Writers of Pro Football Prospectus 2008

Most Recent FO Features

BradyTom00-10.jpg

» Week 4 DVOA Ratings

Five different teams from last year's DVOA top eight rank in the bottom half of the league through four weeks of 2014. What can we learn from other teams with similar starts in the past?

16 Jan 2007

Postseason DVOA II

by Aaron Schatz

Hey kids, time for postseason DVOA ratings. Last year, we ranked just the 12 playoff teams. This year, based on reader requests, we've ranked all 32 teams, whether they are in the playoffs or not. Any week where a team did not play in the postseason is treated as a bye week. All numbers are weighted DVOA. That means that Weeks 1-5 are not included, while Weeks 6-11 are somewhat discounted.

Let's be perfectly straight here: by this point, this is resulting in some goofy rankings. The weighted DVOA formula is not designed to be used like this. Teams that played a couple strong games at the end of the year show up very high, because they have two empty weeks at the end of the season. Green Bay and Buffalo in the top ten? Not if they had actually played games the last two weeks.

If you want to see the regular total season ratings, you can just use the Just the Stats menu above.

Commentary now up on FOXSports.com as of Wednesday morning. It includes just the 8 playoff teams, and the order has been changed so that teams still alive are listed above teams that were eliminated this weekend.

* * * * *

To save people some time, we remind everyone to put their angry troll hatred into the official zlionsfan angry troll hatred Mad Libs form:

<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>

I'm not going to bother to run the whole DVOA explanation; if you are new to the website, you can read about it here. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.


TEAM WEI.
DVOA
LAST
WEEK
W-L WEI OFF
DVOA
OFF.
RANK
WEI DEF
DVOA
DEF.
RANK
WEI S.T.
DVOA
S.T.
RANK
1 BAL 39.5% 1 13-4 6.1% 12 -29.1% 1 4.3% 6
2 IND 32.4% 3 14-4 32.8% 1 0.3% 12 -0.2% 19
3 SD 32.1% 4 14-3 24.6% 2 -3.2% 11 4.2% 7
4 JAC 23.7% 5 8-8 11.8% 8 -11.8% 6 0.1% 17
5 NE 22.1% 2 14-4 6.2% 11 -11.9% 5 4.0% 8
6 NO 16.6% 8 11-6 18.0% 5 1.2% 13 -0.2% 18
7 PHI 14.1% 6 11-7 18.2% 4 2.6% 17 -1.5% 24
8 CIN 10.0% 9 8-8 19.3% 3 10.7% 29 1.3% 16
9 GB 9.0% 10 8-8 -7.1% 21 -19.8% 2 -3.7% 29
10 BUF 7.0% 16 7-9 -6.8% 20 -5.8% 10 8.0% 1
11 CHI 6.6% 7 14-3 -11.2% 26 -12.5% 4 5.2% 3
12 DAL 5.8% 13 9-8 11.8% 7 9.4% 26 3.4% 10
13 CAR 4.8% 11 8-8 -7.4% 22 -14.7% 3 -2.5% 26
14 PIT 4.6% 12 8-8 7.8% 10 1.3% 14 -1.8% 25
15 MIA 3.1% 17 6-10 -10.0% 25 -11.4% 7 1.6% 14
16 TEN 2.1% 14 8-8 1.1% 17 3.8% 19 4.7% 5
TEAM WEI.
DVOA
LAST
WEEK
W-L WEI OFF
DVOA
OFF.
RANK
WEI DEF
DVOA
DEF.
RANK
WEI S.T.
DVOA
S.T.
RANK
17 NYG 1.6% 15 8-9 5.6% 14 3.4% 18 -0.6% 20
18 NYJ 0.2% 18 10-7 3.4% 15 8.0% 20 4.8% 4
19 KC -3.7% 19 9-7 5.9% 13 8.8% 23 -0.9% 22
20 ARI -4.7% 20 5-11 8.5% 9 12.4% 30 -0.8% 21
21 SEA -6.6% 27 10-8 -8.3% 24 1.9% 15 3.6% 9
22 ATL -10.7% 24 7-9 0.1% 18 8.2% 22 -2.6% 27
23 STL -11.2% 21 8-8 15.2% 6 19.6% 32 -6.8% 32
24 DEN -12.1% 22 9-7 -12.4% 28 1.9% 16 2.2% 11
25 MIN -14.8% 23 6-10 -19.1% 30 -9.2% 9 -4.9% 31
26 DET -15.6% 29 3-13 -11.8% 27 9.9% 28 6.1% 2
27 WAS -15.7% 28 5-11 1.5% 16 19.4% 31 2.1% 12
28 SF -16.0% 26 7-9 -7.9% 23 9.4% 27 1.3% 15
29 HOU -16.4% 25 6-10 -3.6% 19 9.0% 25 -3.7% 30
30 TB -25.6% 30 4-12 -18.4% 29 9.0% 24 1.8% 13
31 OAK -26.0% 31 2-14 -34.3% 32 -11.3% 8 -3.1% 28
32 CLE -28.7% 32 4-12 -19.3% 31 8.0% 21 -1.4% 23

Here are the one-game DVOA ratings for the first round of the playoffs. Remember that these include opponent adjustments. Yes, those really are the ratings for San Diego and New England, even though the Patriots won the game. That's why the Patriots drop from second to fifth in the table above.

