sunday Morning Quarterback, which looked at college football using the kind of play-by-play breakdown that Football Outsiders has been using on the NFL for five years. With the main man behind SMQ heading off to Yahoo, Bill moves over to FO. His new column will be called "Varsity Numbers" and debuts this Friday. In addition, Brian Fremeau's column will run weekly throughout the season, rather than waiting until the initial BCS ratings debut in mid-October.

Comings are often joined by goings, and things are no different when it comes to the staff of Football Outsiders. Of particular note, Michael David Smith will be moving on from Football Outsiders after four years. Mike was the first new writer added to the site after our launch in 2003, and he was a big part of our growth those first couple of years. He's gone from one little unpaid column per week on FO to appearances on what seems like every website on the planet (starting, of course, with AOL Fanhouse, Pro Football Talk, and College Football Talk). I'm proud to we helped him get his start. Ryan Wilson is also leaving, to concentrate on his role at AOL Fanhouse, and Stuart Fraser, one of our new writers for PFP 2008, is leaving to concentrate on his studies. We wish them all the best and of course we hope they'll stick around the FO discussion threads.

MDS is taking Every Play Counts with him, but we'll be introducing a new column this year by Doug Farrar called "Cover-3," which will take a similar scouting-oriented look at three different players or units each week.

Unfortunately, the biggest change at Football Outsiders isn't happening... yet. For a couple months, we've been working on a major redesign. It was supposed to debut on Monday, but it is simply not quite ready yet, and we didn't want to rush our product out there before it had been properly vetted. As of now, the redesign is scheduled to launch in mid-September. The changes include:

  • Updating the look of Football Outsiders
  • Improving ease of navigation
  • Dramatically improving our currently almost-useless search feature
  • Improving discussion boards with a "slashdot"-style model that will still allow for commenting without registration
  • Better merging our store, our open discussion threads, and our articles
  • Most importantly, a massive overhaul of our back-end server structure that should hopefully put an end to those annoying FO server slowdowns once and for all

Pray for that last one. Seriously.

* * * * *

One more plug: Make sure you get in on the fifth annual Loser League contest, your chance to pick the worst NFL players of the season. Go here for rules and to pick your team. The deadline has been moved to Saturday. Yes, that means if you feel like it, you can wait and figure out who plays badly for the Giants or Redskins, but remember: There's no guarantee those players will continue to suck the next eight weeks.

* * * * *

Now, onto the projections, with the requisite link to an explanation of DVOA. For anyone new to our site, it stands for Defense-adjusted Value Over Average and measures a team's performance on every play of the season compared to league average in the same situation, adjusted for opponent. I know a lot of people are coming here from various message boards and this is just going to look like a jumble of pointless numbers. Trust me, there is a method to the madness, and over the past eight seasons DVOA has been a far more accurate predictor of future performance than wins or points.

Offense, defense, and special teams DVOA are all projected separately using a system based on 2000-2007 numbers. The equations include a number of variables based on performance over the past two seasons in different splits (by down, passing vs. rushing, red zone vs. whole field) plus variables based on recent draft history, injury history, offensive and defensive pace, coaching experience, quarterback experience, and even weather. Strength of schedule was then figured based on the average projected total DVOA of all 16 opponents for 2008 (yes, projected performance, not 2007 performance).

The projections here are updated from Pro Football Prospectus 2008 based on changes in some of the variables, usually related to injuries, offensive line continuity, and quarterback experience. There are no manual adjustments. The numbers we are presenting here are exactly what the projection system spit out. A few of them will look strange to you. A few of them look strange to us. There's no reason to really regurgitate the reasons why our projections for some teams (Dallas, Cleveland, Baltimore) differ so greatly from the conventional wisdom -- we've gone over it in numerous articles and in numerous promotional interviews over the past few weeks.

There is one projection that does need some explanation, however, because it is brand new. I've been hyping up Houston as a potential playoff team, despite the tough schedule that comes from playing in the AFC South. I wrote the cover story in the latest issue of ESPN the Magazine that has Mario Williams on the cover. So imagine my shock when I went through and re-did our projections based on changes in the expected starting lineups of each team. Houston ends up with a much younger lineup than we originally projected, primarily on the offensive line. The Texans will be starting first-round pick Duane Brown over veteran Ephraim Salaam at left tackle, and an injury means right guard Fred Weary (age 31, seventh year in Houston) is now replaced with Mike Brisiel (age 25, third year in Houston). Because of those changes, the Houston offensive projection has dropped enough to move the Texans to a projected record around 8-8. I hate to do this because I've been driving the bandwagon, and I don't like to create a family squabble, but I've got to go with the numbers, because that's what we do around here. The official Football Outsiders projections for 2008 give the second AFC wild card spot to Denver, not Houston.

We know that there are a couple of teams whose projections in offense and defense are far beyond the rest of the league. That's the system expressing a high amount of confidence.

Other than Houston, the team whose projection has changed the most since the book is the New York Giants. I assume I don't have to explain that one.

The projected wins are going to be a little different from those you found in PFP 08. Rather than being based on Ben Alamar's simulations, these are based on the simulations we run to figure the DVOA Playoff Odds each week. The first version of those odds will be up soon, sometime this evening. The main difference between the two simulations is that the playoff odds equations end up dragging every team even closer to the mean of 8-8.

Projected division champions are colored in light yellow. Projected wild card teams are colored in light blue.


TEAM TOTAL
DVOA
TOTAL
RANK
PROJ
WINS
OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
SCHED SCHED
RANK
NE 40.0% 1 12.0 39.8% 1 4.2% 20 4.4% 3 -7.3% 32
PHI 32.3% 2 11.1 24.8% 3 -9.0% 5 -1.5% 24 -5.1% 31
GB 31.5% 3 10.4 11.6% 8 -17.5% 2 2.4% 6 2.4% 13
SD 25.5% 4 10.7 20.2% 4 -2.6% 13 2.6% 5 -5.0% 30
MIN 22.5% 5 9.9 11.6% 7 -8.6% 7 2.3% 7 1.8% 14
IND 21.7% 6 9.4 28.9% 2 2.5% 18 -4.6% 32 7.0% 1
SEA 19.3% 7 9.9 16.1% 6 -3.0% 12 0.2% 17 -2.2% 25
TB 18.8% 8 9.8 6.7% 11 -9.9% 4 2.2% 8 -2.2% 26
JAC 16.0% 9 9.3 19.4% 5 5.4% 22 2.1% 9 2.7% 9
DAL 10.7% 10 8.9 4.6% 12 -6.2% 8 -0.1% 19 3.5% 8
PIT 5.6% 11 8.1 -9.4% 24 -13.7% 3 1.3% 12 6.9% 2
NYG 5.2% 12 8.3 11.4% 9 6.6% 24 0.4% 16 2.6% 10
BAL 3.1% 13 8.3 -25.5% 31 -24.1% 1 4.6% 2 0.2% 16
CAR 2.9% 14 8.4 0.4% 16 -5.5% 9 -3.0% 29 -1.9% 23
HOU 1.2% 15 7.9 -0.3% 17 0.2% 15 1.7% 10 2.5% 11
DEN 1.1% 16 8.4 0.8% 13 -3.3% 11 -3.1% 30 -3.5% 27
TEAM TOTAL
DVOA
TOTAL
RANK
PROJ
WINS
OFFENSE
DVOA
OFF.
RANK
DEFENSE
DVOA
DEF.
RANK
SPECIAL
DVOA
S.T.
RANK
SCHED SCHED
RANK
NO -0.4% 17 8.2 0.8% 14 -0.5% 14 -1.7% 26 -4.2% 29
ARI -2.2% 18 7.9 -9.7% 25 -8.8% 6 -1.3% 22 -1.0% 20
CHI -3.0% 19 7.0 -9.9% 26 1.6% 16 8.5% 1 5.1% 3
CIN -3.0% 20 7.3 8.5% 10 12.1% 27 0.6% 15 2.4% 12
TEN -3.8% 21 7.0 0.6% 15 3.0% 19 -1.4% 23 4.6% 5
NYJ -5.9% 22 7.7 -4.5% 20 2.2% 17 0.8% 14 -3.7% 28
WAS -10.3% 23 6.6 -15.0% 29 -4.6% 10 0.1% 18 1.5% 15
CLE -14.2% 24 6.1 -8.0% 22 4.6% 21 -1.6% 25 4.4% 6
BUF -15.6% 25 6.6 -9.4% 23 10.6% 26 4.3% 4 -1.9% 22
KC -17.6% 26 6.3 -10.5% 27 5.8% 23 -1.3% 21 -1.7% 21
MIA -22.3% 27 6.0 -3.7% 19 16.6% 30 -1.9% 28 -0.9% 19
DET -22.7% 28 5.2 -7.4% 21 13.4% 28 -1.9% 27 4.8% 4
OAK -24.8% 29 5.7 -2.5% 18 17.9% 32 -4.3% 31 -0.1% 17
STL -28.4% 30 5.7 -18.0% 30 9.5% 25 -0.9% 20 -2.1% 24
SF -28.8% 31 5.4 -13.5% 28 16.8% 31 1.4% 11 -0.7% 18
ATL -41.7% 32 4.9 -27.7% 32 15.2% 29 1.2% 13 3.7% 7

Tomorrow, we'll have our annual look at subjective staff predictions for 2008.

