Week 1 DVOA Ratings

Week 1 DVOA Ratings
Week 1 DVOA Ratings
Photo: USA Today Sports Images

by Aaron Schatz

You love them when your team is high! You hate them when your team is low! Once again, the famous Football Outsiders DVOA and DAVE ratings return for 2010.

Now, some of you may be familiar with DVOA, but you have never met our good friend DAVE. Well, DAVE is our rating that combines the preseason projection with the results of early games to give us a better prediction of how each team will rank at the end of the year. For those who don't know the story, this metric is called DAVE as a reaction to criticism that our stats are too much alphabet soup. I mean, who can argue with a guy named Dave? (Technically, it stands for DVOA Adjusted for Variation Early.) In this week's DAVE ratings, the preseason projection counts for 90 percent, and the current VOA counts for 10 percent.

There's never much to say after just one week of games and this year is no exception. The exception to this lack of exception is the Dallas-Washington game, where the Cowboys completely outplayed the Redskins according to our ratings and lost anyway. It's also interesting to note how much just one win or loss can change the FO playoff odds. Kansas City is now listed with the third-highest chance to make the postseason out of any team in the league, even though their DAVE rating isn't that much higher than San Diego's. The one team that didn't really get a huge boost is Houston, because our poor preseason projection keeps their DAVE rating low. Of all the teams where FO differed from conventional wisdom this season, that's the one where we really looked wrong on opening weekend. If the Texans truly are as good as they looked against the Colts, they'll quickly climb up the DAVE ratings -- and in the playoff odds report -- as the preseason projection gradually becomes a less important part of the formula.

All stats pages should now be updated with 2010 data except for OFFENSIVE LINE and DEFENSIVE LINE, which will be updated after Week 2. The FO Premium splits database will also be updated for 2010 after Week 2, next Tuesday.

* * * * *

These are the Football Outsiders team efficiency ratings through one week of 2010, 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 VOA 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.

There are no opponent adjustments in VOA until the fourth week of the season, which is why it is VOA right now rather than DVOA. 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 VOA to get a more accurate forecast of how a team will play the rest of the season. Right now, the preseason projection makes up 90 percent of DAVE.

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>

1 SEA 63.7% 0.3% 17 1-0 2.6% 13 -52.0% 2 9.1% 7
2 TEN 59.1% -0.2% 18 1-0 21.5% 7 -28.4% 6 9.3% 6
3 NE 57.2% 25.0% 3 1-0 56.8% 1 16.0% 25 16.4% 2
4 NYG 56.6% 13.6% 9 1-0 0.9% 14 -74.9% 1 -19.3% 31
5 JAC 40.3% -6.2% 20 1-0 40.2% 3 14.8% 24 14.9% 4
6 GB 38.2% 17.7% 4 1-0 26.2% 6 3.3% 21 15.3% 3
7 DAL 36.6% 6.1% 15 0-1 31.3% 4 -15.5% 11 -10.2% 27
8 CHI 29.6% 15.7% 6 1-0 -13.5% 24 -39.9% 4 3.3% 15
9 HOU 25.6% -14.1% 27 1-0 46.8% 2 34.7% 30 13.5% 5
10 KC 25.2% 1.9% 16 1-0 0.8% 15 0.3% 19 24.6% 1
11 MIA 23.4% 14.0% 8 1-0 9.1% 12 -25.2% 8 -10.8% 28
12 NO 23.4% 11.0% 11 1-0 17.9% 10 -9.2% 14 -3.7% 21
13 PIT 13.5% 25.5% 2 1-0 0.0% 16 -12.4% 13 1.0% 16
14 ARI 10.2% -3.7% 19 1-0 -8.5% 20 -25.2% 7 -6.5% 25
15 BAL 8.8% 29.2% 1 1-0 -32.1% 27 -45.5% 3 -4.5% 22
16 TB 8.4% -9.5% 24 1-0 -11.0% 23 -15.5% 12 3.8% 13
17 ATL 3.7% 17.6% 5 0-1 -3.6% 17 -3.6% 18 3.8% 14
18 CLE -1.4% -18.6% 28 0-1 -9.6% 21 -17.0% 10 -8.8% 26
19 PHI -4.1% 12.5% 10 0-1 12.1% 11 13.9% 23 -2.2% 19
20 NYJ -7.4% 10.4% 12 0-1 -49.4% 29 -34.7% 5 7.2% 8
21 DEN -16.4% -7.6% 22 0-1 18.9% 9 31.8% 29 -3.6% 20
22 DET -17.6% -31.7% 31 0-1 -39.6% 28 -18.0% 9 4.0% 12
23 BUF -19.6% -12.1% 25 0-1 -24.5% 25 1.8% 20 6.7% 9
24 STL -19.9% -21.8% 29 0-1 -30.4% 26 -5.2% 16 5.3% 11
25 SD -29.0% -7.8% 23 0-1 -6.7% 18 -8.9% 15 -31.2% 32
26 MIN -30.5% 9.1% 13 0-1 -8.1% 19 23.4% 27 0.9% 17
27 IND -34.3% 14.9% 7 0-1 28.6% 5 49.9% 31 -13.0% 29
28 WAS -38.7% 6.9% 14 1-0 -10.9% 22 26.0% 28 -1.9% 18
29 CIN -50.4% -7.4% 21 0-1 20.9% 8 52.5% 32 -18.8% 30
30 CAR -60.2% -13.0% 26 0-1 -71.6% 32 -5.1% 17 6.3% 10
31 SF -76.1% -24.9% 30 0-1 -63.6% 31 6.2% 22 -6.3% 24
32 OAK -84.9% -36.0% 32 0-1 -56.3% 30 23.1% 26 -5.4% 23


190 comments, Last at 17 Sep 2010, 2:20pm

#1 by Lyford (not verified) // Sep 14, 2010 - 4:06pm

is clearly ranked because . is way better than this.

Points: 0

#2 by justme_cd // Sep 14, 2010 - 4:09pm

Interesting. I'd been feeling like the only Cowboys fan who doesn't think our offense was completely doomed based on preseason and one game. I thought they drove well all game and the low scoring games increase the importance of plays similar to the end of the half relative to that same play in a high scoring affair.

Still, 4th in offensive VOA? Can someone explain what the computer sees to spark some hope in Dallas?

Points: 0

#5 by TA (not verified) // Sep 14, 2010 - 4:13pm

Dallas moved the ball like it was going out of style. Pretty easy to see that as an upside, but the redzone offense and penalties certainly taper enthusiasm.

Points: 0

#10 by Rusty G. (not verified) // Sep 14, 2010 - 4:18pm

I actually didn't come out of that game all that concerned either.

Miles was uncoverable, Dez seemed to be an improvement at WR2 when Tony wasn't forcing him smoke screens against press coverage and the backs did their thing. Pass protection was crap, but what else is new?

Fluky outcome, but that's the NFL. Or maybe this team is just predisposed to commit godawful penalties and make singularly horrible decisions from time to time that create a 15-20 point drag.

Under Wade, all such things are possible!

Points: 0

#101 by BigCheese // Sep 15, 2010 - 12:47am

"Or maybe this team is just predisposed to commit godawful penalties and make singularly horrible decisions from time to time that create a 15-20 point drag."

As long as Barron's starting I'd say that's a pretty safe bet.

- Alvaro

Points: 0

#74 by AlanSP // Sep 14, 2010 - 8:50pm

If I remember correctly, VOA doesn't include penalties, so Alex Barron helped the VOA cause a bit by blatantly holding whenever he was beaten rather than giving up negative plays. Doug Farrar pretty much nailed it in last week's article on scouting when he wrote, "Some players don't fit their schemes and are unfairly criticized for it. Then, you have your Alex Barron and Levi Brown-types, guys whose sub-par play is scheme-transcendent."

Points: 0

#3 by Key19 // Sep 14, 2010 - 4:12pm

Well, at least I don't have to complain about Dallas being ranked too low this week.