TEAM TOT OFF DEF ST
IND 67% 6% -50% 11%
BAL 10% -56% -53% 13%
NO 33% 35% 5% 4%
PHI -2% 16% 16% -2%
CHI -30% -19% 9% -2%
SEA 37% 17% -11% 9%
NE -25% -45% -13% 6%
SD 71% 15% -59% -2%

The Colts had a defensive DVOA of -76% against Kansas City and -50% against Baltimore, which has moved their weighted DVOA on defense all the way up from 26th to 12th over the last two weeks. Is there any precedent for this? Actually, it turns out there is.

I went looking for every team with a below-average defense that had two straight very good defensive games in the playoffs, as far back as the DVOA ratings go, which is 1997. There are a few of them, but none of them turned things around quite like the Colts.


Year Team Season Change 18 19 20 21
2006 IND 11.3% -74.6% -76.2% -50.4% ? ?
2005 NE 10.5% -40.3% -47.9% -11.6%    
1998 DEN 0.4% -40.0%   -31.5% -47.8% -27.8%
1999 WAS 2.5% -31.6% -20.8% -37.4%    
2006 SEA 4.5% -17.4% -24.1% -10.7%    

"Change" represents the difference between the team's defensive DVOA during the season and the average of its first two games in the playoffs. The 2005 Patriots actually come the closest, but unlike the Colts, the Patriots' defense had been improving during the second half of the season -- the change didn't show up suddenly in the playoffs. This list also doesn't help us figure out if the Colts can keep this going. Only one of these teams survived to play three games in the postseason, and the 1998 Broncos defense was far better during the regular season than the 2006 Colts defense was.

What about on offense? There isn't even one team with an offensive "change" above 30%. Here are the three mildly similar teams:

Year Team Season Change 18 19 20 21
2005 CAR -3.2% +26.1% 28.5% 23.8% -62.0%  
2001 PHI -3.4% +24.3% 21.9% 26.7% 7.4%  
2003 CAR -7.2% +12.6% 4.8% 20.4% 19.0% 39.8%

The 2005 Panthers did play a third game -- and their postseason offensive improvement completely disappeared. The 2001 Eagles weren't that spectacular in their third game either. The 2003 Panthers did keep their offensive improvement going all the way through the Super Bowl. Unfortunately, their defense -- which had been good all year -- had a bad game in the Super Bowl and they lost to... wait for it... the New England Patriots.

Posted by: Aaron Schatz on 16 Jan 2007

421 comments, Last at 25 Jan 2007, 9:02pm by Pat

Comments

301
by Pat (not verified) :: Thu, 01/18/2007 - 1:15pm

#298: The original graph is actually theoretical, not average. It's just based on the fact that the value of the football at your own goal line is -2 points (it's a safety) and the value of the football at your opponent's goal line is 7 points (it's a touchdown). Draw a straight line inbetween.

When you actually measure it, you find it's basically identical to the theory. I agree that it could change based on the type of offense, but not as much as you think, I'd imagine. The basic rules of football don't change that much.

302
by Pat (not verified) :: Thu, 01/18/2007 - 1:17pm

#300: I searched for "predictions" in this thread, and I didn't find DVOA predictions regarding the New England defense mentioned once. I have no interest in discussing that, and it's immaterial to the point.

303
by Rich Conley (not verified) :: Thu, 01/18/2007 - 1:20pm

302

The fact that DVOA predicted New England to have the 27th best defense in the league is immaterial to our contention that DVOA is missing something about the way the New England defense works?

304
by Pat (not verified) :: Thu, 01/18/2007 - 1:23pm

We have seen that DVOA consistently ranks the Patriots lower on defense than their points and turnovers would indicate they should be.

They are ranked third in defensive VOA. They are ranked third in points per drive.

The only reason they're ranked lower in DVOA are opponent adjustments. OK?

305
by Pat (not verified) :: Thu, 01/18/2007 - 1:23pm

#303: Yes. Go read how the DVOA projections work.

306
by Rich Conley (not verified) :: Thu, 01/18/2007 - 1:24pm

And opponent adjustments are part of DVOA. So therefor....

307
by Pat (not verified) :: Thu, 01/18/2007 - 1:26pm

#306:

In other words, you don't agree with adjusting for the quality of the opponent.

308
by B (not verified) :: Thu, 01/18/2007 - 1:28pm

303: You're confusing actual DVOA with the pre-season DVOA projections.

309
by mactbone (not verified) :: Thu, 01/18/2007 - 1:29pm

Re 307:
Well, obviously the Pats only play as hard as they need to.

310
by Rich Conley (not verified) :: Thu, 01/18/2007 - 1:31pm

307, no I do agree with adjusting for quality of opponent.

I just dont think we're doing it correctly. I also dont think it should be done in the linear fashion that we're doing it.

I also think the emphasis on yards is given too much credence. DVOA just has a problem with the Patriots, and I havent looked at enough other teams to see if it misvalues things the same way. The patriots offense is nowhere near as good as DVOA says, and the defense is better than it says.

311
by Pat (not verified) :: Thu, 01/18/2007 - 1:36pm

I just dont think we’re doing it correctly.

So what? It doesn't matter if we're doing it correctly or not - that just affects where the Patriots end up. The Pats faced worse offenses than average. Therefore, their defense, by any opponent adjustment, would go down.

