So the new 2007 DVOA projections were released today ('boys ranked 26th)

[thekavorka]

http://www.footballoutsiders.com/2007/09/05/ramblings/dvoa-rankings/5437/

Top 5

New England
Philly
Jacksonville
Baltimore
Washington

here's a note they made about the cowboys:

Dallas: A quick summary of the biggest issues here: 1) Third-down offense far better than overall offense; 2) Romo regressed at the end of the season; 3) worst defense in the NFL last six weeks of 2006; 4) new coordinators means learning new systems; 5) extremely unlikely that entire offensive line will remain healthy for all 16 games again.

For an explanation of what DVOA is and how they calculated it: http://www.footballoutsiders.com/methods.php

 

[tomson75]

It's a wonder we managed to win ONE game last year, nevertheless NINE...and make the playoffs.

 

[DBoys]

Washington top 5 after last year?? This dude is a crackhead.

 

[theogt]

They're probably basing it off of the last few games, which makes total sense.

 

[Oh_Canada]

Baltimore and there merry-band of pensioners will be lucky to win eight games this season....lucky.

Washington??? Improved yes....but top five....are you kidding me??

Jacksonville....with David Garrard at QB and no real threat at wideout??

Other than that....it's ok.

1.New England
2.Indy
3.San Diego
4.New Orleans
5.Philly

6.Seattle
7.Dallas

 

[AdamJT13]

http://www.footballoutsiders.com/2006/09/06/ramblings/dvoa-rankings/4194/

Last year's projections -- further proof that past performance is never accurate at projecting what will happen in the future.

 

[FuzzyLumpkins]

http://www.footballoutsiders.com/2006/09/06/ramblings/dvoa-rankings/4194/

Last year's projections -- further proof that past performance is never accurate at projecting what will happen in the future.

I would say it has more to do with how they determine which statistics to use and how they weigh them as it correlates to success would be the issue than it is an indictment of statistical analysis. That and injuries.

 

[AdamJT13]

I would say it has more to do with how they determine which statistics to use and how they weigh them as it correlates to success would be the issue than it is an indictment of statistical analysis. That and injuries.

Injuries, free agency, coaching changes, players improving or getting worse, luck, rookies, etc., etc., etc.

Anytime you try to project the future using past statistics, you're going to be only so accurate. And when the team you project to be the worst in the NFL ends up in the conference championship game, you should realize that your efforts are futile.

 

[TunaFan33]

http://www.footballoutsiders.com/2007/09/05/ramblings/dvoa-rankings/5437/

Top 5

New England
Philly
Jacksonville
Baltimore
Washington

here's a note they made about the cowboys:



For an explanation of what DVOA is and how they calculated it: http://www.footballoutsiders.com/methods.php

While I'm not going to argue what they said about Dallas...

Jax with David Gerard?? Balt sans Adalius Thomas and a p-poor O?? Philly and their "can't stop the run even if it means to save their own lives" D??? WASHINGTON????????????????????????!!!!!!!!!!!!!!!!!!!!!!

Obviously these guys have had too much to drink.

 

[FuzzyLumpkins]

Injuries, free agency, coaching changes, players improving or getting worse, luck, rookies, etc., etc., etc.

Anytime you try to project the future using past statistics, you're going to be only so accurate. And when the team you project to be the worst in the NFL ends up in the conference championship game, you should realize that your efforts are futile.

Their entire metric is based on determining what they think is a good play or a bad play in a specific arbitrary situation and then assigning arbitrary points to it. It fails for the same reason zone ratings and most defensive statistics do in baseball as its based on arbitrary nonsense.

I will admit that statistical analysis of teams at best is sspecious because there are too many interrelationships that are not represented by raw numbers.

At the same time statistical analysis of individual performance has shown to be very successful especially in baseball.

Football is a different animal because you typically have much smaller sample sizes other than with QBs and RBs to work with and there are a TON more injuries to deal with that skew the data. At the same time i would imagine if you were to look at progressions of players and league averages etc that you could forecast rather well how QBs, RBs, and WRs will perform barring injury.

just because these guys came up with arbitrary nonsense doesnt meant that statistical forecasting has no merit.