Looks like we're going to be witnessing another defensive-minded game against Chicago this week. Holy cow, what an offensive/defensive split for those guys.

Points: 0

#11 by Thomas_beardown // Sep 14, 2010 - 4:18pm

Our defense is not likely to look quite that good again. Unless Shaun Hill is also QBing the Cowboys.

However, I'm thinking the offense might get stronger as the year goes on. If the key members can stay healthy.

Points: 0

#12 by Thomas_beardown // Sep 14, 2010 - 4:18pm

Our defense is not likely to look quite that good again. Unless Shaun Hill is also QBing the Cowboys.

However, I'm thinking the offense might get stronger as the year goes on. If the key members can stay healthy.

Points: 0

#52 by TomC // Sep 14, 2010 - 6:56pm

Interesting contrast between naive application of DAVE and not-naive opinion of Vegas for that game. DAVE plus Dallas home field of +18% (or whatever the correct figure is) implies narrow Dallas win or reasonable probability that Bears can upset. Vegas says Dallas by 9, i.e. Bears have no shot. Despite my delicious glass of blue and orange kool-aid, I'm inclined to believe Vegas.

Points: 0

#56 by RichC (not verified) // Sep 14, 2010 - 7:18pm

Vegas doesn't say anything.

Vegas says the public doesn't think the Bears have a shot.

Points: 0

#71 by Jeff Fogle // Sep 14, 2010 - 8:34pm

Vegas/offshore is focused on professional wagerers and the public both. Cowboys are -400 on the moneyline, Bears +320...so not quite no chance.

Points: 0

#99 by Eddo // Sep 15, 2010 - 12:35am

A very common misconception is that Vegas lines reflect the betting public. In actuality, they reflect how oddsmakers (both employed by casinos and sports books and those who do the big-money betting) view the strength of the teams in question.

Sure, the general public might see a line of Dallas -9 and think, "The Cowboys are more than nine points better than the Bears", and bet the line up. However, the smart money coming in - the big-money bets - comes from, for lack of a better term, "professional" gamblers, who generally have their own systems for rating teams and setting lines. If the line gets bet by the public up to say, Cowboys -11, but the consensus opinion of the smart money, which is similar to Vegas, thinks the line should be -9, then they'll start placing larger bets on the Bears +11, so the line will come back to an equilibrium point.

Does this make sense?

Points: 0

#103 by BigCheese // Sep 15, 2010 - 12:56am

Actually I'm pretty sure the initial lines are set with at least an equal wieght given to public perception than to team strength. Vegas isn't interested inpredicting the outcome of the game. They are interesting in setting a line that will get about equal action on both sides of it, which is how they make their money. THAT is why a line moves. If one side is seeing too much action, the line moves so that the other side will see more action and things will balance out.

This is why lines on teams like the Cowboys, the Steelers or most New York teams usually are a little higher than they would be if the team played anywhere else, because the huge fan-base is going to slant the betting towards those teams. Or doesn anyone really beleive either Vegas, the sharks or the general public saw no bigger miss-match in week one than Giants/Panthers?

- Alvaro

Points: 0

#107 by jbrown (not verified) // Sep 15, 2010 - 3:03am

Yeah, and having spent quite a bit of time behind the scenes in Vegas, more of this is done by computer than some expert despite what people think. The goal is to divide the bets in half and limit risk, not predict the outcome of the game.

Points: 0

#127 by Eddo // Sep 15, 2010 - 12:24pm

Dividing the bets is the goal of the casinos, yes, but what I was trying to say is that the high-volume betting of the professional gamblers, when they see a line tilted too far in one direction, brings the lines back down to what is a pretty good estimate of how "Vegas" views the true strengths of the teams involved.

Points: 0

#155 by BJR // Sep 15, 2010 - 4:29pm

That isn't true. Or at least it shouldn't be. The aim of the bookmaker is to set the line correctly so that if the game was played a million times, the line would be the median outcome. Of course that is difficult. However, it's a bad bookmaker that moves his line simply because he has a position one way, to try and neutralise that liability. Maybe if the liability is too large, but in that he case he should not have risked that liability in the first place. What the bookmaker should do is assess WHO is placing the bets, and whether they think those persons will win money betting in the long term. More often than not, the answer is no, so don't move the line. When the answer is yes, that is when the line should move.

Points: 0

#157 by Jeff Fogle // Sep 15, 2010 - 4:45pm

Agree with the "who" element here BJR. Only thing I'd add though is that once oddsmakers have a read on the market...they'll take gladly take positions against the "who" rather than try to post a median outcome. Let's say you're an oddsmaker who thinks the median outcome is a six...but you KNOW the WHO is going to bet the favorite (because the public loves betting favorites of less than a TD). You post a 7.5 or an 8 so that the median, everything below it, and a few of the results above it cash your side of the ticket.

Oddsmakers put the 11/10 and this knowledge of the public to work in their favor. Professional wagerers try to find spots where they can squeeze out some of those public losses for themselves. The settled widely available lines reflect the elements of that dance. I think your last few sentences really explain real world decisions of oddsmakers very well. I've ghostwritten for a few over the years. That's how they describe it.

Points: 0

#186 by otros // Sep 17, 2010 - 12:14am

Getting the expected value of the game so they "get it right if the game is played a million times" doesn't make sense, because it isn't difficult, it's downright impossible. The good oddsmaker, not being a psychic, has to play to reduce his risk, not betting himself that he's the best liner maker in the sense that you describe, but the one that reaps benefits no matter what the final outcome of the game is.

Points: 0

#190 by Jeff Fogle // Sep 17, 2010 - 2:20pm

In the real world though, the nature of gamblers makes it difficult if not downright impossible to split the action ON EVERY GAME so that one reaps the benefits no matter the result. Oddsmakers must find a balance between reducing risk and maximizing return. Taking a position at a good number (from their point of view) along with the 11/10 advantage is how they've done this. That's a potent combination over a high volume of games. The public is laying -8 when it should be -7, AND ALSO paying 11 on their losses and getting only 10 on their wins. The oddsmaker has +8 and 10/11.

The combination of deep pockets, and plenty of time for the math to work itself out in the big picture, allows sportsbooks to take positions against the public. That's not a theory. That's how recent and current oddsmakers describe what they do in the modern market.

Points: 0

#114 by Rusty G. (not verified) // Sep 15, 2010 - 8:25am

Not to get all "Wisdom of Crowds" on everyone here, but it's not apparent to me why a market-negotiated view of the most likely outcome should somehow be worse than a randomly selected expert's view.

And in this particular case, we're not necessarily talking about whether the spread is right, but whether the ridiculousness of the spread is at least informative. I think it is.

I also think 9 points is preposterous.

Points: 0

#124 by Jetspete // Sep 15, 2010 - 11:52am

You also need to remember that Dallas is a heavily bet team, thus 9 might be purposefully inflated by a 1/2 to a point. Also i have to believe vegas is putting safeguards in the lines to try and minimize the damage the public did to the books the first few weeks of last year, when seemingly all big spreads for heavily backed teams hit, meaning parlays hurt the books. the line is almost vegas' way of begging the general public not to put dallas in their parlays.

Points: 0

#130 by BigCheese // Sep 15, 2010 - 12:42pm

What about a high point spread makes it less succesptible to parlays than any other bet?

I think it's more about preventing teasers (which are the worst bet you can make anyway, so they shouldn't be too worried anyway) and, more importantle, getting killed on the moneyline if the underdog wins.

- Alvaro

Points: 0

#149 by jebmak // Sep 15, 2010 - 4:06pm

The books want as many teasers and parlays as they can get.

Points: 0

#156 by Jeff Fogle // Sep 15, 2010 - 4:34pm

Favorites sticking at -9 has become more common since professional wagerers have really started attacking two-team teaser nominees that cross both the 3 and the 7 when you move the line 6 points.

These would be moving favorites in the -7.5 to -8.5 range down to -1.5 to -2.5 in two-team teasers...and moving underdogs in the +1.5 to +2.5 range up to +7.5 to +8.5.