DVOA just has a problem with the Patriots

By ranking them 3rd without opponent adjustments? That's a problem? Do points per drive have a problem with the Patriots, as well, because that's where they rank there, too.

What statistic doesn't have a problem with the Patriots? (And why is there a global anti-Patriots statistic campaign?)

312
by Pat (not verified) :: Thu, 01/18/2007 - 1:47pm

The patriots offense is nowhere near as good as DVOA says

Patriots offense, DVOA: 7th.
Patriots offense, points per drive: 5th.
Patriots offense, points per game: 7th

Why isn't the Patriots offense as good as DVOA says again? Every metric imaginable basically puts them at between 5th and 7th.

313
by Rich Conley (not verified) :: Thu, 01/18/2007 - 1:51pm

Pat, ideally, when you look at a team's VOA, it should be a pretty inconsistent number, while the DVOA should actually help smooth the graph out. It should bring down the peaks where you're getting high VOA, because chances are, you were playing a crappy team. It should bring up the Valleys where you're playing a good team, and your VOA suffers.

The problem is, from what I can see, its not doing that, its doing the opposite in a lot of cases. Teams blowing out awful teams and getting a DVOA boost. Teams losing to good teams and getting knocked down. I just think viewing this in a linear way just doesnt work.

314
by Pat (not verified) :: Thu, 01/18/2007 - 1:58pm

#313: Actually, it does. There's a blog post from Aaron about that. The variance in DVOA is less than the variance in VOA.

315
by Rich Conley (not verified) :: Thu, 01/18/2007 - 1:59pm

312.

If they were as good as the metric said, they should have torched the charger's defense this week. They couldnt move the ball.

I've seen too many games where you get 3 patriots drives that are 3 or 4 plays and then a punt, followed by one drive that goes 80 yards on 14 plays and gets a TD.

DVOA sees that as 14 successful plays, and 9-12 failed plays.

I see it as one successful drive, and three failed drives.

For the last two years, DVOA has said the Pats offense is better than their Defense. In almost every loss, it has been the offense that underperforms, and not the defense. This year, see the colts/miami/jets/broncos/chargers. ALL offensive failures. Look at last year's playoff loss..offensive failure.

316
by Rich Conley (not verified) :: Thu, 01/18/2007 - 2:00pm

314

Again, it is across the league, but isnt in a lot of specific cases.

317
by Pat (not verified) :: Thu, 01/18/2007 - 2:06pm

If they were as good as the metric said, they should have torched the charger’s defense this week. They couldnt move the ball.

That's cherry-picking. All offenses have bad days. All defenses have bad days. It's a game. Teams don't play consistently.

In almost every loss, it has been the offense that underperforms, and not the defense.

Well, in almost every loss you quoted, they were facing a very good defense. Your eyes are not opponent adjusted.

Again, it is across the league, but isnt in a lot of specific cases.

Do you have proof for that? Are those cases consistent, or just random noise? According to Aaron, who has the actual numbers, it's just random noise.

For the last two years, DVOA has said the Pats offense is better than their Defense.

By VOA, their defense was better than their offense this year. Their defense looked better because they played crappy offenses (and above-average defenses, mind you), so it's not surprising that in your eyes, the defense looked better.

318
by Rich Conley (not verified) :: Thu, 01/18/2007 - 2:28pm

Pat, its not random noise when its consistently that way.

colts 11.3
miami -8.9
jets 10.7
broncos -3.7
chargers -1.1

The only good defense there is Miami. The rest are average to awful. They got shut down, and badly by all of those.

I think the Patriots offense is just schemed wonderfully. When an opposing coach fails to adapt, they crush people. When the coach makes adjustments, the offense just isnt talented enough.

319
by Pat (not verified) :: Thu, 01/18/2007 - 2:34pm

Pat, its not random noise when its consistently that way.

What's funny is I can actually just repeat my previous statement:

Do you have proof for that? According to Aaron, who has the actual numbers, it’s just random noise.

320
by Pat (not verified) :: Thu, 01/18/2007 - 2:45pm

I think the Patriots offense is just schemed wonderfully. When an opposing coach fails to adapt, they crush people. When the coach makes adjustments, the offense just isnt talented enough.

That's the Tennessee Run Defense Problem again - you can only measure what you see. Yes, you can cherry pick examples where the offense failed when they should've succeeded. But they've also succeeded where they should've succeeded (vs Green Bay, vs Cincinnati, vs Tennessee, vs Minnesota).

Any offense can be beaten, and any defense can take itself out of the game.

321
by chris clark (not verified) :: Thu, 01/18/2007 - 3:10pm

:316 I don't think I can be a Peter Libero on this point. Moreover, I tend to agree with Rich Conley on the idea that DVOA isn't an equally effective measure of all teams performances. However, at the same time, Pat has a good point, in that statistically, that is doing measurements across the entire league and over series of games, it is about as good as one can do, well as good as any measure we have found.

I think Rich's idea of a non-linear measure is a potential fix. Different teams have different tendancies, some play better against good teams and some better against poor teams (the old "quality wins" argument, aka playing up or down to ones opponent). Moreover, some teams play better earlier in the season, some better later. This used to be the dolphins/bills split. You also have HFA and road-warrior teams.