Posted by: Aaron Schatz on 03 Sep 2008

128 comments, Last at 11 Sep 2008, 4:52pm by NE fan

Comments

1
by Temo (not verified) :: Wed, 09/03/2008 - 7:59pm

I'm so glad you waited until I left work to post this. +1 for productivity.

2
by John Z (not verified) :: Wed, 09/03/2008 - 7:59pm

The Eagles are projected as the second best team in the league, yet the team that has made the playoffs 3 years in a row and won the Superbowl last year won't make the playoffs...Do the Giants ever get any respect?

3
by Crushinator (not verified) :: Wed, 09/03/2008 - 8:03pm

Why do the FO projections always love Philly and Tampa?

4
by KJT (not verified) :: Wed, 09/03/2008 - 8:04pm

I am stunned that Cleveland is projected as the worst team in the AFC North. Cincinnati is looking like a complete trainwreck again this year, on both sides of the ball (as opposed to just defense like in years past). I don't mind the Ravens projection nearly as much, even though I am a Steeler fan.

5
by Some Dude (not verified) :: Wed, 09/03/2008 - 8:17pm

So, the Giants are the new Patriots: "We Don't Get No Respect!" Cheating scandal follows in 2013, i guess.

6
by Temo (not verified) :: Wed, 09/03/2008 - 8:18pm

Quick note though: If the Cowboys go 11-5, you're not allowed to tout that your system predicted a Cowboy regression.

7
by Bomb Ann Arbor Now (not verified) :: Wed, 09/03/2008 - 8:21pm

Is there a rough rule of thumb for home field advantage in terms of DVOA?

8
by TBW (not verified) :: Wed, 09/03/2008 - 8:22pm

Re 2: Perhaps it's strength of schedule. The Eagles have the 2nd easiest schedule, the Giants the 10th hardest.

9
by Insancipitory (not verified) :: Wed, 09/03/2008 - 8:25pm

I think the FO system likes the West Coast Offense so much is because it's predicated on a lot of lesser value successful plays. Over a season, that's a lot of successful datapoints. 40 yards in a West Coast offense might consist of a vastly larger number of successes than a system that relies more on say "unleashing the dragon." One can easily imagine a situation which a WCO team and a UtD team have the same number of yards, same number of TDs, and perhaps even same number of offensive series, but which one will have the larger proportion of successful plays? More over since the WCO team has more plays in general, the system would be more confident in that prediction.

In that respect WCO's might be over valued. But I would also argue that WCO's are more valuable even in that circumstance. After all the WCO's are creating more opportunities for a mismatch or breakdown to unleash their own dragon. By that same token they also increase the risk that supelative defensive play or butterfingers might strike. But if you're a mostly successful team, that additional opportunity is worth something. Maybe DVOA "knows" that, maybe it over values it.

10
by Temo (not verified) :: Wed, 09/03/2008 - 8:26pm

8. DVOA is, by definition, adjusted for strength of opponent. Adjust wins are not (which is why Indianapolis is projected to win fewer games than the next two teams).

11
by Josh (not verified) :: Wed, 09/03/2008 - 8:27pm

Why do the FO projections always love Philly and Tampa?

The West Coast Offence is very DVOA-friendly. Much like I expect a NCAA DVOA would like Mike Leach's spread.

I think DVOA might over-value WCO's and under-value deep passing (the REAL West Coast Offence, see Dr Z) like the Giants, Browns, and Cowboys.

12
by Josh (not verified) :: Wed, 09/03/2008 - 8:28pm

Is there a rough rule of thumb for home field advantage in terms of DVOA?

I think it's around 34% or so.

13
by Mike W (not verified) :: Wed, 09/03/2008 - 8:29pm

It appears DVOA feels the NFC has achieved parity with the AFC. Just noticing.

14
by Josh (not verified) :: Wed, 09/03/2008 - 8:31pm

As a Titans fan, this saddens me. DVOA is obviously biased against Jack Daniels, Country Music, and drunk redneck trailer trash.

15
by Marko (not verified) :: Wed, 09/03/2008 - 8:31pm

2: Who cares how your team is projected? Your team won the Super Bowl last year. Be happy. They don't give out hardware based on projections, DVOA, power rankings, etc. Maybe the Giants can use the lack of respect they receive as a motivating force. If so, they certainly wouldn't be the first team to do so.

At the end of the year, if the Giants beat out the Eagles, you can just point to the standings.

16
by Temo (not verified) :: Wed, 09/03/2008 - 8:33pm

10. Meant "projected", not "adjusted".

9. While many of the teams that DVOA rates above conventional wisdom seem to be WCO teams, it seems to be a red herring. I don't see how a preference for WCO teams would lead to greater projections in the future.

17
by shake n bake (not verified) :: Wed, 09/03/2008 - 8:34pm

Why so down on the Indy D?

They were 3rd last year and return every starter and most of the backups (only 5 of the 25 defenders to appear in at least 8 games are no longer with the team, only Brandon Condren of those 5 is currently on an NFL roster, and the Rams D hardly counts).

They added a big dose of pass rushing depth with their rookie class. Marcus Howard and Curtis Johnson on the edges and Eric Foster as an undertackle. Plus 2nd year player Daniel Muir to give a Ed Johnson a legitimate backup as the over-tackle.

It can't be injuries, the Colts were the most injured team in the league last year with a good amount of that falling on the defense.

Is the awful 2006 regular season D still effecting the projection?

Some 3rd down or Fumble recovery trends suggesting a fall to earth?

Anything I'm missing?

18
by the silent speaker (not verified) :: Wed, 09/03/2008 - 8:48pm

How come there's such a huge gap between the Eagles' strength of schedule and the Giants'? They play 12 out of 14 of the same teams, and those two games push the Giants 22 slots ahead?

19
by jsa (not verified) :: Wed, 09/03/2008 - 8:49pm

Could you tell us where MDS is taking his Every Play Counts column?

20
by Jerious Norwood for President (not verified) :: Wed, 09/03/2008 - 8:53pm

Can you just leave Atlanta off the chart? I can't really handle the thought of this year being worse than last.

21
by MJK (not verified) :: Wed, 09/03/2008 - 9:02pm

One comment on readjusting projections. You said:

Houston ends up with a much younger lineup than we originally projected, primarily on the offensive line. The Texans will be starting first-round pick Duane Brown over veteran Ephraim Salaam at left tackle, and an injury means right guard Fred Weary (age 31, seventh year in Houston) is now replaced with Mike Brisiel (age 25, third year in Houston).

I get re-doing projections if there is an injury--you assume the replacement is weaker than the guy that would have been your starter (although I wonder about the drop associated with a 31 year old to a third-year 25 year old...it almost seems like that would be an improvement. Is there evidence that age on the O-line leads teams to perform better the following year?).

But part of the Texan's younger O-line, and hence part of their drop, is because a rookie beat a veteran out for the starting job? If you think a team will be X good, and a rookie outperforms one of the players that you expected to contribute to X such that he wins the starting job, shouldn't the new projection be X+ good?

22
by Matt (not verified) :: Wed, 09/03/2008 - 9:06pm

Defensive DVOA in general seems a couple seasons behind, not just for the Colts. I'll take the under that Baltimore doesn't match that number, and I'll be stealing your money. Or I could take the over on Tennessee. New England's older, sure, but I'd still feel pretty comfortable taking the over - though I guess their opponents this year are even weaker than last.