This specific teaser strategy has been called "basic strategy" or "Wong teasers" because author Stanford Wong discussed them in his book "Sharp Sports Betting" that came a few years ago. I know guys who were doing this back in the 1980's though...so it's not like it's a new thing, and it makes me nutso when people call them "Wong teasers" (lol).

Teasers in general are a bad bet, particularly in college football. They're actually so good in this line range though in the NFL that sportsbooks have had to start playing defense against them. Many Vegas books now charge higher vigorish on them or don't even offer them. Offshore places (which typically drive the line these days while Vegas stores copy the "settled" numbers after the smart money has hit offshore), would often rather avoid the range and just take a position on the game. You assume the store is going to get one sided in action because games don't often truly split out evenly. They can live with rooting for either team at the nine and accepting the result. They really don't want to get flooded with teaser bets that have them rooting for Chicago +2.5 to spike all the teasers.

Note that all five nominees that hit the range at some point late last week covered the teaser number (all underdogs, Pittsburgh, Cleveland, Denver, Houston, and Baltimore)...meaning all possible combinations hit, and sportsbooks had to pay out on any basic strategy teaser that was bet.

So, part of this line of Dallas -9 may be "we know it should be lower, but we're not going there because of teasers" line of thinking. If they moved it all the way down to -7 to miss the window, then there's a chance for very high exposure because the public would surely hit Dallas -7. As others have pointed out, Dallas does get support from the public as a home favorite generally.

Should be fun seeing if and how this line moves. Stores that don't get much teaser action because they don't offer them or make them unattractive to bet with higher vigorish may sit at 8 or 8.5, while those who take teasers stay at 9. This area around the 9 has been causing headaches for sportsbooks since the sharps started getting really aggressive with the basic strategy teasers in recent seasons. Newly arriving quants in particular have an "oh my god" type moment once they've crunched the numbers. It's not like every week is a board sweep like Week One. But, over time, it's an edge for the sharps that sportsbooks are having to deal with.

Points: 0

#4 by jmaron // Sep 14, 2010 - 4:12pm

Vikings offence was actually better than their defence according to DVOA. That seems counter to my sense of the game. I thought the offence was horrendous whereas the defence held NO to 14 points, albeit there were two missed FG's but one was after a 3 and out stuff after a Favre interception.

In my mind holding NO - in NO to 14 is damn fine D. Scoring 9 points...horrendous.

Points: 0

#13 by drobviousso // Sep 14, 2010 - 4:19pm

DVOA loves long, sustained drives of many plays for few yards with many first downs. The one drive that took approximatively 1,324 minutes and was made up of runs of exactly 3.33334 yards (or whatever) was probably worth a lot.

I think its based on the fact that the system is a per play, instead of per drive, metric is the underlying reason. Also might be why it likes small ball O like Philly and not long ball O like Dallas, but that's just a personal pet theory. No one really knows why that is.

Points: 0

#43 by RichC (not verified) // Sep 14, 2010 - 6:05pm

It would be awesome if we didn't have to hear this oft repeated, and often proven wrong statement every week.

DVOA loves drives of long plays a lot more than it loves long drives.

Points: 0

#88 by RickD // Sep 14, 2010 - 11:44pm

We really have no idea what DVOA loves or hates since Aaron keeps the formulae secret.

Points: 0

#50 by tally // Sep 14, 2010 - 6:45pm

It does love sustained drives, but not for the reason you cite.

It's simply sample size. Multiple first downs on a drive allow you to draw a more reliable conclusion about the true offensive ability than a boom or bust plays.

Points: 0

#82 by MJK // Sep 14, 2010 - 10:11pm

This is a good point that a lot of people overlook...I just said something similar further up in the column talking about defense.

If Team A has three offensive possessions, and on each one go no gain, no gain, break a long TD run, then the system will think that 1/3 of the time A has a chance of doing awesome, and 2/3 of the time S will do nothing (greatly oversimplified, because the system adjusts for context and scenario, but bear with me).

On the other hand, if Team B has three possessions where B gains 5 yards on every play, and scores a TD after each 16 play drive, then the system will think that B has a nearly 100% chance of getting 5 yards every time B touches the ball.

Now imagine each team gets the ball for a fourth posession. Based only on what you've seen so far, which do you think is more likely to move the ball/score a TD/be able to run out the clock in a critical situation? Unless you're bad at probabilty, you'll almost certainly say B.

(In mathematical terms, if A has a 1/3 chance of scoring a TD each time they touch the ball and a 2/3 chance of getting no yards, and if B has a 100% chance of getting exactly 5 yards each time they touch the ball, then B will always score a TD on every possession, but A will score a TD only 19 out of every 27 posessions).

Points: 0

#113 by Subrata Sircar // Sep 15, 2010 - 6:33am

Now introduce turnovers into the mix, and watch the math reverse. Say that each team fumbles once every 10 non-TD plays and the opponents always recover.

Team A's plays are 33% TD, 6.7% TO, 60% 0.
Team B's plays are 90% 5, 10% TO.

If each team gets the ball on the 20, Team A will almost certainly score before it turns the ball over on downs or otherwise. Team B, on the other hand, will almost certainly not score before they turn the ball over.

While there is a value to controlling the ball, there is also a risk; the longer the drive, the bigger the chance that something bad happens and the drive ends. Sooner or later, someone drops a pass, or blows a block, and you've only made it 40 yards on 8 plays and have to punt anyway, while the boom-n-bust guy scored 10 minutes ago.

Hyperbole, of course - it is good to control the ball, and first downs lead to victory, but there are a few points worth thinking about.
1. You can often slow down the relentless, beloved-of-DVOA style offense with a reasonable offense of your own and good punting. Make them go 80 yards every time and they probably won't.
2. The home-run offense reverses the execution factor - it only has to execute/get-lucky once, while the defense has to be lucky every time.

Points: 0

#173 by MJK // Sep 15, 2010 - 10:26pm

Agreed...and so the rates of gain versus rates of turnovers do matter. If you turn your turnover percentage down really, really low, B is still favored, whereas if you turn the turnover percentage up, then you're right...A is favored.

That's where success points come in...bust plays get none, "successful" plays get 1, boom plays get a lot more than 1, and turnovers get negative. By tuning the definition of "success" and the weighting factors (the value of boom plays and the negative value of turnovers), DVOA is tuned to take these factors into account. At least, I assume this is what Aaron does when he tries to adjust the system to maximize correlation with wins.

And so the fact that the system does seem to like offenses that can sustain drives more than it likes quick strike offenses that can score quickly but that go 3-and-out a lot implies that the turnover rates are relatively rare and that the "B" type team really does win more than the A-type team...otherwise, the weighting factors would end up such that the A-team would have a better DVOA.

Of course, it doesn't always work, as you point out. There are ways to stop a team that DVOA loves, exactly as you suggest. However, the fact that such teams still correlate well to winning implies that opponents can't figure out how to stop most of them, or possibly that they can't stop them more often than not...

Points: 0

#115 by Rusty G. (not verified) // Sep 15, 2010 - 8:32am

Which is a little worrisome. I'd suspect that successful plays are a bit autocorrelated (momentum, "setting up" plays, tiring the defense, whatever it actually is), but if there's anything a process like DVOA is going to struggle with, it's the less binary aspects of "success."

I've long suspected (but obviously have no evidence) that ALY, for example, overstates the back's importance and understates backside blocks, second level blocks and the size of the initial hole opened by the line. You can draw corrolaries in this treatment to long passes, I think.

So maybe it's just all those years of Philadelphia riding the top all season, win or lose, but I will stubbornly, if still wrongly, insist that efficient, short passing teams are disproportionately favored by the system.

Points: 0

#139 by Thomas_beardown // Sep 15, 2010 - 3:10pm

ALY is the worst stat on FO. It hold barely any information that just plain old RB ypc doesn't have.

As for the setup plays, and things of that nature, I think over the course of the season it all averages out because no team is successful on every play.