Now, it is questionable whether we have enough games in a season to include all those factors reliably in DVOA. However, I know that Aaron wrote back last time I asked that is was in his plans to include at least HFA in the calculation sometime. If you really think something would improve DVOA, the best way to prove it is to take the stats, crunch the numbers, and show us your results.

Personally, I think the opponent adjustment could be improved. Hopefully, I'll eventually get the chance to show how.

322
by chris clark (not verified) :: Thu, 01/18/2007 - 3:32pm

:318 I can't argue the other numbers, but when DEN/NE played, their defense wasn't -3.7. At that time, DEN's defense was highly rated and improving. It was only after game 9, that "something" happened, and the defense started going downhill. Thus, using the end of season #s for DEN, underrates them for what level they were playing at when they played NE.

In fact, that's part of my argument that one could do better opponent adjustments.
As I recall, Aaron tried iterating on the adjustments to see if they convereged and decided that a 2 or 3 attempt fit was best. However, that assumes that the numbers are fixed for a team for the entire season. If one assumed that opponent adjustment was a polynomial, one might get a better fit. I don't know for certain, I haven't crunched the numbers. However, it is my hypothesis.

Still, my point to you holds, you are assuming the DEN's defense was poor when they played NE, because it was poor at the end of the year, and that brought their whole year ranking down. I think that DEN's defense was actually quite good against NE and that's NE's offense deserves full credit for how it played in that games. (The same thing holds for last years playoff matchup. It wasn't the offense playing poorly. It was the offense getting beat.)

323
by hwc (not verified) :: Thu, 01/18/2007 - 3:58pm

It seems to me that the very nature of DVOA statistics will underrepresent the winning potential of teams with coaches like Bill Belichick and other highly strategic coaches.

Why? Because DVOA is based entirely on individual plays. By design, it ignore larger issues of game management and the whole concept of tilting the odds in your favor over the entire game.

Here's a perfect example. With a 3 point lead in Sunday's game, the Pats kicked off to the Chargers with a minute left on the clock and no timeouts. The entire defensive strategy was based on shutting down sideline patterns and preventing the deep ball. The Pats were perfectly willing to give up 10 yard pass plays over the middle. In fact, they invited the Chargers to complete 10 yard pass plays over the middle by leaving that portion of the field undefended. Why? Because with no time outs, every completed 10 yard pass over the middle brought the Pats one play closer to winning. There simply wasn't enough time to complete enough 10 yard passes over the middle to win the game outright or even get into position for a legitimate field goal opportunity to tie.

Because it is based entirely on individual plays, DVOA "docked" the Pats defense for making plays that, in reality, increased the chances of winning.

This is not a knock on DVOA. It is simply a recognition of what the system attempts to measure. As long as we understand that, then we can gain tremendous insight into the game from DVOA, especially when broken down into down and distance details, etc.

If it is less than perfect at providing a "big picture" view, so what? The Sagarin style rankings based on wins, points, game location, and opponent strength give us that side of the equation.

324
by Pat (not verified) :: Thu, 01/18/2007 - 4:05pm

It should be noted that opponent adjustments can be tested. Chart resulting DVOA measured in the matchup as a function of DVOA difference between the two units.

You could do this with points per drive, too. When a 2.0 point per drive offense faces a 1.0 point per drive defense, how many points per drive on average does it result in? Plot that as a function of difference between the two (You could also try a 2D fit, but that cuts down on the numbers).

You can't just stab in the dark and say "try this" or "I don't think it's working". You can measure and test these things. Yes, you can only test them against average behavior, but claiming "this team is an outlier" is almost always a subjective determination.

If the opponent adjustments don't work, you should be able to show it in a non-subjective way, not just argue it. I don't doubt that the opponent adjustments aren't perfect. The only question is how bad are they? And I doubt they're that far off.

In fact, that’s part of my argument that one could do better opponent adjustments.

What you're basically saying is "use the weighted DVOA at the time of the game for the opponent adjustment for that game". That's not the smartest way to do it - weighted DVOA is the best you can do at any given week. It's not the best you can do once the season's over. What you'd want is some sort of weighted measure centered on that week.

Moreover, some teams play better earlier in the season, some better later. This used to be the dolphins/bills split. You also have HFA and road-warrior teams.

There's no way - none - you'd ever be able to determine which teams are truly doing that, and for which teams it's just noise. Too little data.

325
by Brad (not verified) :: Thu, 01/18/2007 - 4:07pm

"If your team has higher expected points further back on the field, than field position becomes worth less. You can not compare across the league for the same reason you can’t say that teams should pass every down because passes get more yards than runs: you’re ignoring too many factors"

But what you are saying is that if NE has an above average offense because their expected points are higher than other teams for a given field position. What you are also saying is that if their offense was just average then they would not be able to play their give up yards not points defensive strategy. So it is the New England OFFENSE that allows thier defense to give up few points and that is what DVOA is saying.

326
by Rich Conley (not verified) :: Thu, 01/18/2007 - 4:19pm

"What you are also saying is that if their offense was just average then they would not be able to play their give up yards not points defensive strategy. So it is the New England OFFENSE that allows thier defense to give up few points and that is what DVOA is saying."

No, its not. DVOA is saying thier defense is poor because of that.

You can't consider how the Patriots Defense would be if they didnt have a good offense, because they do have a good offense.