I'm biased, but the Colts definitely seem off in Defensive DVOA the most. This chart's predicting a drop off of over 13 percentage points, despite no significant changes in the roster, and Dwight Freeney was out for most of 2007. I guess DVOA is expecting exponentially more injuries this season??

23
by Temo (not verified) :: Wed, 09/03/2008 - 9:11pm

Well since Aaron didn't do it, for whatever reason, here's a look back at 2007:

The correlation between 2007 Predictions and 2007 actual DVOA was 0.582-- enough to say it wasn't random variation, but (for me, anyway) not quite good enough to hang my proverbial hat on. Predicted DVOA missed most on San Fran (overestimated by 36.5%), Dallas (underestimated by 33.7%) and Carolina (overestimated by 31.8%). It hit most with New Orleans (off by 0.6%), Buffalo (off by 0.9%) and Jacksonville (1.0%).

The absolute value of the difference between predicted and actual DVOA in 2007 had a mean of 13.6% with a standard deviation of 9.7%.

24
by Mig (not verified) :: Wed, 09/03/2008 - 9:20pm

One would think that after last season, FO would have learned the lesson to not understimate the Dallas Offense.

As usual TB and PHI are over rated, specially the aging Bucks.

Other than that, even thou I expect a rebound by Baltimore's DEF, kinda surprising they are projected #1.

Last, amazing that halt the projected Play Off teams have a schedule ranked # 25 or easier. So much for the NFL preaching parity !

25
by mikeabbott (not verified) :: Wed, 09/03/2008 - 9:22pm

slashdot ?
really ?
meta moderation?
I'm not sure you have a big enough user base for that.

26
by Quentin (not verified) :: Wed, 09/03/2008 - 9:22pm

plus variables based on... offensive and defensive pace...

Just out of curiousity, how does pacing factor into the equation? Fast= good for offense, bad for defense? And is there any type of consistency in defensive pacing from year to year?

Why do the FO projections always love Philly and Tampa?

The system doesn't necessarily always love Tampa Bay, it just seems to have a better read on them than anyone else. If memory serves, it projected success for '05, serious decline for '06, and then success again for '07. It just seems to have them pegged.

Regarding Philly, I'm in agreement with the system that Philly is a dangerous team, and has been for years. The only time in the last 5 years or so that I'd consider them to be just plain bad was the year McNabb got hurt after falling out with T.O.

27
by Quentin (not verified) :: Wed, 09/03/2008 - 9:27pm

21: Maybe it's lack of experience + lack of continuity.

24: These are just raw numbers spit out by the projection system. You can cry about your team and bash the writers tomorrow when they give their subjective opinions.

28
by mikeabbott (not verified) :: Wed, 09/03/2008 - 9:32pm

And congratulations to Mr. Barnwell.
I expect a full time Bill Barnwell to be a very good thing for the site.

29
by Temo (not verified) :: Wed, 09/03/2008 - 9:49pm

So I looked at 2006 predictions now. There was considerable improvement from 2006 to 2007 in terms of predictions.

The correlation between 2006 Predictions and 2006 actual DVOA was 0.358. The absolute value of the difference between predicted and actual DVOA in 2006 had a mean of 15.8% with a standard deviation of 11.5%.

In terms of consistently being wrong on certain teams, there was no indication of that. Correlation between the absolute value difference between 2006 and 2007 was -0.21 (in other words, just about no correlation at all). But for the two years, the System was most wrong on Baltimore, Seattle, and the NYJ (the Jets are the only team to differ by more than 20% in DVOA both years). They were wrong on Seattle by an astounding 53.2% in 2006 (they thought Seattle would be the best team in the league-- it was in the bottom 10). It was most right about Arizona, NYG, and Cincinnati.

30
by Zac (not verified) :: Wed, 09/03/2008 - 9:49pm

Re: 18. Don't forget the two games they play against each other as well. So that's 12 common opponents. Then Philly plays the Giants twice (ranked 12th), and then Chicago and Atlanta (19th and 32nd). The Giants, meanwhile, play Philadelphia twice (ranked 2nd), then play Carolina and Minnesota (5th and 14th).

So 2x 12, 19, and 32 are worth the 31st schedule.
2x 2, 5, and 14 is worth the 10th schedule. It seems a little surprising, but perhaps the teams at the extremes are weighting the average down or up.

31
by Dave (not verified) :: Wed, 09/03/2008 - 9:51pm

Will we still be getting EPC in the Extra Points?

32
by Led (not verified) :: Wed, 09/03/2008 - 9:56pm

Congrats on continued growth of the franchise, Aaron. One request on the redesign -- is it possible to make the site more accessible by blackberry? I'd love to be able to visit while commuting or otherwise on the road, but the site is darn near impossible to navigate remotely right now.

33
by Raiderjoe (not verified) :: Wed, 09/03/2008 - 9:57pm

Dvcoa really wrong.
No way aare Raiders 29th best team in league. More than 9th best team. Lets see team resigneed T Kelly. Team get D Hall, G Wilson, Ashle y Lelie, draft Chaz Schilens, draft Darren McFadden, Jamarcus Russell in 2nd year and was in camp this summer, Z Miller great up and cominger at TE, T Branch great backup defensive back guy, coaching staff one of best ib league, team plays in same dsivision as Chiefs and Broncos, so that is 4 easy wins right there. raiders always play good sv Sd Chargres, so call that a split for 2008 season.
Raiders 11-5
Chargers 11-5
Pates, browns, Colts, and other team will make the Afc playoff.s

dvaa to be really embararsed again

34
by Temo (not verified) :: Wed, 09/03/2008 - 10:03pm

26. Philadelphia:

2007 Predict:26.0%
2007 Actual: 7.7%

2006 Predict: 12.0%
2006 Actual: 24.2%

And in 2005... well the predictions are in a strange format, but it did include this gem: It’s Philadelphia’s world and we’re just living in it. They are the only team with more than ten projected mean wins. They are the only team that projects to be in the top ten with all three units. They have one of the league’s easiest schedules. As we noted in the book, the DVOA win projection system gives Philly a 4.6% chance of a perfect season.

Of course, Philly finished 2005 with a 6-10 season-- in the "TO" year.

35
by Temo (not verified) :: Wed, 09/03/2008 - 10:06pm

I'm not sure, but I may have done the correlation between 2006 predict and 2006 actual DVOA wrong. But I have to go to sleep now and wake up early in the morning, so bleh.

36
by Crosseyed and Painless (not verified) :: Wed, 09/03/2008 - 10:59pm

I refuse to let your "projections" ruin the only enjoyable week of being a Lions fan.

37
by TomC (not verified) :: Wed, 09/03/2008 - 11:13pm

And for the third year running, here are my sure-to-be-wrong, nobody-asked-for-them final standings predictions. I pick the winner of every game every week then tally them up, so the final win totals are often quite a surprise. Never more so than with Seattle this year. I must have been sipping some serious cane cola when I did that. And yes, that's Carolina winning the NFC South with a losing record.

AFCE:
NE 13-3
NYJ 7-9
BUF 5-11
MIA 4-12

AFCN:
PIT 9-7
BAL 9-7
CIN 6-10
CLE 6-10

AFCS:
IND 11-5
JAC 10-6
TEN 9-7
HOU 7-9

AFCW:
SD 14-2
DEN 9-7
OAK 6-10
KC 5-11

NFCE:
PHI 11-5
DAL 11-5
NYG 8-8
WAS 8-8

NFCN:
GB 11-5
MIN 9-7
CHI 9-7
DET 5-11

NFCS:
CAR 7-9
ATL 6-10
TB 6-10
NO 5-11

NFCW:
SEA 15-1
ARI 6-10
SF 5-11
STL 4-12

DIV: NE, PIT, IND, SD, PHI, GB, CAR, SEA

WC: JAC, TEN, DAL, MIN

38
by Raiderjoe (not verified) :: Wed, 09/03/2008 - 11:15pm

Loisn have good chance to start off season 1-0 . So you get at least one more enkoyable week after this one.
Lets see , if Liosn beat Packers at Detroit in week 2, tream would go to 2-0. very possible as Rdogers is not very good. M Flynn could be godd but wont happen till 2009 season, Brian Broom just suck, couldn't even beat out Flynn for #2 job, but maybe because Flynn is really good, he might be. not as good as former LSU great Jamarcus Russell, but playing behind Russell and learning from Russell had to help Flynn.