Points: 0

#161 by tally // Sep 15, 2010 - 5:20pm

Success isn't binary in DVOA. I believe that there are degrees of success incorporated into the formula, so that a 9 yard run on 1st and 10 is more successful than a 5 yard run on the same down and distance, even though both are successful.

Something like RB success rate is binary, but that's not exactly what's being used in DVOA.

Points: 0

#163 by Thomas_beardown // Sep 15, 2010 - 5:30pm

Oh yeah, this too.

Also, each play is compared to how successful other teams are from the same situation. It's just measured in success points instead of say yards.

Points: 0

#15 by ammek // Sep 14, 2010 - 4:21pm

It's just VOA for the moment: no opponent adjustments, obviously.

But I had the same thought about the Packers. The D looked strong in holding Philly to 0 net yards passing in the first half, although it didn't fare so well against Mike Vick. The offense looked out of sorts with 2 picks. But apparently the running game and Rodgers' traditional renaissance on third down more than made up for it.

Also: I am taking this moment to observe the Packers' 3rd-ranked special teams. A once in a lifetime opportunity.

Points: 0

#20 by jmaron // Sep 14, 2010 - 4:24pm

I also thought the Packers D looked very good. I don't put much credence in a lot of offence by a team that is way down the way Philly was. Even if it ends up close I think such yards are much easier to come by. Teams get conservative when they have big leads.

Points: 0

#34 by Will Allen // Sep 14, 2010 - 4:51pm

I think there is a good chance that a large segment of the group that has been ripping McNabb for years is about to get an advanced education as to what average to below average quarterbacking does for offensive efficiency over the course of a season.

Points: 0

#72 by >implying impl… (not verified) // Sep 14, 2010 - 8:36pm

So, in your estimation, the Packers' D only really got aggressive on that final 4-and-1 stop? Because they were just giving up yards left and right before that. If it weren't for that 4-and-1 stop, and Philly went on to win a dramatic comeback, would you still say the same thing about Packers D being conservative for the entire second half? Your position seems awfully contingent upon Andy Reid's retarded decision to call a QB sneak from the shotgun.

Points: 0

#92 by RickD // Sep 15, 2010 - 12:03am

That would seem to depend on your definition of "conservative."

It's clear that the Packers' defense was excellent for the first half of the game.

In the second half...well, they only gave up four plays that that were longer than 20 yards (and two were Vick scrambles). But they had 9 plays where they gave up between 10 and 20 yards. Some people might think this is 'conservative', giving up ground while trying to make sure really huge gains are not made.

I don't see how calling this 'conservative' has anything to do with whether it worked or not. A lot of fans hate the 'prevent' defense, and it seems to fail a lot, but that doesn't mean it's not a conservative approach.

Points: 0

#19 by Will Allen // Sep 14, 2010 - 4:23pm

I believe that defensive ranking is the VOA, not DVOA, and thus does not account for the fact that the Vikings defense was facing the Saints' offense.

Points: 0

#21 by jmaron // Sep 14, 2010 - 4:27pm

good point. But still - 300 yards - 14 points. 5 yards per play.

Is that really bad defence even against an average team.

As one other poster mentioned DVOA ranks plays so a drive of 10 consecutive 5 yards plays is valued far higher than a bunch of busts and a big play or two.

Points: 0

#29 by jimbohead // Sep 14, 2010 - 4:46pm

Remember that NO missed two field goals, and that had nothing to do with quality of def. From MIN def VOA perspective, they allowed 20 pts.

Points: 0

#33 by Will Allen // Sep 14, 2010 - 4:49pm

One of those missed field goals came after a drive of very little yards, following a Favre interception.

Points: 0

#32 by Will Allen // Sep 14, 2010 - 4:48pm

Well, when the game was in the balance, and the Vikings needed possession of the ball, the Vikings defense failed to obtain it, so we should not discount the value of getting first downs. Having said that, this is how the interdependence in football is hard to capture statistically. I think the Vkings had 5 possessions in the 2nd half, and three times the Vikings offense went three and out, which certainly exposes the defense to yielding first downs.

Points: 0

#45 by Dan // Sep 14, 2010 - 6:19pm

The Saints had 5 pretty good drives in 10 possessions, two ending in touchdowns, one in a missed field goal, one in the end of the game, and one in a punt in Vikings territory after they started from their own 4. And no turnovers. That's a pretty good showing by an offense, even if it didn't result in many points.

Points: 0

#6 by Scott P. (not verified) // Sep 14, 2010 - 4:14pm

Surprised the Patriots' defense is 25th after Week 1. They looked pretty good to most observers. They did give up some yards and points late but not sure that is diagnostic.

Points: 0

#93 by RickD // Sep 15, 2010 - 12:12am

Well, only 6 teams gave up more points than the Patriots did.

Yeah, the defense looked good in the first half, but you can only ask so much of any statistical system. The defense was soft in the second half, and it seems only appropriate for the rankings to reflect that.

Points: 0

#119 by nat // Sep 15, 2010 - 10:33am

Yes, in the second half the Patriots defense was poor at preventing the things that DVOA values. The question is, did DVOA value the correct things?

The Bengals first drive of the second half started with 29:48 left. They were down four TDs, and could (un)reasonably hope to have the ball at most 20-24 minutes. That means that they needed to score - assuming the Patriots offense would go three-and-out on each and every possession - a TD every 5-6 minutes. They scored one in 6:54. Advantage Patriots, but a big DVOA win for Palmer and company.

Their second drive started with 20:13 left. They were down three TDs, and could hope to have the ball for 14-16 minutes, best case. They needed to score a TD every 5 minutes. They scored one in 4:51. Slight advantage Bengals, but a big DVOA win.

Their third drive started with 7:41 left. They needed to score three TDs, and could expect to have the ball maybe 4-5 minutes if they were successful, longer only if they failed. They needed to score a TD every 1:30 to have a chance. They scored in 3:44. Huge win for the Patriot's defense. Big DVOA win for the Bengals.

It might be possible to improve DVOA by tweaking the formula for success points when teams are up or down by many scores in the second half. I suspect there are other complications that would make it difficult to get this right. As you say, you can only ask so much from statistics. After that, you need to remember their limitations and context. For DVOA, it's weakness is that it's idea of success fits reality less the more the clock becomes the issue.

Points: 0

#125 by Aaron Schatz // Sep 15, 2010 - 11:56am

Here's the problem: Giving up yardage when you are winning by four touchdowns in the third quarter has, in fact, been shown to be a strong indicator that you will give up yardage in general. I specifically test the adjustments in DVOA to try to get them to balance the correlation between a) DVOA and wins and b) past DVOA and future DVOA by the same team. Adding stronger adjustments based on the score of the game in the third quarter *does not* improve these correlations. I know we want to believe that giving up points when you are up by four touchdowns is meaningless, but it isn't meaningless until you get into the fourth quarter. Anecdotally, we certainly know that teams can come back from 21-28 point deficits at halftime.

Points: 0

#131 by nat // Sep 15, 2010 - 12:51pm

Thanks for the comment, Aaron.

I understand there are complications involved. For example, DVOA measures the ability to optimize the next score. Adjusting it to measure a more appropriate idea of success during desperation time would naturally make it less predictive of itself. And that same adjustment would have very little if any benefit in overall correlation to wins, because we're talking desperation time, not close games, so the numbers are small.

Take the Bengals situation: Once they were down by four scores to a team that was dominating them, they were going to lose that game. So it does not matter one whit what measure of success DVOA used in the second half for the correlation to wins. But by using DVOA's optimize-next-score success formula, you naturally get a higher correlation to future DVOA.

As I recall, the adjustments you have tried have been about emphasizing or deemphasizing "garbage time" plays in calculating DVOA. I don't think you have attempted morphing the success formula itself. Have you?

The experiment I suggest is this: Considering ONLY plays in the last 20 minutes of a game while down 17 or more (three scores), determine what success formula correlates best with winning, while at the same time correlating well with itself (with opponent adjustments, naturally). I suspect the answer is a different success point formula than the one you use for the entire game.