You can't say if they didnt have a good offense theyd do anything, because they DO have a good offense. You're making an assumption on how the defense would perform if the offense wasnt good, and that assumption isnt necessarily accurate. DVOA makes the same assumption you are, and I think its a bad one.

"Why? Because DVOA is based entirely on individual plays. By design, it ignore larger issues of game management and the whole concept of tilting the odds in your favor over the entire game."

And that is exactly the problem, and why people need to stop taking these things as being all that predictive.

327
by Pat (not verified) :: Thu, 01/18/2007 - 4:40pm

No, its not. DVOA is saying thier defense is poor because of that.

First: 8th in the league, and -8%, is far from poor. You're exaggerating significantly.

Second: no, it's not. The only reason DVOA says that the Patriots defense is anything less than third in the league is opponent adjustments. Which is the same place that pace-free points-based metrics have them.

Where do you think the Patriots defense should be ranked? First? No way. Baltimore's better in just about every metric. Second? Chicago's given up more points in games, but they've done it in more drives. In points/drive - a pace-free statistic - Chicago's noticeably ahead of New England.

The points/yards thing is completely immaterial to the discussion.

328
by B (not verified) :: Thu, 01/18/2007 - 4:41pm

323: That's not entirely accurate. DVOA Takes plays and compares them to plays by other teams in similar situations. So how the Patriots graded on that particular series is compared to how other teams did protecting small leads late in the game.

329
by Pat (not verified) :: Thu, 01/18/2007 - 4:46pm

Because it is based entirely on individual plays, DVOA “docked� the Pats defense for making plays that, in reality, increased the chances of winning.

Except that those situations are compared to similar situations. And since every team knows that with only a minute left, with the lead, you can give up 10 yard passes - this isn't rocket science - you don't 'dock' the defense in those situations, because everyone gives up those plays.

330
by hwc (not verified) :: Thu, 01/18/2007 - 4:47pm

...people need to stop taking these things as being all that predictive.

We already know that DVOA is not predictive. After all, based on DVOA, the Chargers won a massive rout over the Pats on Sunday and are preparing for the AFC Championship game in San Diego. But, of course, the Chargers did NOT win on Sunday.

Why did the Chargers lose despite a lopsided DVOA game? Simple. They did not capitalize on their opportunities as efficiently as the New England Patriots did. The DVOA stuff doesn't matter much when you have three highly efficient Tom Brady drives to close the first and second halves of the game.

If I'm looking for pure predictive capability, I go to the Sagarin, Elo Chess, and Pure Points predictions because the end of the day, it's all about points and wins. I enjoy the DVOA statistics because they can provide some insight into how a team performs in certain situation.

331
by Pat (not verified) :: Thu, 01/18/2007 - 6:00pm

We already know that DVOA is not predictive. After all, based on DVOA, the Chargers won a massive rout over the Pats on Sunday and are preparing for the AFC Championship game in San Diego. But, of course, the Chargers did NOT win on Sunday.

I'm really, really tempted to write a guest article explaining exactly what the word "predictive" means.

Predictive doesn't mean you can predict the outcome of games all the time. Nothing can do that. The game itself can't do that.

Predictive, in this sense, means that DVOA can predict itself in future games. And it can. That's what the midseason projections do, for instance.

If I’m looking for pure predictive capability, I go to the Sagarin, Elo Chess, and Pure Points predictions because the end of the day, it’s all about points and wins.

Even those aren't predictive. They're like 65% predictive on average. You can get that with just picking the team with the higher Pythagorean wins. You can't predict football games any better than about 65% consistently. This just has to do with how close the teams are in true strength.

332
by VarlosZ (not verified) :: Thu, 01/18/2007 - 6:12pm

#330: "We already know that DVOA is not predictive. After all, based on DVOA, the Chargers won a massive rout over the Pats on Sunday and are preparing for the AFC Championship game in San Diego. But, of course, the Chargers did NOT win on Sunday."

First, you're confusing 'predictive' with 'descriptive.' The DVOA numbers for the NE-SD game are describing what happened, not predicting what will happen.

Second, DVOA is descriptive. For the game in question, DVOA reveals that San Diego was significantly better on a per-play basis -- the Chargers' average play, on both sides of the ball, was much more successful than the Patriots average play.

Finally, and more importantly, DVOA is predictive. That is the standard by which it is measured, and all changes to DVOA are made with an eye towards making it a more accurate forecaster of future events. The reason SD comes out so much higher than NE in DVOA is that the kind of success the Chargers had in the game is the kind that tends to be repeated from one game to the next, and also tends to lead to wins. OTOH, the kind of success that NE had (recovering a muffed punt, stripping the ball from a defender after a failed 4th Down, watching your opponent miss a game tying field goal, etc.) does not tend to carry over to future games.

Of course, those aren't the only things that NE did well, but those were the plays that won them the game in spite of being outplayed on both sides of the ball for most of the day.

I don't know how good Sagarin is at forecasting results, but I'd wager that DVOA does better -- in any event, DVOA does a very good job, say, picking games against the spread, to the point that it would be almost impossible for Sagarin to be much better.

333
by throughthelookingglass (not verified) :: Thu, 01/18/2007 - 6:19pm

"I’m really, really tempted to write a guest article explaining exactly what the word “predictive� means."

Please.