Then Loisn go to Sf for Week3 game. should be a win for them too. 3-0 start not out of questuins. Then bye.
Then home game vs Beras. That could be win too. Loins could definitely be 4-0 team then they start to lose. ebecausue have games at Minnesota and Houston

39
by JasonK (not verified) :: Wed, 09/03/2008 - 11:18pm

#15: Exactly.

Expecting a wholly rational projection system to conform to your hopes and dreams for your favorite team is just silly. If the '07 Giants proved anything, it's that it is a hell of a lot more fun when your team defies all rational projections.

I just watched the premiere of the 2007 "America's Game." We Giants fans have that forever. (Of course, this is not to say that I don't still enjoy reading FO-type rational analysis. I just try to keep the "raving fanboy" and "quiet scholar" parts of my brain at a safe distance from each other.)

Anyhow, one number that jumps out to me: Pittsburgh's Offense. From 12th in '07 all the way down to 24th in '08? I know you're expecting the OL to be bad, but it's not like it was a team strength last season.

40
by Bugs Meany (not verified) :: Wed, 09/03/2008 - 11:30pm

I think the blatant Jeff Garcia bias in the DVOA system needs looking into. Might be some kind of bug in the code.

41
by Joe in Seattle (not verified) :: Wed, 09/03/2008 - 11:33pm

As a Seahawk fan I fully endorse the 15-1 prediction.

42
by sam! (not verified) :: Wed, 09/03/2008 - 11:36pm

23:

Since Carolina couldn't find enough QBs, I don't think it's fair to say DVOA "missed" on the Panthers.

43
by Drew (not verified) :: Wed, 09/03/2008 - 11:41pm

Every bias I had against Raiders fans has just been confirmed.

44
by TomC (not verified) :: Wed, 09/03/2008 - 11:50pm

#40 - Encyclopedia Brown reference. I'm speechless.

45
by Bobman (not verified) :: Thu, 09/04/2008 - 12:00am

Time for my fifth anual (probably 4th really) bashing of the DVOA projections regarding the Colts.

I second Shake n Bake's queries at #17--what is the rest of the NFL world missing--even the folks who predict the Jags to win the AFCS would not predict fewer than 11 wins for Indy. And adding D players to a highly ranked D somehow diminishes it. Did half the league improve so much more on D that they surpassed one of the top teams? That seems highly implausible. I am willing to accept it if there's a cogent explanation.

The simplest question, and one that is unanswerable because of the way this is generated, is which 6+ games would Indy lose? I know the only way their sked could be tougher is if they have to play themselves (waah!), but I just don't see more than 4 likely losses in the mix.

But the projection system always tends to squash teams toward 8-8. The Pats win total looks low as does SD's (esp with those 32 and 30th ranked skeds! The only way those skeds would be easier is if they played with themselves! Waah! Thank you thank you, I'm here all week.) And we know the worst records will be below 5 wins. There will probably be 3-4 teams with 3-4 wins each, as opposed to all those 5-win teams at the bottom. So in reality there should be fewer wins to spread among the bottom teams and more for the to teams. Which is pretty consistent here year-over-year.

So while the relative rankings make sense (except for Indy's D), the win totals have always looked flukey to me.

46
by Josh (not verified) :: Thu, 09/04/2008 - 12:17am

#40 - Encyclopedia Brown reference. I’m speechless.

As am I.

username of the year, perhaps?

47
by Rich Conley (not verified) :: Thu, 09/04/2008 - 12:34am

I love the "patriots are old" meme.

Has anyone actually looked at the age of patriots starters?

48
by Tom D (not verified) :: Thu, 09/04/2008 - 12:44am

I'm a little surprised at Chicago's defense ranking. Does the projection system just not think they are very good, or does it think they will have a lot of injuries?

49
by Boston Dan (not verified) :: Thu, 09/04/2008 - 12:55am

Going to miss it!

Every Play Counts

50
by MossyCade (not verified) :: Thu, 09/04/2008 - 1:06am

First, just when I was getting comfortable with dvoa and DYAR, Rjoe springs "dvcoa" on the world (I'll need an explanation of that one, too-and slowly).

Then he hits hard with "up and cominger". "And some others" was my catch phrase on this site for August. September's is "up and cominger".

Finally, he finishes off with not one, not two, but THREE references to the "Liosn." And then spells the fourth reference "loins".

Happy Football Season to you, Raiderjoe! I'm eagerly anticipating your commentary during the next several weeks.

51
by Greg (not verified) :: Thu, 09/04/2008 - 1:14am

I'm surprised -- neh, shocked -- that DVOA sees a total regression for the Raiders' defense over last season. I mean, what (realistically) is the worst case scenario for the 2008 Raiders' pass D? 2007 Broncos, maybe? That pass D finished 20th. Moreover, what's the chances that despite a large defensive overhaul, they finish 32nd in Rush D DVOA in back to back seasons. 1/32 at best?

I think that if DVOA correctly sees the Raiders as a horrible defense, it might be the crowning achievement in the projection system. Until then, I'll just have to consider the factors that DVOA seems to be oblivious to.

Oddly enough, the Raiders' offensive projection seems, if anything, a bit high.

52
by Andrew (not verified) :: Thu, 09/04/2008 - 1:20am

#9 and 11:

Except that year in and year out, most teams have about 900-1000 plays, regardless of offensive system. Since 1978 and excluding 1982 and 1987, there have been 806,922 pass attempts and rushes combined spread over 825 team seasons (978.08 per season).

The Gruden Bucs have had 5719 plays in 6 years (953.17 per season).

The Holmgren Seahawks have had 8776 in 9 years (975.11).

The Shanahan Broncos have had 13,222 in 13 years (1017.08).

The Walsh to Mariucci 49ers had 22431 in 22 years (1019.59).

The Lurie owned Eagles since 1995 have had 12,665 in 13 years (974.23).

The Packers from Holmgren's start had 15,986 in 16 years (999.13).

The Lions under Mooch and Mornhinweg 4697 in 5 years (939.40).

The Gruden/Callahan Raiders with 5952 in 6 years (992.00).

The distribution of those well known WCO teams doesn't show any sort of real bias toward a lot more plays per year than the league as a whole.

53
by Boboblaw (not verified) :: Thu, 09/04/2008 - 1:28am

I'm not sure of the exact factors that the FO system likes in PHI and TB, but I'd guess it's combo of efficiency, and returning starters/coaches, particularly on the o-line. That's what it usually is, if memory serves.

54
by Dunbar (not verified) :: Thu, 09/04/2008 - 1:50am

I can't believe someone else knows who/what Encyclopedia Brown was. No one I know read those books. I always get blank stares when I talk about them.

55
by kibbles (not verified) :: Thu, 09/04/2008 - 2:06am

Oakland projected to the worst defense in the NFL? Now *THAT* is a surprise. Wasn't too long ago that Oakland was supposed to be an unbearable offense weighing down a quietly elite defense.

56
by David C (not verified) :: Thu, 09/04/2008 - 2:09am

RE: 51

Seems like a trend to me. True, there's only five more plays per season for those eight teams, but because of the rise of the passing game and the subsequent rise in yards per play, there are fewer plays in today's seasons than there were back when running the ball was primary.

Notice that the only two teams significantly below the average are the two teams with numbers exclusively from the post-2000 era. And the next two above them are also fairly recent.

Yeah, Shanahan's Broncos are way up there, but they've always had a strong running game. Other than them, the top three teams are the older ones (49ers/Raiders/Packers).

I think the era makes more of a difference than the type of offense (because different offenses of the same era are more similar than the similar offenses of different eras). But still, the type of offense does make a difference in the number of plays. I think if you account for the changes in number of plays from year to year, and stuck exclusively with teams from after the liberalization of passing rules, you'll definitely notice a bias.

57
by tally (not verified) :: Thu, 09/04/2008 - 2:10am

53: I'm still waiting for the Hardy Boys and Nancy Drew love.