I would not expect the success formulas to be completely unrelated. Giving up yardage is still worse than not doing so. First downs still have value. But the emphasis is likely to be different, and time-per-play may have a measurable value that correlates to winning and predicts future desperation-time performance.

Points: 0

#160 by Jerry // Sep 15, 2010 - 5:14pm

Once they were down by four scores to a team that was dominating them, they were going to lose that game.

Any time a team is down four scores, they're being dominated. On rare occasions, they do come back.

The emphasis is likely to be different, and time-per-play may have a measurable value that correlates to winning and predicts future desperation-time performance.

I've always thought that Clock DVOA would be a useful piece of the puzzle. I'm sure there are combinations of score differential and time left where average time per play changes, and whether teams are eating clock or using very little of it makes a difference. If you're up three points, kneeling three times to run out the clock after the two minute warning is a successful series.

Points: 0

#177 by nat // Sep 16, 2010 - 11:30am

Yes, I overstated the case a smidgen. Teams do sometimes come back. That's why I've taken to calling the situation "desperation-time" rather than "garbage-time".

It's the time when the game situation demands that your goal on each play be something other than "maximize the net expected value of the next score". By definition, it's the time when VOA's success formula doesn't correctly correlate with winning, even if DVOA still correlates well with future DVOA.

It's pretty obvious that desperation-time exists. There are lots of obvious to the point of being silly but still common situations where playing to maximize the next score on offense or defense is completely boneheaded. There are even more situations where DVOA-maximizing strategies look reasonable, but probably fail to maximize wins.

It's an open question as to how far "maximize DVOA" and "maximize chance of winning" measures of success diverge and for how long.

Points: 0

#165 by Eddo // Sep 15, 2010 - 5:34pm

Aaron, in the past you've said that you generally "bucket" the time remaining and score difference values to increase sample size. Now that the amount of play-by-play data you have has grown, as you have more years measured in real time and more historical years, have you revisited the buckets you use? Could you make the buckets more granular?

Points: 0

#175 by Aaron Schatz // Sep 16, 2010 - 9:58am

I revisit it all the time. Remember, this is version 6 of DVOA we're using right now, which means five major rounds of improvements and changes, not to mention the various minor improvements made in between re-numbering versions.

Points: 0

#178 by nat // Sep 16, 2010 - 11:46am

Aaron, I know you use the time remaining and point difference buckets to calculate the average success against which plays are judged. (that's the "OA" in VOA)

Do you also use time remaining and point difference as inputs to the success point formula itself? Or is the value of a particular play outcome (the "V" in VOA) determined by down and distance only?

Points: 0

#182 by Thomas_beardown // Sep 16, 2010 - 2:52pm

It's down distance and field position.

Points: 0

#183 by nat // Sep 16, 2010 - 3:51pm

Sorry. I knew that.

Points: 0

#180 by Eddo // Sep 16, 2010 - 1:50pm

Cool. Thanks for replying, Aaron.

Points: 0

#7 by drobviousso // Sep 14, 2010 - 4:16pm

When does Premium get updated with this information? I'm a first time Premium subscriber, and quite curious about Pittsburgh's punt and kick coverage VOA.

(Not complaining that it's slow, just curious about general time frame to be expected.)

Points: 0

#17 by Aaron Schatz // Sep 14, 2010 - 4:22pm

Updating Premium after Week 1 is pretty pointless -- a lot of the splits are just blank, or have one or two plays -- so that stuff all will be updated for the first time next Tuesday.

As far as the Patriots defense and the Bengals offense, there's no doubt that most of the Bengals offensive success came late in the game, but it still did happen, and those plays count. There actually are no adjustments for score in the third quarter -- we found that they made the system less predictive -- and the adjustments for score in the fourth quarter are much larger for running plays than for passing plays.

Points: 0

#31 by dolo (not verified) // Sep 14, 2010 - 4:47pm

That seems strange, I would think DVOA would account for what the average team does when up by three touchdowns in the second half. How does a team's probability of winning change when they go from three touchdowns up at the half to two touchdowns up with one quarter left to play? I would bet the winning percentage goes up, but giving up a 15 minute drive is horrible for defensive DVOA.

Points: 0

#81 by MJK // Sep 14, 2010 - 10:02pm

I would expect this is the effect of small sample size. Simply put, it's probably pretty rare for a team to be up by 3 TD's in the third quarter. And the fact is, giving up TD's in the 3rd quarter is never good... It totally makes sense that adding a "blowout" adjustment would reduce the predictive value of the system overall. Yes, there are the occasional games when the other team just gives up after halftime (this one, the blizzard game from a couple of years ago against the Cards, etc.), but most of the time, playing poorly in the 3rd quarter lets the other team get back in the game.

Also, I think you make a good point about giving up a 15 minute drive. It's more the number of successes you allow the offense to have.

If I hold a team to 3-and-out five times, and then give up one 15-play drive where the other team moves the ball well on every play, then drive (and score) wise, I've succeeded defensive five out of six times, but on a per-play basis I've succeeded only half the time.

Which means that, looking at my total history of defensive plays, I can expect only a 50% chance of success on any given defensive play (this is greatly oversimplified, of course).

Points: 0

#8 by Thomas_beardown // Sep 14, 2010 - 4:16pm

7 teams have +10% DVOA passing or better while having -17% rushing DVOA or worse (a number are more like -30%).

So much for making teams one dimensional.

Points: 0

#9 by Arkaein // Sep 14, 2010 - 4:18pm

Just checked the QB DVOA/DYAR numbers. Does Kolb's rushing DVOA set any kind of record (and not in a good way)?

Rate stats are always fun after a single game.

Points: 0

#18 by Thomas_beardown // Sep 14, 2010 - 4:22pm

Holy mackerel, that looks like a typo.

Did he sprint back 50 yards before fumbling?

Points: 0

#24 by Arkaein // Sep 14, 2010 - 4:30pm

Actually, he tried to run away from Clay Matthews and got just past the line of scrimmage (preventing Matthews from getting what could have been the 2nd of three sacks for the day), before Matthews stripped the ball and gave him his concussion.

As you can see in the table the official line is 1 run for 1 yard and a fumble.

Being a rate stat, -1644.7% sounds about right for a player who would run for a gain of one and then fumble on his AVERAGE play.

Points: 0

#105 by Yaguar // Sep 15, 2010 - 2:18am

The average run is worth maybe a 2 or 2.5 yard gain. 3 at best.

Yes, the league average is 4 yards per carry, but that can often come in groups like (0, 1, -1, 4, 3, 17), which is not nearly as useful as a constant 4. Not to mention the fact that fumbles can happen and such.

If you count his run and fumble as worth about -30 yards, then -30 could legitimately be seen as worth about -1500% of an average running play.

Points: 0

#176 by nuclearbdgr // Sep 16, 2010 - 10:14am

Kolb's numbers should improve:

Upon further review, Matthews had his first career three-sack game Sunday. The Elias Sports Bureau changed Matthews' second-quarter takedown of Eagles quarterback Kevin Kolb from a rush of 1 yard to a sack for no yards.
(courtesy of the Milwaukee Journal-Sentinel Online)

Points: 0

#181 by prs130@gmail.com // Sep 16, 2010 - 2:18pm

I would love to see the Eagles' defensive DVOA pre- and post-Bradley injury and the Eagles' offensive DVOA pre- and post-Kolb injury.

Points: 0

#14 by Yaguar // Sep 14, 2010 - 4:20pm

Arian Foster has a VOA of +68% against the Colts. Chris Johnson last year was +15%. Um, this is not good.

Points: 0

#95 by RickD // Sep 15, 2010 - 12:14am

Sounds good for Texans' fans.

Points: 0

#116 by Mr Shush // Sep 15, 2010 - 8:45am

Actually, I think it mostly sounds good for Titans and Jaguars fans, as their awesome running games get to go against that shonky run defense twice each over the rest of the season. Certainly no bad thing for Texans fans, or indeed Bengals fans, Raiders fans and some others, but above all, as Yaguar says, very, very bad news for Colts fans.