334
by Eddo (not verified) :: Thu, 01/18/2007 - 6:24pm

330:
1. Aaron has never claimed that DVOA can predict single game outcomes.
2. Measuring DVOA against a game that happened in the past is no way to judge the predictive ability of DVOA.

335
by hwc (not verified) :: Thu, 01/18/2007 - 6:24pm

The reason SD comes out so much higher than NE in DVOA is that the kind of success the Chargers had in the game is the kind that tends to be repeated from one game to the next, and also tends to lead to wins. OTOH, the kind of success that NE had (recovering a muffed punt, stripping the ball from a defender after a failed 4th Down, watching your opponent miss a game tying field goal, etc.) does not tend to carry over to future games.

I dunno.

I would say that having a special teams gunner grab the returner's arm and prevent him from picking up a muffed punt is the kind of thing that leads to future wins.

I would say that having WR's drill on ball stripping skills is the kind of thing that leads to future wins.

I would say that forcing the other team to burn their timeouts, executing good kickoff coverage to pin the other team back on a long field, and managing the defense to keep the clock running and force a 54 yard field goal effort is exactly the sort of execution under pressure that leads to future wins.

The fact that DVOA doesn't have a stat for any of that doesn't change the underlying reality.

336
by Pat (not verified) :: Thu, 01/18/2007 - 6:33pm

Well, let me clarify it a bit here, for now. :)

Suppose you've got a ranking system. You build this ranking system by creating a ranking such that r(A,B) = probability that A will beat B. You then find the rankings that maximize the probability that all the results you've seen have happened (this is called a Bradley-Terry model).

Now, team A plays team B. Team A is higher than team B, and r(A,B)=0.55. Team B then beats team A. Oh no, you say. My model isn't predictive!

No. Game outcomes aren't absolute - like I said, football can't even predict itself, so there's no guarantee that the game that was played didn't fall exactly in line with what you predicted. If the score was, say, 24-21, you could easily do a regression and realize "hey, 24-21: if those teams played again, team A (the team that lost!) has a 45% chance of winning, maybe +/-10%."

So now you've got a situation where your model predicted a 55% chance for team A to win. And the actual result of the game said that in a rematch, team A has a 45% chance (+/- 10%) of winning. 45% is within 10% of 55%.

In other words, the model was predictive. Even though the outcome of the game was wrong.

The way you determine whether or not the model was predictive is by looking at the results from all teams who were ranked A,B, and comparing to either the win percentage (i.e. do 55% of those teams win?) or the distribution of game output functions (i.e. are most of those games close?).

If all you do is just compare prediction percentages, they're all going to peg at around 65%. Predicting a game wrong doesn't mean you were wrong, because you never predict with absolute certainty.

337
by Pat (not verified) :: Thu, 01/18/2007 - 6:39pm

#335:

I would say that having a special teams gunner grab the returner’s arm and prevent him from picking up a muffed punt is the kind of thing that leads to future wins.

Why?

It led to a previous win. But what makes you think that it will lead to future wins? Will that situation ever happen again? Has it before? And what makes you think that every other team in the league doesn't have the exact same ability?

I'm not trying to criticize you - it's just that giving reasons which led to a previous win don't mean they lead to future wins.

Let me give an example: Obviously, Seattle has the ability to win the coin toss in overtime. I would think that would lead to future wins. After all, more teams who win the coin toss win overtime.

What's the difference between this statement and yours? Other than we know that coin toss calling isn't predictive, and

338
by VarlosZ (not verified) :: Thu, 01/18/2007 - 6:45pm

#335: "I would say that ... executing good kickoff coverage to pin the other team back on a long field, and managing the defense to keep the clock running and force a 54 yard field goal effort is exactly the sort of execution under pressure that leads to future wins.

The fact that DVOA doesn’t have a stat for any of that doesn’t change the underlying reality."

Those are absolutely factored into DVOA -- why would you think otherwise?

"I would say that having a special teams gunner grab the returner’s arm and prevent him from picking up a muffed punt is the kind of thing that leads to future wins.

I would say that having WR’s drill on ball stripping skills is the kind of thing that leads to future wins."

Yes and no -- mostly no. Those are skills, and good plays, and they help. However, they so rarely occur no matter how good a team is at them that, even if the skill is indeed repeatable, its impact on future games lies below the threshold of statistical detection. Moreover, as you're probably aware, fumble recovery among NFL teams is, basically, totally random; no teams show repeatable success either from one year to the next or within the same season.

Look at it this way: if this were the regular season and both NE and SD had games next week, which would be more likely to be repeated: San Diego's consistent efficiency at moving the ball and stopping the other team from doing so, or New England's recovering more than their share of fumbles, and benefitting from dumb personal fouls, and having WRs force fumbles, and so on?

Again, what San Diego did well on Sunday is the kind of thing that is much more likely to be repeated in the future than what New England did well. Hence, DVOA says they played the "better" game.

"I would say that forcing the other team to burn their timeouts. . ."

Well, you're right that DVOA doesn't account for that, and some teams are (probably) better at it than others. However, it's such a small thing on balance that its impact may be unquantifiable. At any rate, it certainly remains unquantified, as (I assume) no one has studied the matter.

339
by hwc (not verified) :: Thu, 01/18/2007 - 6:49pm

It led to a previous win. But what makes you think that it will lead to future wins?