58
by Mystyc (not verified) :: Thu, 09/04/2008 - 2:33am

Just noting that Indianapolis took over hardest schedule from Pittsburgh. The AFC South has the 1st, 5th, 9th, and 11th toughest schedules. That is going to be one brutal division this year.

59
by Tom (not verified) :: Thu, 09/04/2008 - 2:34am

I would think the G-men's D ranked #24 might jump out at you too! Explanation is linked!

60
by Bobman (not verified) :: Thu, 09/04/2008 - 2:38am

56, That would be just peachy-Carolyn-Keene.

Sorry, but when I see a punchline just waiting, I can't help but be a total Franklin W Dixon.

Next up for this father of 3 boys under 8: a 2008 season preview done entirely in Seussian rhymes. My wife may kill me, but it's worth it, right? Glory lasts forever.

61
by Bobman (not verified) :: Thu, 09/04/2008 - 2:40am

57, about the same as last year, IIRC. And three of them went to the playoffs in 2007 (rightly or wrongly) with the 4th going 8-8 despite QB and WR injuries.

62
by Temperance Society Chip Bailey (not verified) :: Thu, 09/04/2008 - 2:49am

#54 I agree. It also seems at odds with the number of quality players they have on that unit. Asomugha is an elite CB and Hall is at the very least above average. Derrick Burgess is a quality pass rushing end and Kirk Morrison is an excellent MLB. Thomas Howard is a great pass coverage OLB, if still pretty weak against the run. I dunno, there just seem to be enough good components that unless the D is being run by a moron, it should be able to outrank some of the units out there (Falcons, anyone?).

63
by Kulko (not verified) :: Thu, 09/04/2008 - 3:17am

Re 45:
As for the win totals you are missing the fact that the highest and lowest win totals are random noice, so they are unpredictable.

A few of this 5-11 Teams will hit bad breaks and end up 3-13 or such, but it is completely unpredictable which is it.

For instance last year the 1-15 Dolphins had 4.2 estimated wins, and only STL was below 4.

In an extreme scenario imagine a league of 16 teams of absolutely equal strength (lets say the just bflip coins), which will play a K.o. tournament (winners always playing winners and losers playing the losers) for 4 rounds. I can guarantee you that there will beexactly 1 4-0 team and one 0-4 team, but any projection system worth its money would predict all teams at 2-2. And of course any columnist worth his money would pick a team to win and toute that they will win because of theirt coin flipping swagger.

Re MJK (21):
Its sounds a bit dubious, but there is another scenario too.
The System projected the old starter at X but he hit the wall and is only x-2. of course they have a young backup worth x-1 who wins the starting job, but all the negative variables (Continuity, experience) are there. so the projection falls of. I don't believe Aaron can tell from news reports which scenario is more to the truth, so he plays it by the book (new prohejction). He is probably slightly undervaluing them, but since he is probably undervaluing Indy too, there estimated wins should come about right.

64
by Boo Cocky (not verified) :: Thu, 09/04/2008 - 3:22am

I've been waiting eagerly for these projections since Eli's pass to Plaxico sealed the upset February 3. My question is about San Diego's defense: how much has Merriman's injury affected it, i.e. SD defensive DVOA before and after his injury? Thanks.

65
by Lou (not verified) :: Thu, 09/04/2008 - 3:48am

Congrats to Aaron and Bill. I'm so incredibly excited for football to start.

53, 46, 40

i have to admit i missed the reference, but i read the hell out of the local library's Encyclopedia Brown collection when i was a kid. No one else i know has ever even heard of them.

66
by Richard (not verified) :: Thu, 09/04/2008 - 4:28am

I'm curious why the Chargers defense is expected to be so much worse this year than last. They haven't lost anyone other than Florence.

67
by MC2 (not verified) :: Thu, 09/04/2008 - 4:34am

Interesting read, but I have one little nitpick. Why does this year not include the standard deviation for the win projection, as in previous years?

I always thought that made the projections more interesting, especially given the annoying (albeit logical) tendency for the projection to put every team somewhere between 6-10 & 10-6.

68
by ammek (not verified) :: Thu, 09/04/2008 - 4:59am

I'm still scratching my head at the Packers' projected defensive ranking. It finished 15th last season - was that just a question of third-down weakness? Is the bounce because they're returning 10 starters (even if those include Nick Collins and Atari Bigby)? Why does Dallas regress after staying injury-free but not Green Bay? Does the system understand that the d-line is one injury/felony hearing away from starting KGB on all three downs?

PS: What happens to the Browns' special teams?

69
by gilthorp (not verified) :: Thu, 09/04/2008 - 5:08am

I believe that the overall win projection for the Texans is about right, but for the wrong reasons. I will be very surprised if the offense is not significantly above average (barring critical injuries). The offensive line will probably play at a very similar level to last year, but the running back situation is less disastrous, Schaub has a year of starting experience under his belt and Andre Johnson, the one player whose loss would be really damaging, is unlikely to miss as much time as he did in 2007. This is a pretty good unit which should get slightly better.

The defense, on the other hand . . .

I'm guessing that the projection system's confidence (if it has such a rating) in the Texans defensive projection is very low. That "about average" prediction probably means "could be great, could flat out suck, we just don't know". The evidence of preseason and camp, however, suggests that if a breakout is coming for the Houston D it will be in 2009, not this year. Mario Williams aside, the defensive line has by all accounts looked awful, which means Williams can be double teamed on every single play. They can't stop the run up the middle, they can't get to the quarterback, and every healthy defensive back not named Fred Bennett sucks. Teams are going to come out with 3 wide receivers and have a field day throwing and running to whatever side Bennett isn't on (presumably the offensive left). That may change as the season progresses, if Robinson gets healthy and/or Molden develops, but right now it does not look pretty.

As for the Colts defense, I imagine it's because that unit has been all over the shop, from near-elite to God-awful, over the last few years. Since 2004, it's been above and below average in alternating seasons, despite the key personnel being largely the same. Who knows how good it will be this year?

70
by Mr Shush (not verified) :: Thu, 09/04/2008 - 5:09am

I should really pay attention to what box I'm typing in. #68 was from me.

71
by Mark Finzel (not verified) :: Thu, 09/04/2008 - 5:16am

#54 and #61 - yeah, Raiders defense won't be that bad. I don't think Morrison is excellent, actually I don't think he is as good as Howard, but the back seven should be strong against the pass. The question is up front of course, and the whole defensive concept of Oakland is based upon sand if these guys don't do well. Burgess brings heat, but who else? Teams that play the pass well and the run badly have to score a lot to hide their weakness, and Oakland is still going to struggle just a bit. A few years away there.

I admit it, I root for them, for a very long time. The old Raiders would likely have been a team that won even though their DYAR was not so hot. Even when we won our championships, there was always someone with better stats - often several teams. FO's methods frown on "big play" teams, and even I would wonder sometimes how we did it. Not being by the book was part of the fun.

72
by QB (not verified) :: Thu, 09/04/2008 - 5:44am

I find it funny how lopsided Baltimore is. Their defense is projected to be the best in the league, but their offense is projected to be so bad that it completely wipes out the positive defensive projection and then some. Luckily, they have special teams to carry them back over the line of mediocrity.

For some strange reason, I had a hankering to read some Encyclopedia Brown just last week. I haven't read them since elementary school, where we had a whole row of them in the library, but I wanted to revisit it for a day and try my hand at a mystery or two. Unfortunately, I couldn't find any stories online and I wasn't about to go out and buy an Encyclopedia Brown book for myself, so I gave up.

73
by MC2 (not verified) :: Thu, 09/04/2008 - 7:26am

If anyone is interested, I used DVOA (with a 17% adjustment for HFA) to pick every game on the schedule, and the results were:

NE: 15-1
NYJ: 9-7
BUF: 5-11
MIA: 4-12

BAL: 9-7
PIT: 8-8
CIN: 6-10
CLE: 3-13

JAC: 12-4
IND: 10-6
HOU: 8-8
TEN: 5-11

SD: 15-1
DEN: 9-7
KC: 5-11
OAK: 2-14

PHI: 15-1
DAL: 10-6
NYG: 9-7
WAS: 7-9

MIN: 13-3
GB: 12-4
CHI: 6-10
DET: 2-14

TB: 14-2
CAR: 10-6
NO: 7-9
ATL: 1-15

SEA: 12-4
ARI: 8-8
STL: 3-13
SF: 2-14

Of course, this assumes no "upsets", which makes it highly unrealistic, but I still think it's kind of interesting.