Points: 0

#120 by The Other Ben … (not verified) // Sep 15, 2010 - 10:39am

I think it's also good for football fans who are not Colts fans. Let the whining become reality-based for a season, see how that tickles 'em.

Points: 0

#123 by Eddo // Sep 15, 2010 - 11:46am

If the Colts struggling results in this kind of obnoxious schadenfreude, then I would say it's decidedly bad for those of us who are neither fans of the Colts nor their opponents.

Points: 0

#129 by Guy#1 (not verified) // Sep 15, 2010 - 12:38pm

I'm taking pleasure in your dismay over the obnoxious comments... it's layered schadenfreude.

Also, Patriots rule nah nah doo doo

Points: 0

#128 by Mr Shush // Sep 15, 2010 - 12:26pm

I happened to look briefly at the comments on some Colts/Indianapolis sites this week. Egads. Evidently Polian's a lousy GM, Freeney and Mathis are bums, and Peyton's a good player but not a patch on Bert Jones. The Colts are going 0-16. They're finally getting their just deserts for not building a proper football team that can run the ball and stop the run. Makes the people who continually inform the Houston Chronicle that Kubiak's an incompetent fool with the emotional intensity of a sack of spuds and Mario Williams is a horrible bust who will continue to leave the team rueing its failure to pick Bush or Young seem positively reasonable.

Points: 0

#136 by DW94 // Sep 15, 2010 - 2:38pm

Yeah, unfortunately I think this kind of egregious stupidity is common to most fans of all the NFL franchises.

Points: 0

#143 by Thomas_beardown // Sep 15, 2010 - 3:20pm

Due to my own personal belief that human beings cannot be this stupid, I attribute a lot this to a cathartic therapy. After a loss, they just want to yell at someone and they don't care if it makes any sense.

Points: 0

#150 by jebmak // Sep 15, 2010 - 4:11pm

Due to my own personal belief that human beings cannot be this stupid


Points: 0

#168 by DW94 // Sep 15, 2010 - 6:51pm

Well, anyone who thinks that Peyton Manning and Bill Polian are bums would have to be unbelievably stupid, right?

But seriously, I think what you've said is correct. However, I also think it's more than just being really angry that your team lossed that week. I think we could reasonably say that it's a mixture of stupidity, ignorance laziness, and that those conditions also come from multiple sources. Or in other words, people hold stupid beliefs on football and many other matters because of diverse reasons. And I think at this point I'm just stating the obvious at a very general and somewhat vague level so I'll quit.

Points: 0

#169 by nat // Sep 15, 2010 - 7:04pm

In some vague and very general way, this may be my favorite comment of the day. I don't know why I think that. I'm just saying.

Points: 0

#172 by DW94 // Sep 15, 2010 - 8:47pm

I'm glad that it may be your favorite comment of the day . . .I'm specifically glad.

Points: 0

#16 by jmaron // Sep 14, 2010 - 4:21pm

I'm not going to get into a DVOA argument here but where my view of a game differs strongly from DVOA it usually comes down to the following:

1) Team X made a bunch of great special teams plays or had a bunch of turnovers in their favour and were rated much higher than I would have rated them.
2) Team X made a bunch of first downs but not many big plays and didn't score many points.

Points: 0

#25 by Will Allen // Sep 14, 2010 - 4:36pm

I understand your view, but I think there is a tendency to underrate the value of simply maintaining possession of the ball, which is what getting first downs does. I also think there is a tendency to underrate the value of strong special teams play. I don't think DVOA is wholly accurate by any means. At times I don't think it fully captures the value of physically dominating the line of scrimmage, and the value of violently hitting the opposition's ball handlers consistently, or preventing that from happening, but I am cautious to discount the value of forcing turnovers, especially interceptions, racking up first downs, and good special teams play.

Points: 0

#111 by Subrata Sircar // Sep 15, 2010 - 6:17am

The biggest issue for me is that DVOA tries to serve two masters: narration/explanation and prediction. When James Harrison takes back an interception return 100 yards for a TD with no time left at the end of the half, that obviously is a *huge* part of the story of the game and why the Steelers won. It is just as obviously a non-predictive piece of data, in that even if you assume that Warner throws a pick to Harrison 9/10 times there, the return won't go for a TD all nine times.

Points: 0

#30 by Loose On The Lead // Sep 14, 2010 - 4:47pm

I have no problem with the notion that the Cowboys outplayed the Redskins. In fact, I think it's true. However, the degree to which it occurred is distorted somewhat by VOA because VOA doesn't reflect the impact of most penalties. In this case, it's not just that penalties hurt Dallas; it's that the Redskins' defensive pressure caused some of the significant infractions--primarily the holds. In those instances, without the penalties, the Skins might have gotten sacks, or at least a shot at a negative play. Instead, the plays just don't count.

Points: 0

#22 by Q (not verified) // Sep 14, 2010 - 4:27pm

It is amazing GB's Offense rated #6. They had 2 turnovers, gave up multiple sacks, and had under 300 yards of Offense against a non elite Defense.

Points: 0

#23 by Thomas_beardown // Sep 14, 2010 - 4:29pm

Total yards doesn't matter at all to DVOA. It's a rate stat, not a counting stat.

Points: 0

#27 by Arkaein // Sep 14, 2010 - 4:43pm

Subjectively I would have agreed with Q, but looking at the box score GB got 22 first downs in 64 plays (Philly was 16 in 55). That's 0.34 1st downs per play for GB compared to 0.27 for Philly.

GB may not be penalized too heavily for the turnovers because they both came on 3rd and long from near midfield. If you have to have a turnover that's the time to do it since anything short of a fairly big play will result in a punting situation anyways.

GB also had zero fumbles, which helps. I'm a little surprised that Philly wasn't penalized on offense more for their 3 fumbles (only one lost), and they probably should be marked down in charting if not DVOA based on the number of interceptable balls thrown by Kolb (3 by my count, two at Tramon Williams, and one thrown right at Clay Matthews).

Points: 0

#28 by ammek // Sep 14, 2010 - 4:44pm

And we don't yet know whether the Eagles' D is elite, non-elite, or Lions East. You may think you know, but you may also be wrong.

Points: 0

#35 by Eddo // Sep 14, 2010 - 4:56pm

Well, as others pointed out, the non-elite defense part isn't factored in; this is just VOA, not DVOA, so it doesn't factor in opponent strength.

Points: 0

#26 by putnamp // Sep 14, 2010 - 4:42pm

Seahawks? At number 1? Wha? Gonna just freeze time for a while and pretend this week never ends.

Points: 0

#36 by Keasley (not verified) // Sep 14, 2010 - 4:56pm

As a Seahawks fan I feel its incumbent upon me to post in this thread, even if I have nothing to say, to celebrate the Seattle Seahawks #1 total VOA for the season...since this will almost certainly be the last time they will be there until, maybe, 2012.

I'm sure there's a Pete Carollian adjective like 'pumped' or 'jacked' to describe my enthusiasm for Sunday's game but I doubt I share Pete's overall optimism for the season. That was just the kind of game where Seattle's home field advantage is a significant advantage (a bunch of Niners including Alex Smith, Mike Singletary, and Michael Crabtree seemed pretty rattled and some of the false starts, delay of games, timeouts, etc were costly) but Seattle could still have easily lost if SF had taken advantage of their early dominance. But none of that matters now. We're number 1!

Points: 0

#37 by JasonK // Sep 14, 2010 - 5:18pm

I have a strange feeling that the Giants' 1st overall ranking in VOA defense isn't going to last very long.

(Their 31st ranking in special teams, though, might be sticking around for a while.)

Points: 0

#51 by thendcomes // Sep 14, 2010 - 6:47pm

Dodge's performance was surpassed only by San Diego's punter, who at one point recorded a punt of 7 yards. How does someone like Dodge collect a paycheck in good faith? I've never seen so many awful punts in one game, not to mention how god awful he was in preseason.