Because effectively coaching and drilling those kinds of situational plays are exactly the sorts of small differences that can decide close games.

A team that effectively coaches and drills those things will be more likely to get "lucky" than a team that is not as effective in doing so.

The Brown strip and the gunner's attack on the punt returner's arm were clearly intentional plays, every bit as intentional as a running back making an effective backside cut. Without Brown's strip, the defender keeps possession. Without the gunner, the punt returner has the best shot at picking up the muff. He had it in his hands when the gunner ripped his arm away.

340
by hwc (not verified) :: Thu, 01/18/2007 - 6:55pm

Look at it this way: if this were the regular season and both NE and SD had games next week, which would be more likely to be repeated: San Diego’s consistent efficiency at moving the ball and stopping the other team from doing so, or New England’s recovering more than their share of fumbles, and benefitting from dumb personal fouls, and having WRs force fumbles, and so on?

Actually, having seen the rap sheets, suspensions, and behavior of the Chargers players this season and the mental state of Tomlinson and Rivers after the game, I'd bet on the stupid personal fouls being the most likely to occur the next time these two teams face off.

Does DVOA predict what happens to a team when they can't risk juicing in the offseason the following year? It would be interesting to see if both roid rage personal fouls and on field performance decline.

341
by Pat (not verified) :: Thu, 01/18/2007 - 6:56pm

Because effectively coaching and drilling those kinds of situational plays are exactly the sorts of small differences that can decide close games.

So can coin flips!

The Brown strip and the gunner’s attack on the punt returner’s arm were clearly intentional plays,

Sure. But so is a call on a coin flip. The guy definitely says "heads" or "tails."

What you need to show is 1) those results happen repeatably, and 2) other teams don't have those same skills to.

In Brown's case, for instance, why has he never stripped a ball before? He has 44 tackles. He's never forced a fumble before. What makes you think other receivers don't have the exact same skill level Brown has, and they haven't gotten the chance to show it?

Ditto for the gunner's case. Has he done it before? Does he do it more often than gunners on other teams?

342
by hwc (not verified) :: Thu, 01/18/2007 - 7:07pm

In Brown’s case, for instance, why has he never stripped a ball before?

Actually, yes. I can't remember the game now, but it was key play in a playoff game in the last two years. Brown was the receiver. The defender got the pick, and Brown stripped it from behind. In that case, the strip occurred so quickly that the pass was ruled incomplete.

Ditto for the gunner’s case. Has he done it before?

He had never played in an NFL game until October. He was signed off the practice squad.

I would say that kind of big game situational awareness from a practice squad player probably indicates coaching effectiveness that carries over to future games.

343
by VarlosZ (not verified) :: Thu, 01/18/2007 - 7:22pm

#340: "Actually, having seen the rap sheets, suspensions, and behavior of the Chargers players this season and the mental state of Tomlinson and Rivers after the game, I’d bet on the stupid personal fouls being the most likely to occur the next time these two teams face off.

Does DVOA predict what happens to a team when they can’t risk juicing in the offseason the following year? It would be interesting to see if both roid rage personal fouls and on field performance decline."

Ok, if you really believe that, fair enough. However, you have to acknowledge that those opinions are 100% specualtive, and are in fact contradicted by the available research: a team's past rate of personal fouls is not predictive of its personal foul rate in the future.

Also -- and this is very off-topic -- you should realize that by suggesting the Chargers might have lost because they're roided up thugs and criminals, you're only sounding like someone with an axe to grind who has no interest learning something new or changing his mind.

344
by Pat (not verified) :: Thu, 01/18/2007 - 7:26pm

Actually, yes. I can’t remember the game now, but it was key play in a playoff game in the last two years. Brown was the receiver. The defender got the pick, and Brown stripped it from behind. In that case, the strip occurred so quickly that the pass was ruled incomplete.

Let me repeat the previous statement, then: what makes you think Brown's any better than any other receiver at doing this? I've seen Donte Stallworth twice this year already break up an interception in almost exactly the same way.

I would say that kind of big game situational awareness from a practice squad player probably indicates coaching effectiveness that carries over to future games.

What makes you think practice squad players on other teams wouldn't've done the same thing? Cincinnati had a rookie this year who also displayed fantastic situational awareness on a punt return. Does that mean Cincinnati's well coached, too?

The problem here that I'm trying to show is that I can find examples from every team in the league and make exactly the same argument you're doing. I'm not saying you're wrong. But it's an opinion, nothing more. You can't justify it in any way.

345
by VarlosZ (not verified) :: Thu, 01/18/2007 - 7:30pm

So, Pat, you think we should work in shifts, or what? Redundancy is wasteful... :-)

346
by Pat (not verified) :: Thu, 01/18/2007 - 7:37pm

It's funny. I've just noticed recently that NFL analysts/fans tend to work in patterns.

1) Choose good organization-of-the-week (if you're a fan, skip this - choose your team).
2) Find something they do that sounds good/interesting. Maybe it helped them find a great player. Maybe they made a critical play during the week.
3) Write it up as if no/few other team in the league does this.
4) Watch as everyone insists that said "something" is a clear indicator of their brilliance.

I'm waiting for someone to write up that a team wins more in overtime because they have a "coinflip calling strategy."

I'm making a resolution right now - I will never compliment the Eagles on doing something until I can show that 1) less than half of the teams in the league do it, and 2) it's repeatable.