74
by This post is fungible (not verified) :: Thu, 09/04/2008 - 7:32am

Aaron: Question about the weekly picks against the spread:

Do Green have a higher success rate than Yellow rated picks? Or is there little difference between the two success rates? Can I find this out even if I pony up the loot for Premium?

75
by qed (not verified) :: Thu, 09/04/2008 - 7:39am

Is there a detailed explanation of the DVOA projection method somewhere? I'm curious how this year's draft, free agency, and injuries/retirements are taken into account.

76
by mawbrew (not verified) :: Thu, 09/04/2008 - 8:10am

I mentioned this on the AFC Over/Under Scramble last week but I think it's worth repeating...The FO preseason win projection model is not impressive. FO has been stated win projections in the last three PFPs. Obviously it's too early to tell how 2008 will turn out, but for the two years prior (2006 and 2007) their win projections were no better than an assumption of absolute parity (every team wins eight games). Over those two years the FO win projections missed by an average of 2.32 wins per team (per year). The parity assumption missed by an average of 2.38 wins per team (per year).

No offense to Aaron (I love the site and approach), but there's just no evidence (yet)that these projections are better than any random fan's guess.

77
by mawbrew (not verified) :: Thu, 09/04/2008 - 8:16am

Re: 67

I've got no idea on what the system bases it's projection on the decline in Cleveland's special teams. But, from a purely subjective POV they lost their Special Teams coach of the past few years and ended up hiring the guy that had been coaching the Raiders ST. If you check the Raiders ST DVOA for last year, you'll see this doesn't look like a 'trade up' for Cleveland.

78
by JMM (not verified) :: Thu, 09/04/2008 - 8:36am

Is there any evidence to support the use of last year's record as a factor in predicting anything about this year?

In the past I have looked at correlations between one year's record and the next and found the r2 to be less than 0.1. What do you, y'all, yinz know (about this) that I don't?

79
by Frick (not verified) :: Thu, 09/04/2008 - 8:45am

Bobman,

In regards to your question about the Colt's D ranking. Would the high schedule strength tend to move down the D ranking? They could be improved from last year, but will be facing some good offenses.

How much of NE high ranking is based on their blowing teams out last year? Did the Colts have a similar high offensive rating the year after Manning broke the TD record?

80
by shake n bake (not verified) :: Thu, 09/04/2008 - 8:57am

The hard schedule will hurt the Colts VOA, but it should get adjusted up to where it belongs with DVOA.

81
by JCRODRIGUEZ (not verified) :: Thu, 09/04/2008 - 9:17am

I refuse to accept that a team led by "Joe Flacco", a rookie QB from DELAWARE will take the division from the Steelers...lets re-run the numbers and get that story straight...

82
by glengarry (not verified) :: Thu, 09/04/2008 - 9:59am

anyone else hear the awesome Alcoa 'You Make the Call' music when raiderjoe mentioned "dvcoa" ?

83
by Duff Soviet Union (not verified) :: Thu, 09/04/2008 - 10:09am

I can't get over how much of an outlier the Falcons are. I'm pretty sure that isn't the worst DVOA you've seen, but is that the worst that has been projected?

84
by Rich Conley (not verified) :: Thu, 09/04/2008 - 10:19am

" Over those two years the FO win projections missed by an average of 2.32 wins per team (per year). The parity assumption missed by an average of 2.38 wins per team (per year).

No offense to Aaron (I love the site and approach), but there’s just no evidence (yet)that these projections are better than any random fan’s guess."

Do you have the standard deviations for those? I would think that would be more important than the averages here.

Also, the best teams don't always win. Things happen. With a 16 game season, teams are going to be as much as 3 or 4 games off their true ability.

85
by Ben Riley\\\'s Mom (not verified) :: Thu, 09/04/2008 - 10:21am

And no one is surprised that Minnesota, with Tavaris Jackson as their starting QB, is predicted to have the 7th best offense in the league?

86
by Scott P. (not verified) :: Thu, 09/04/2008 - 10:37am

Go Loins! Beat the Beras!

87
by Joseph (not verified) :: Thu, 09/04/2008 - 11:03am

This site has to have the biggest gathering of Encyclopedia Brown readers ever assembled. (I've read every one I've ever seen.) Also read lots of the Hardy Boys and Sugar Creek Gang as a kid.
Okay, Aaron, we need a separate thread for EB, as a kid detective is trying to hijack it.

88
by admin :: Thu, 09/04/2008 - 11:06am

Playoff odds are now up. Click link on my name. The playoff odds, of course, assume that the team with the higher DVOA doesn't always win the game.

I'll try to get to answering some of these questions in a mailbag at some point, but to answer one: Strangely enough, in our "test run" the Yellow picks were slightly more accurate than the Green picks. However, using the formula on the seven years of data that, well, created the formula, the Green picks are much more accurate.

89
by Karl Cuba (not verified) :: Thu, 09/04/2008 - 11:09am

OK I want to make a few points about the niners but I'm afraid of the FOMBC so I'll try to dispell it;

Dvoa Dvoa Dvoa DYAR! (while hopping on one leg, no mean feat when typing into a laptop)

That should hold it.

I suppose I should begin to prepare for the worst, as nearly every writer alive seems to be predicting the niners to only manage 5 or 6 wins this year. However, this niners side has to be one of the most upredictable teams in recent memory. I'm not certain they'll be good but I can see resons for optimism.

Firstly, the offense is unlikely to be as thouroughly inept as last year. If you watched the niners you would have seen that they would consistently end the first half with either zero or a couple of first downs. They were horrible. The coaching was abysmal, for example, what kind of moron leaves Vernon Davis in to pass block, never actually running a route for three whole games? Martz might be a lunatic evil genius from Translyvania but the guy has at least been able to move the ball wherever he's been. The niners quarterback battle/bataan death march/three stooges style farce has reduced me to a valium based view of the world but the niners' offense will almost certainly be an improvement over last year's sorry effort.

Defensively, the only starter from last year's front seven to return in the same spot is Patrick Willis (who played rather well last year). The only other starter to return is Sopoaga, who has shifted to DE where he should be a beast of a run stuffer. I can't think of any other situation where a side has so comprehensively rebuilt their defense without anyone seeming to notice. Justin Smith's arrival and the return of Manny Lawson from injury mean that the niners will actually have two outside linebackers for their 3-4 this year. They demonstrated last year that the 3-4 isn't much of a scheme without those guys, a pass rush of three fat linemen and a fat Tully Banta-Cain just doesn't get to the quarterback enough. Even a half decent pass rush should help a secondary that was often asked to hold coverage for an insurmountable length of time. And the front seven is now huge, the second heaviest in the league I think, which should help the run defense.

The 49ers have decent special teams but basically this season will depend on Martz. If he's a genius then the niners could suprise a lot of people. This season will be a pretty good measure of the effect of a very good offensive coordinator as opposed to a dunce who's well out of his depth.

90
by jonnyblazin (not verified) :: Thu, 09/04/2008 - 11:31am

"I refuse to accept that a team led by “Joe Flacco”, a rookie QB from DELAWARE will take the division from the Steelers…lets re-run the numbers and get that story straight…"

You're right, thats almost as absurd as assuming a rookie QB from Miami of Ohio could lead a team to a 15-1 record.

91
by dryheat (not verified) :: Thu, 09/04/2008 - 11:37am

Is anybody going to pick up MDS's injury column? I do have the right author, correct?

92
by shake n bake (not verified) :: Thu, 09/04/2008 - 12:00pm

The Colts D's low projection isn't from 3rd down performance or trending downward at the end of last year. They were great on 1st down last year, good on 2nd and mediocre on 3rd down, so if anything they should improve based on that. There's nearly no difference between their ranking in total vs weighted DVOA on defense.

I can't figure it out.

93
by Whatever0 (not verified) :: Thu, 09/04/2008 - 12:07pm

81: Yes, there is absolutely zero precedence for the Baltimore Ravens winning games with fantastic defense, great special teams, and atrocious quarterback play. None whatsoever.