Giants will not fix this until they have a punt returned on them and lose in the same game. Or some similar humiliation like punting it from the endzone and the opponent fielding it at the 30.

Points: 0

#63 by Gferguson (not verified) // Sep 14, 2010 - 7:43pm

Actually it was Kansas City's punter who recorded a punt of 7 yards. Scifres had a decent night considering he was punting to primarily short fields (and since no one seemed inclined to cover the kicks, there were a lot of touchbacks).

Points: 0

#76 by JasonK // Sep 14, 2010 - 9:18pm

Dodge only punted twice in the game. His 3rd punt was blocked (through no fault of his), and the only other time he touched the ball was on the ensuing Free Kick, which he boomed. He had some trouble early in the preseason, but by the Baltimore game (which I attended), he was killing it.

I'm willing to give the kid a chance-- 2 punts in his professional debut isn't much of a sample size upon which to make a judgment. I suspect that nerves had him rushing his drop a tiny bit, leading to his foot contacting the ball too early in its swing (thus creating a line-drive).

The craptacularity of the cover teams and the KO return blocking are more what concern me. (I was happy with Reynauld's game, though.)

Points: 0

#38 by glengarry // Sep 14, 2010 - 5:24pm

god, i love being a Titans fan right now. Catching Pittsburgh in Week 2? yes please!

Points: 0

#39 by Seattleite (not verified) // Sep 14, 2010 - 5:27pm

Every team is clearly ranked too randomly because the D in DVOA is absolutely vital. DAVE is way better than this. All tiems sucks.

You must have some defense data in order to have the preseason projections and DEF. RANK. Is there any way to estimate the defense adjustment with those data?

Points: 0

#60 by Mountain Time … // Sep 14, 2010 - 7:26pm

I have to disagree with you here. In years past I have found DAVE to be close to 100% worthless, since it places so much importance on the subjective pre-season guesses of FO writers. I just ignore DAVE altogether. Even VOA is merely interesting, and not all that valuable until we have enough data to make DVOA possible.

"Just look at that pumpkin."
-John Madden, looking at the moon.

Points: 0

#67 by Jerry // Sep 14, 2010 - 8:22pm

The projections are from Aaron's formula, not "subjective pre-season guesses of FO writers." You're still welcome to ignore them and DAVE, but that's what you're ignoring.

Points: 0

#77 by Thomas_beardown // Sep 14, 2010 - 9:22pm

There is a lot of subjective stuff in the projections, and a lot of guesses about what role the writers think a player will fulfill.

Points: 0

#89 by Dan // Sep 14, 2010 - 11:45pm

This guy ran the numbers on how predictive the preseason projections and early season VOA are, and found that DAVE does a pretty good job of combining them (although it tends to give a little too much weight to the projections and not enough to VOA).

Points: 0

#110 by Kulko // Sep 15, 2010 - 5:27am

Well having read his article I would say, that the idea behind Dave is sound, but the percentages used are pretty crap.

What he is finding, that they give way too much weight to their preseason predictions. This gets especially problematic if you add the finding by Advanced football Stats, that FO preseason predictions are not better than anybody elses guess.

Points: 0

#141 by Thomas_beardown // Sep 15, 2010 - 3:15pm

That finding was horribly flawed, and I believe only looked at one year's worth of predictions. Which was also FO's worst year of predictions.

Points: 0

#187 by drobviousso // Sep 17, 2010 - 11:52am

Also, they assume a collection of predictions around 8 wins and a collections of wins more spread out of the average of 8, when having the same correlation to actual wins, carry the same amount of information.

They do not.

Points: 0

#188 by Arkaein // Sep 17, 2010 - 12:40pm

Actually, they didn't use correlation as a test metric, in part because you CAN'T calculate correlation with a set of all equal values.

Instead they used metrics like mean error and mean squared error, which as you pointed out doesn't convey nearly as much information. You can get every record wrong in a division, but if you get the order right (which will produce a strong correlation) you've still determined the division champion. A fairly good mean error won't necessarily get the ordering right.

Points: 0

#189 by drobviousso // Sep 17, 2010 - 1:01pm

Ah, right. Thanks for the correction.

Points: 0

#40 by Anonymous1 (not verified) // Sep 14, 2010 - 5:32pm

I'm somewhat amused at NE's defensive rating considering they obviously changed coverages after going up 31-3. That said, it is only one game and it will all even out at the end of the year if they play consistently like they did in the first half.

Points: 0

#41 by brooklynfunbus // Sep 14, 2010 - 5:35pm

Is there a way to take the difference between two teams' DVOA and translate that into a point spread? For example, Atlanta's DVOA is 3.7%, Arizona's is 10.2%, yet the Falcons are 6.5 point favorites (as of Tuesday). Based on that data, it seems like there would be a lot of value in betting the Cardinals this week, but is there a way to quantify that?

Seattle (63%) getting 3.5 @ Denver (-16.4%) looks like another good value.

Points: 0

#42 by Arkaein // Sep 14, 2010 - 5:43pm

I think that's what the Premium Picks does.

If you don't feel like paying money, I had a field day in my (non-betting) Pick 'Ems league picking using the Preseason projections. I only got Indy wrong, although I went against the projections in picking GB over Philly (although GB was rated slightly higher, home-field advantage at estimated +17% DVOA would have given the edge to Philly).

Do not use this early season DVOA for picks. That's basically relying solely on the results of the first week. use DAVE instead, which combines Preseason DVOA projections with actual DVOA.

Points: 0

#117 by Podge (not verified) // Sep 15, 2010 - 8:55am

I'd urge you to consider these two facts:

Atlanta put up 3.7% VOA against the Steelers.

Arizona put up 10.2% VOA against the Rams.

I suspect that if you looked back at these figures at the end of the season Atlanta's DVOA for the game will be way higher than 3.7% and Arizona's will be way lower than 10.2%.

IIRC someone once said that an adjustment of about 12% or so is homefield advantage. You can use that as a vague indication. Not sure how you'd factor it in for large points spread, but its useful for games that have a small spread.

I think Premium picks does a lot better job of picking winners against the spread.

Points: 0

#44 by tabsports // Sep 14, 2010 - 6:18pm

Don't worry, Aaron, I have already been stabbing my eyeballs out for a day and a half. It's absolutely no shock we outplayed Washington in that game.

Points: 0

#61 by Mountain Time … // Sep 14, 2010 - 7:27pm

I didn't know you were involved in the Cowboys organization. What is your job title?

"Just look at that pumpkin."
-John Madden, looking at the moon.

Points: 0

#46 by UK Mr B (not verified) // Sep 14, 2010 - 6:21pm

...and the Raider Nation dies another quiet death...

Points: 0

#49 by Raiderjoe // Sep 14, 2010 - 6:41pm

Raiders going to win next week. Just had blimp on radar week 1. new QB and other new players need weke to get going. Patriots once lost 31-0 in openenrr then go on to win Super Bowl.

Raiders goingn to be fine.

Points: 0

#84 by zlionsfan // Sep 14, 2010 - 10:39pm

The Lions once opened a season losing 45-0 and went on to make the NFC Championship (where they lost to the same team 41-10).

You're right, the first week doesn't always tell you in which direction your ship is sailing.

Points: 0

#96 by RickD // Sep 15, 2010 - 12:18am

Heh, nice 'typo'.

Personally, I think the Raiders' problems are more like a blimp on the radar than a mere blip.

Points: 0

#112 by Raiderjoe // Sep 15, 2010 - 6:27am

Not typo. Blimp on radar correct. When lookign at radar screen, peoeple lokking for alien spaceshipps or terrorists. Not blimp. Blimp not oging to bother anyone. So thats why say blimp on rsdar screen- not concerned about it. JUst like Raiders not cocnern about loss on Sunday. GBoing to rebound against Ramms, get to 1-1 and be one way to playoffs.