 

[AdamJT13]

Their entire metric is based on determining what they think is a good play or a bad play in a specific arbitrary situation and then assigning arbitrary points to it.

They claim that it's all based on what the NFL averages are in identical situations, so it's not simply an arbitrary judgement about what is "good" or "bad." My problem with their system is that it has never been revealed, so there's no way to know whether the system is flawed. (Even if they'd prefer not to reveal all of the specific numbers to keep others from using their data, they should at least fully explain the system.)

At the same time statistical analysis of individual performance has shown to be very successful especially in baseball.

Projecting individual performance is different from projecting how teams of 53 individuals -- teams that are changing and evolving from year to year and from week to week -- will perform. Players don't change identities from one year to the next.

just because these guys came up with arbitrary nonsense doesnt meant that statistical forecasting has no merit.

I should have specified that I was talking about projecting how NFL teams will do. Statistical forecasting of teams' wins and losses from one year to the next has very little merit, unless you don't mind a system that's often wrong.

 

[superpunk]

http://www.footballoutsiders.com/2006/09/06/ramblings/dvoa-rankings/4194/

Last year's projections -- further proof that past performance is never accurate at projecting what will happen in the future.

Once they get into next season projections, they're really outside their wheelhouse. It gets better once they use the season data to project playoff teams and winners, btu it's still not what they do best.

 

[Yakuza Rich]

Injuries, free agency, coaching changes, players improving or getting worse, luck, rookies, etc., etc., etc.

Anytime you try to project the future using past statistics, you're going to be only so accurate. And when the team you project to be the worst in the NFL ends up in the conference championship game, you should realize that your efforts are futile.

There's also the issue of them using some questionable ways of projecting the future.

In their Commanders overview they claim that the Skins should be much better on defense simply because their 3rd down defense was so bad last year it almost has to go up. While that's true, this neglects the fact that 3rd down defense correlates with 1st and 2nd down defense and the Skins were poor on 1st and 2nd down defense and atrocious on 3rd down defense. So it's more likely that the Skins will improve on defense, but they would still be a lousy defense since they are so poor on 1st and 2nd down.

Then they talk about the "Peter Principle" where a team is usually going to bounce back to their old form after one off season. It's a commonly held principle in baseball and they believe it can work in football, although they show no statistical backing of that.

Then there's their claim that Parcells' defense was more aggressive than Phillips defense because Parcells blitzed 6 or 7 pass rushers more than Wade. However, that neglects the fact that Parcells' blitzing of 6 or 7 is a small amount of the time and Phillips actually blitzes much more, but usually only sends 5 players and eschews sending 6 or 7 guys.




YAKUZA

 

[blindzebra]

Aren't these the same clowns that predicted records by the baseball equations for runs scored and runs against?

They had Philly with 10 wins and us with 6 when our points for and against were nearly identical, and when you used their equation to arrive at the won loss record we actually would have a better record than Philly.

Nothing better than math guys with a bias.

 

[percyhoward]

1) Third-down offense far better than overall offense

IOW, our QB was clutch.

 

[theebs]

that website is garbage.

 

[theogt]

that website is garbage.

FootballOutsiders.com? No way, dude. They're one of the best.

 

[superpunk]

FootballOutsiders.com? No way, dude. They're one of the best.

Indeed...

They just get a little nuts when they move into the realm of predictions.

 

[thekavorka]

The DVOA is a better predictor of a team's wins than either the team's previous seasons's wins, points allowed/points scored, or yards allowed/yards scored.

Then again, I think it's pretty pointless in the first place to use stats to form a prediction. There is no magic formula, objective or subjective, that can predict an NFL season. That's just how the NFL is now.

 

[AdamJT13]

The DVOA is a better predictor of a team's wins than either the team's previous seasons's wins, points allowed/points scored, or yards allowed/yards scored.

I think you mean indicator (same season as the stats used), not predictor (the next season).

According to one blogger's research, DVOA is no better at predicting each team's wins in the following season than are the Las Vegas over/unders -- and are even worse than simply expecting every team to win the same number of games as the previous season, as well as worse than simply expecting each team to go 8-8.