94
by MCS (not verified) :: Thu, 09/04/2008 - 12:56pm

#73. Thanks for the effort. I thought the adjustment was 17% for offense and 17% for defense.

95
by George (not verified) :: Thu, 09/04/2008 - 1:17pm

I look at your work every year and see that you are no closer now than you were when you first started. The Eagles are the #2 team in the NFL. The 49ers #31. You have the Seahawks way ahead of the Cowboys, who were the most potent team in the NFC, though the Seahawks were the beneficiaries of the worst division in pro-football.

You need to find appropriate limits in your data in order to construct a reasonable model. Because right now, it's pretty clear that you're NOT applying appropriate filters and your results are, at best, only marginal and, in my eyes, pretty much Garbage-in, Garbage-out.

96
by Jimmy (not verified) :: Thu, 09/04/2008 - 1:28pm

I note from the playoff odds page that you seem to be saying that there is zero chance (or at least no significant chance) of New England having the first pick in the draft next year. Simply to clarify, is this because there is no chance they could be that bad? Or is it because they are unlikely to still have a draft pick due to forfeiture?

97
by starzero (not verified) :: Thu, 09/04/2008 - 1:59pm

don't want to start a fire, but what does the n.e. creampuff schedule the past few years say about that team's overall ability? are any players possibly overrated because of this? if so, what's that say about other teams with harder schedules?

98
by weinsteinium (not verified) :: Thu, 09/04/2008 - 2:16pm

I agree with Raiderjoe here, you project the raiders to have the worst defense in the NFL?!?!?!? I decry your statistical engine as a Raider Hater.

99
by Keith Cockrell (not verified) :: Thu, 09/04/2008 - 2:31pm

The strength of schedule is nonsense. Philly plays the same teams Dallas plays with two exceptions and then plays Dallas (instead of themselves) which won more than Philly last year. They might be one or two places harder or easier than Dallas, but they are NOT 31.

100
by mawbrew (not verified) :: Thu, 09/04/2008 - 2:33pm

Re: 84

I don't see why s.d. would be more appropriate here. To simplify, if we are predicting the win total for just two teams and are off by 0 for team A and 6 for team B, why is that inferior to a predition that is off by 3 games for both teams? I jut don't see why one would be inherently better than the other.

That said, the standard deviation (assuming I've calculated it correctly)for the FO 'prediction delta' is slightly lower than that of the parity (every team 8-8) model. But the difference isn't great (2.85 for the FO vs. 3.05 for parity), about 7%. Again, I don't think this is particularly impressive even if you assume the s.d, is a better indicator.

"Also, the best teams don’t always win. Things happen."

Of course, things happen. That's what makes it difficult to predict. That's also pretty much my point. It's so difficult to predict, that the best model developed by a top notch expert in the field is essentialy no better than random chance.

101
by Tom (not verified) :: Thu, 09/04/2008 - 2:40pm

If the west coast offense is so DVOA friendly, then how come the Saints are projected so low?

102
by RMoses (not verified) :: Thu, 09/04/2008 - 2:57pm

Let the haters hate... DVOA "still" loves Philadelphia!!!

103
by josh o. (not verified) :: Thu, 09/04/2008 - 3:54pm

@98 - your comment is nonsense. Philly and Dallas play a lot of the same teams, but it is significant where they play them (home/road) and the teams that are NOT the same are not on the same level (as far as DVOA is concerned).

Philly plays Seattle on the road (ranked #7 total DVOA) while Dallas plays Green Bay on the road (#3); AND Philly plays Arizona (#18), Cleveland (#24) and Atlanta (#32) at home while Dallas is visited by Seattle (#7), Tampa Bay (#8), and Cinci (#20).

Those differences are pretty significant. If you assume that the team ranked higher in DVOA wins every time (which, of course, it doesn't) than Dallas would win three of four given Philly's schedule and only one of four given their own.

104
by Wilford Wiggins (not verified) :: Thu, 09/04/2008 - 4:18pm

@40,44,...,87: Wikipedia Brown is where it's at these days (see link)

105
by Waverly (not verified) :: Thu, 09/04/2008 - 4:23pm

I think a team's strength of schedule calculation should never include the opponents' games against the team itself.

106
by TomC (not verified) :: Thu, 09/04/2008 - 4:40pm

#94 -

Here, let me help you with that:

Seattle is clearly ranked too high because Dallas has bigger potency. Appropriate filters and data limits is way better than this. I like pro-football!.

107
by zlionsfan (not verified) :: Thu, 09/04/2008 - 4:59pm

Re 36: That shouldn't be an issue. Nothing FO prints can change January 5-11, 1991.

raiderjoe, you're right, the Lions really could start 4-0. Of course, after last season, I understand the perils of a back-loaded schedule a bit better ... 5-11 sucks no matter which 5 games you win.

108
by Pat (not verified) :: Thu, 09/04/2008 - 5:09pm

Over those two years the FO win projections missed by an average of 2.32 wins per team (per year). The parity assumption missed by an average of 2.38 wins per team (per year).

That's a terrible comparison. A parity projection (everyone at 8-8) will always have the lowest error of any brain-dead projection, but that's simply because it's making a projection which has the least "room for large error."

So you can't really compare the two predictions. The 8-8 prediction is "safe," so it has less chance of being violently wrong, but it also has less team-by-team information (as in, none), which means it has no value.

Imagine a weather forecast - one channel comes out and says there's a 10%, 20%, 30%.. etc. chance of rain. And then another channel says "screw it, there's a 50% chance of rain each day." Channel B might actually have a relatively close error to Channel A! But Channel A's prediction could still be better. How? Well, Channel B's prediction contains no day-to-day information. None. But Channel A's does. So you can say "when Channel A says there's a 75% chance of rain, what's the likelihood there's rain?" If it's greater than 50%, then Channel A is providing information, whereas Channel B is not.

Try this:
1) Start off with everyone at 8-8. Compare the average error.
2) Now, draw from a Gaussian distribution around 8-8, with a spread of 0.5 wins. Compare the error now. Repeat, say, about 1000 times. Average.
3) Repeat #2, increasing the Gaussian spread to 1.0 wins.
4) Repeat #2, with 1.5 wins.
etc., etc.

Plot the error versus the Gaussian width, and you'll find that the average error increases as the width increases (it's just the convolution of the pick spread and the natural NFL spread).

Now, calculate the standard deviation of the FO predictions, and compare them to the "brain dead" pick with the same standard deviation, and you'll find that the FO picks do much better (the standard deviation for the FO picks is, what, ~1.5? So the braindead parity assumption should have a spread of ~2.8 or so, and the FO predictions are significantly lower than that.

(From a betting perspective, if you care, it's also easy to prove that you can bet more accurately using the FO picks than a parity assumption.)

109
by Pat (not verified) :: Thu, 09/04/2008 - 5:24pm

What really, really amazes me is all the people saying "how can team X's strength of schedule be so much different than team Y?!"

All the schedule ranking is is just the average DVOA of all of a team's opponents. Want to know why Dallas has so much higher schedule rank than Philly? Because Philly is projected to be much better than Dallas (~20% DVOA difference), Seattle is projected to be much better than St. Louis (~50% DVOA difference), and the Packers are projected to be much better than the Bears (~30% DVOA difference).

That's 1/4 of the Eagles schedule right there, with an average DVOA 30% lower than the Cowboys.

110
by Insancipitory (not verified) :: Thu, 09/04/2008 - 6:35pm

Pat, or whomever, does FO's strength of scheduel factor in where the game is going to be played? Home vs Away are pretty significantly different circumstances to be facing the Seahawks.

In my eyes, the best part about the Seahawks scheduel this year is that they get, aside from Dallas, their strongest opponants in Qwest.

111
by mawbrew (not verified) :: Thu, 09/04/2008 - 7:58pm

Re: 107 - "The 8-8 prediction is “safe,” so it has less chance of being violently wrong, but it also has less team-by-team information (as in, none), which means it has no value."

Again, that's pretty much the point. If the FO model can't beat a model that has no value, what does that say about the FO model? Providing information that is no better than a brain dead model isn't really helping me.

The more elaborate analysis is beyond my capability these days (not sure my stats skills were ever that advanced). Hopefully yoou or someone else can accept that challenge and provide it as well as a explanation about why it's superior to the simple comparison I did. By the way, the s.d. for the two is listed in post 99 and is very close, but I'm not sure that's what you were looking for.