Points: 0

#133 by Will Allen // Sep 15, 2010 - 1:34pm

Sure, Raiderjoe, this thinking works just fine.....until the day comes that aliens build their spacecraft in the shape of blimps, and fly them reeeeaaalll slooooooow! Ya' look up in the sky, and expect to see Snoopy, and instead get sucked up in a tractor beam, and have medical experiments performed on you by what appear to be giant, air-breathing, crustaceans! Who no doubt are Bronco fans!!

Points: 0

#146 by Marko // Sep 15, 2010 - 3:43pm

How do we know that this hasn't actually happened to Raiderjoe? Or to Al Davis, for that matter?

Points: 0

#140 by bouch (not verified) // Sep 15, 2010 - 3:12pm

Well, based on all the other typos in your messages, its completely understandable how someone would think "blimp" was a typo. Ye gods, get a spell checker!

Points: 0

#145 by jimbohead // Sep 15, 2010 - 3:22pm

Asking raiderjoe to use a spell check is like asking Mona Lisa to smile for the painter.

Points: 0

#148 by Dan // Sep 15, 2010 - 4:03pm

I thought it was an eggcorn rather than a typo.

Points: 0

#47 by ammek // Sep 14, 2010 - 6:37pm

Obscure query of the week: weather adjustments. Are the numbers a projection for the whole season, or just week one?

If it's the latter, it seems that weather adjustments can be very different for two teams playing in the same game. Witness TB/Cle, where the Bucs have +1.2 pts and the Browns only +0.4. Yet Tampa had only one more punt (7:6) and one more kickoff (4:3).

If it's the former, it's messing up the non-adj VOA number.

Points: 0

#48 by Raiderjoe // Sep 14, 2010 - 6:39pm

didnt read comments yet,. only saw rankings. Computer obivously broken. no way Raiders are 32nd best tmea in league. Loso t best tema in league in week 1 -titnas. Tennesssee Tiants cant argue best showing of any team in week 1. They ranked 2nd, so how Raiders rnaked 32? that is wrong. Raiders should be 17th- the best loser team in 1st fwewek. Dave going to rise netx week for sure.

Points: 0

#53 by Josh // Sep 14, 2010 - 7:02pm

You didn't follow the zlionsfan template.

Points: 0

#54 by TomC // Sep 14, 2010 - 7:08pm

Raiderjoe doesn't follow The Man's "rules" and "templates" and "conventions of syntax, diction, spelling, and grammar."

Points: 0

#69 by Jerry // Sep 14, 2010 - 8:31pm

The autumn wind isn't a template.

Points: 0

#64 by DGL // Sep 14, 2010 - 7:48pm

Is computer drnuk?

Points: 0

#73 by Raiderjoe // Sep 14, 2010 - 8:49pm

yes, iti is like computer durnk

just nor tight. Raiders lose to maybe bestt team in league. Titams real dela. Going to make playoffs this year.
No shame in Raiders losiing to Titans. Definitely not derserve 32 rnaking. 17-20 area more like it.

Points: 0

#184 by ChicagoRaider // Sep 16, 2010 - 5:04pm

Um, did you see the same offensive line that I saw? If Campbell can't get a pocket, can he at least get an ejection seat?

Points: 0

#185 by Raiderjoe // Sep 16, 2010 - 5:29pm

problems oging to be smoothed over this weke and Raiders going to get job doen on Sunda

Points: 0

#55 by MainerRaider (not verified) // Sep 14, 2010 - 7:11pm

For once, Raiderjoe has a point: it is odd that the Titans are #2 after beating #32. I'm not saying this to argue about their rankings, because FO's stats are excellent; I'm saying this with the faint hope of improvement for that 8-8 season we dream of these days.

Points: 0

#58 by RichC (not verified) // Sep 14, 2010 - 7:25pm

Its not odd at all. In fact, its expected. Its just VOA.

Points: 0

#97 by RickD // Sep 15, 2010 - 12:20am


In similar news, the biggest winner and the biggest loser often leave the same poker table in the casino.

Points: 0

#59 by Josh // Sep 14, 2010 - 7:25pm

These ratings aren't adjusted for opponent strength.

Points: 0

#70 by Big Johnson // Sep 14, 2010 - 8:31pm

im actually curious how theres not a direct coorelation to where each of the teams that played each other ended up. in a system like VOA wouldnt it make more sense that if....

the seahawks are #1 then the niners are #32?
the titans are #2 then the raiders are #31?
and so on and so on.

Isnt it true that the average of all the VOA's are 0? This would make it that lets say the seahawks have a VOA of 70 then the niners would have a VOA of -70.

Obviously im way off in my thinking but i guess im wondering how the system works

Points: 0

#75 by Andrew Potter // Sep 14, 2010 - 8:52pm

Not necessarily. Take a missed field goal (not blocked, just plain old wide right missed), for example. If I understand this right, for the "offensive" specials teams, that's a bad play with a negative predictive value. For the "defensive" special teams, it's not predictive at all so has no value in this system. One team's special teams loses DVOA, but the other doesn't necessarily gain it.

Points: 0

#78 by Thomas_beardown // Sep 14, 2010 - 9:24pm

Also, fumble recoveries and certain penalties are basically counted as luck.

Points: 0

#109 by Mountain Time … // Sep 15, 2010 - 4:57am

The VOA of each team is calculated independently of each other, so if you add them together it will not necessarily equal 0.

"Just look at that pumpkin."
-John Madden, looking at the moon.

Points: 0

#57 by Fan in Exile // Sep 14, 2010 - 7:24pm

I think the Bronco's ratings are pretty screwed up, because they don't line up well with the game. The Defense played much better than 31.8% although they needed more pressure.

The special teams was way worse than that, whatever penalty you take away from the defense should be given to the special teams. The offense just wasn't that good, as shown by their inability to actually put points on the board which is kind of the object of the game.

Points: 0

#85 by tornadot // Sep 14, 2010 - 10:50pm

Offense seemed to show promise when they seemed to go quick strike, 2 minute drill style. Why they didn't keep doing that, who knows. Though those super short offensive series might wreak havoc on the poor D in that Florida heat. Defense did okay but gave up quite a few rushing yards (Okay this is Jaguars and they run the ball very well)...do penalty yardage get accounted into these ratings? Those two McBean penalties hurt.

Points: 0

#122 by tgt2 (not verified) // Sep 15, 2010 - 11:45am

Most penalties are not included as they didn't have predictive power. Defensive pass interference is factored in, and possiby false starts. I'm not sure about the latter.

Points: 0

#62 by glkatz81 // Sep 14, 2010 - 7:30pm

Hi guys. New to the site - love it. Book was best purchase I made in a long time (and Kubiak has me off to a good start in fantasy football for first time in forever). So last week I took the mean wins figure (http://www.footballoutsiders.com/dvoa-ratings/2010/2010-dvoa-projections), did some tweaking, and ended up doing pretty well against the spread. I was wondering if the mean wins figure is something that is updated weekly / distributed by FO. Thanks for opening the eyes of someone who thought they understood football to what is really going on in this great game.


Points: 0

#65 by Josh // Sep 14, 2010 - 7:58pm

You'll only see mean wins on the playoff odds page, but you will see Estimated Wins and Pythagorean Wins on the weekly DVOA analysis page.

Points: 0

#66 by glkatz81 // Sep 14, 2010 - 8:16pm

Thank you. I see the mean wins - thanks for pointing me to that. This might be a foolish question but isn't this page the weekly DVOA analysis page? If so I don't see estimated wins or Pythagorean wins.

Probably a newbie mistake and something I'm just missing that is starting me in the face.

Also, I went to glossary - what is difference between estimated wins and mean wins? I understand means wins to be per the 10k simulations. I understand (I think) Pythagorean wins from the book, however, not sure what estimated wins are. I just don't recall reading about it.


Points: 0

Save 10%
& Support Aaron
Support Football Outsiders' independent media and . Use promo code SCHATZ to save 10% on any FO+ membership and give half the cost of your membership to tip Aaron.