I'd much prefer that the FO model be vindicated, but I just don't see it (yet).

112
by Scott de B. (not verified) :: Thu, 09/04/2008 - 8:46pm

Again, that’s pretty much the point. If the FO model can’t beat a model that has no value, what does that say about the FO model?

The point is, you are using the wrong metric.

113
by johnnyxel (not verified) :: Fri, 09/05/2008 - 4:21am

@mawbrew:
I'll see if I can take up the challenge of describing the difference here.
Let's assume that both the FO win projection and a parity assumption of 8 wins for everyone have a standard deviation of exactly 2.5 wins.
In general, there is a 95% probability that a result will fall within two standard deviations, and roughly 2/3 probability that a result will fall within a single standard deviation.
That means, based on a parity assumption, one would expect 95% of NFL teams to have a record between 3-13 and 13-3.
What the FO projections would do, assuming a 2.5 standard deviation, is predict that a particular team will have a specific number of wins, +/- 5 with 95% confidence, so instead of the Patriots winning between 3 and 13 games, FO predicts with 95% confidence a season with between 7 and 16 wins, centered on 12.
They are equally accurate, but the FO prediction is more informative, because it helps figure out which teams will be above or below the 8-win mark, which ten or eleven teams will probably win less than 6 games, and which will probably win more than 10, and just how likely it could be that the Patriots go undefeated (again).

114
by Rod Pullmann (not verified) :: Fri, 09/05/2008 - 11:43am

The whole notion of predicting outcomes for season X based upon analysis of play-by-play performance from seasons [X-n,X-1] is a bit flawed in the first place. Given the high degree of player and coaching turnover from year to year in this era of free agency and free-swinging coach scapegoat hatchets, it's a bit of an apples-to-lawnmowers comparison.

Teams change schemes year in and year out, and there's no accounting for the appropriateness of players for various positions within those schemes: plugging a Vince Young or Michael Vick type into an Andy Reid pass-wacky offense, for example. That type of personnel mismatch happens more often than it should.

Perhaps DVOA represents a set of objective metrics, which seems "better" than a bunch of subjective journalistic and fan-spewed swags, but it seems like those metrics are construed in a number of digits beyond the decimal point far beyond their significance. It should mostly be regarded as simply another tool, albeit a thorough one, to be used in the black art of NFL prognostication.

115
by Dr. Slashdot Fan (not verified) :: Fri, 09/05/2008 - 11:50am

Improving discussion boards with a “slashdot”-style model that will still allow for commenting without registration

Thank you, thank you, thank you. That's the best news I've heard in a while.

116
by Possuum (not verified) :: Fri, 09/05/2008 - 2:11pm

Why is DVOA so down on da Bills defense? I can understand the offense.

117
by Pat (not verified) :: Fri, 09/05/2008 - 3:14pm

Again, that’s pretty much the point. If the FO model can’t beat a model that has no value, what does that say about the FO model?

Your problem is that you're assuming that average error is the "end all, be all," and if prediction A has a lower average error than prediction B, A "beats" B.

This is wrong. Completely. Especially in this case, where the "8-8 prediction" is a completely different kind of prediction than the FO predictions. You can't compare predictions that naively.

If you read the previous post I had more carefully, specifically the weather forecaster example, it might help. But just as a simple bit of help, think of it this way: the "8-8 prediction for everyone" is not predicting an actual NFL season, because an actual NFL season has far more variation than that on average (the standard deviation of each team's wins is more than zero). So the predictions are meaningless. The FO predictions, however, are predicting something "closer" to an actual NFL season.

When you increase the standard deviation of the predicted wins, you increase the average error. And therefore, if you want to compare the FO predictions to a "brain dead, 8-8" prediction, you have to spread the 8-8 samples by a Gaussian equivalent to the spread in the FO prediction.

Another "simple" explanation is this: I could likewise predict "every team is going to have between 0 and 16 wins," and I'll have less error than any prediction! It's also equally worthless.

118
by turbohappy (not verified) :: Fri, 09/05/2008 - 3:51pm

I think the Colts will end up doing better on D as well, although I think they are likely to end up a few places lower on offense. However, I think the biggest difference from the projection will be on special teams. I think this will finally be the year they don't totally suck on returns and coverage.

119
by Chris (not verified) :: Fri, 09/05/2008 - 5:30pm

81, 90, 93...Let's put it this way - the Ravens defense will have to be historically good, given how much of the game they will be on the field cleaning up after that cover-your-eyes-awful offense.

Man, I just love this part of the year.

120
by Duff Soviet Union (not verified) :: Fri, 09/05/2008 - 11:20pm

Chris, Baltimore's defense is projected here to be historically excellent. There's a fair gap betweeen them and #2 Green Bay. Their offense is projected to stink on ice (#31, even that might be too high), but they should be average given great defense and special teams. You know, like pretty much every other Ravens team this decade. The questions are whether their defense will be THAT good and whether the rest of the division will be that bad. I say yes to the first. They had terrible, historically bad fumble recovery luck last year and have some great young players. I also say yes to the second question. Pittsburgh has a brutal schedule (San Diego and New England instead of Oakland and Miami), Cinci look like a train wreck and Derek Anderson has some serious regression coming his way. So, I think they'll win by default pretty much.

121
by mawbrew (not verified) :: Sat, 09/06/2008 - 5:19pm

Re: 113 & 117

I appreciate your taking the time to attempt to educate me on this. Really. I still don't quite get it, but when enough people tell me I'm wrong, I know it's time to shut up.

I'll provide one last piece of info for this discussion. Pat's comment that it would be easy to demonstrate that the FO predictions would be superior for bettting purposes got me wondering how these two approaches would compare on the over/under win total bets that initially sparked my interest in this. For 2006 and 2007, the FO predictions were winners 60% of the time versus the vegas number. Pretty good. The weird thing is that the parity model was 65% over this same period vs the vegas number. Not a large sample size obviously, but interesting anyway. May indicate that the past two years were just especially good for the parity model.

122
by GoodKingJohn (not verified) :: Sat, 09/06/2008 - 9:08pm

forget who made it first, but the point about WEst Coast offenses being overrated made sense. I am not sure of the exact formula for DVOA, but perhaps all of those little successes add up to more than they should. perhaps also, looking at number of plays over the year is a little misleading.

for example, a WCO offense team has 1000 plays and 800 are successful.

however, since they NEED a high rate of success (for dink passes to worth their way down the field), they dont accomplish as much as it appears. a non WCO offense has 750 successful plays, but perhaps was more successful since they did not NEED to be as successful.

also, i do not understand the difference mentioned regarding simulations.. can somebody explain it?

"The projected wins are going to be a little different from those you found in PFP 08. Rather than being based on Ben Alamar’s simulations, these are based on the simulations we run to figure the DVOA Playoff Odds each week. "

thanks

123
by Peter (not verified) :: Sun, 09/07/2008 - 9:59am

122: You're just talking about success rate. DVOA still takes into account how successful a successful play is... a 20-yard-pass on 3rd and 10 is worth more than a 10-yard-pass on 3rd and 10.

124
by GoodKingJohn (not verified) :: Sun, 09/07/2008 - 10:45am

yes, I understand that, but what I am saying is that perhaps that high success rate (even with a lower value per success) is distorting the value.
for instance, perhaps in reality (and nobody knows) a success unit for gaining 4 yards on third and three should be x, but it is given more than x)

125
by t.d. (not verified) :: Sun, 09/07/2008 - 11:30am

I don't see how you could predict a historically great defense with the personnel in place in Baltimore. Their back seven has age in a few places, they have some good young pieces, but noone who is a lock to make 'the leap'. Also, Philly and Tampa are always overrated. I guess we'll see

126
by davidCO (not verified) :: Mon, 09/08/2008 - 12:41am

Looks like things already need to be recalculated now that Tom Brady is out for the year.

127
by t.d. (not verified) :: Mon, 09/08/2008 - 12:48am

If Brady is out for the year, the AFC is going to be totally wild. It's just one game, but Pittsburgh looks like the class of the conference.

128
by NE fan (not verified) :: Thu, 09/11/2008 - 4:52pm

#96,

chuckle, chuckle:)

A long time and loyal Pats fan.

Go, er, Matt?