347
by hwc (not verified) :: Thu, 01/18/2007 - 7:40pm

what makes you think Brown’s any better than any other receiver at doing this?

I didn't say that he was better. The Patriots aren't the only team that wins games in the NFL.

348
by Pat (not verified) :: Thu, 01/18/2007 - 7:43pm

I didn’t say that he was better. The Patriots aren’t the only team that wins games in the NFL.

What makes you think even the receivers on the teams that lose aren't as good as him at that skill?

That would make the skill immaterial to winning, after all, since everyone has it.

349
by hwc (not verified) :: Thu, 01/18/2007 - 7:46pm

Also — and this is very off-topic — you should realize that by suggesting the Chargers might have lost because they’re roided up thugs and criminals, you’re only sounding like someone with an axe to grind who has no interest learning something new or changing his mind.

Not really. If I look at Sunday's game objectively, starting with the week's preparation, the game itself, and the Chargers' whining after the game, I would suggest that the fundamental reason the Chargers lost is that they are less disciplined and poised than the team that beat them.

The steroid abuse and the criminal rap sheets are relevant in that they speak directly to team discipline and mental toughness.

350
by hwc (not verified) :: Thu, 01/18/2007 - 7:57pm

That would make the skill immaterial to winning, after all, since everyone has it.

That's not really the issue. All NFL teams (well, except maybe the Raiders) are all capable of doing the same "good" things. What separates the winners from the losers is how often they do them, when they do them, and whether they can continue to do them even when injuries force them to go deep into their rosters or practice squads.

Look, I have no beef whatsoever with DVOA. I find it a useful analytical tool or at least an interesting diversion. Is it terribly relevant to this weekend's games? No. Probably not. The fact that the Pats dropped in DVOA after beating a very good team coming off a bye on the road probably has little or no predictive power regarding the AFC Championship game. Frankly, I think anyone trying to predict Sunday's Colts/Pats game is doing so for entertainment value only. Might as well be a juggling act. The game will likely hinge on unpredictable events. Tipped pass INTs. An unconscious player maintaining possession of a fumble as he lands out of bounds with his leg on the ball. A false start penalty at the wrong time. Who knows?

I would be surprised if the game goes the way anyone predicts. I don't recall anyone's system predicting the Colts winning field goal contest in Baltimore last weekend.

351
by VarlosZ (not verified) :: Thu, 01/18/2007 - 7:59pm

#349: "Not really. If I look at Sunday’s game objectively, starting with the week’s preparation, the game itself, and the Chargers’ whining after the game, I would suggest that the fundamental reason the Chargers lost is that they are less disciplined and poised than the team that beat them.

The steroid abuse and the criminal rap sheets are relevant in that they speak directly to team discipline and mental toughness."

I'll certainly grant that the Chargers made a ton of mental errors, and they probably cost them the game. By itself, however, that doesn't say anything about what the Chargers are -- it's only relevant to what they were on Sunday. It would be a different story if you could show that San Diego routinely made dumb mistakes over the course of the season, but given that had the best record in football that seems exceedingly unlikely.

You have no evidence that the off-field problems of the Chargers' players are at all relevant to their mental mistakes against New England. Anecdotally, you could probably point to a few felonious teams that made dumb mistakes on the field. I could play that game too, however: the most recect Cowboys dynasty probably had more problems with drugs, violence, and the law than any of their contemporaries off the field, but it didn't seem to stop them from being a smart and successful team on it. Just like it didn't stop the Chargers from being dominant for four months during the regualr season.

352
by hwc (not verified) :: Thu, 01/18/2007 - 8:07pm

It would be a different story if you could show that San Diego routinely made dumb mistakes over the course of the season, but given that had the best record in football that seems exceedingly unlikely.

I believe that tournament play and regular season play are qualitatively different due to the pressure of a single elimination format and the rarity of an NFL player ever getting to a Super Bowl.

It's the same qualitative difference that you see between the Greater Hartford Open and the Masters. The guy who wins in Hartford can hit all the shots. But, can he hit them on the back nine on Sunday afternoon in Augusta? Different question. Usually a different answer.

I'm sure that is heresy in the hallowed halls of DVOA statistics, but that's the way I see it.

353
by Pat (not verified) :: Thu, 01/18/2007 - 8:12pm

hwc, my comments have nothing to do with DVOA at all. It has everything to do with predictive and non-predictive events.

What separates the winners from the losers is how often they do them,

Yup. Which is why you need to show that Brown and that gunner (or New England, etc.) does those things more often than other teams. Repeatably. If you want to say "well, they do them at better times" that makes it even harder - you'd have to show that they do it more often than other teams at appropriate times.

Without showing that, there's no reason to believe that those events indicate anything about the quality of New England's team compared to the rest of the teams in the NFL.

354
by hwc (not verified) :: Thu, 01/18/2007 - 8:25pm

Without showing that, there’s no reason to believe that those events indicate anything about the quality of New England’s team compared to the rest of the teams in the NFL.

Ultimately, there's on one indication of the quality of teams in the NFL. Who hoists the Vince Lombardi trophy.

You are focusing on one specific issue as if "the number of Troy Brown strips" is predictive. It's not Troy Brown strips; it may be hundreds of other little events, any one of which can be the difference between hoisting the trophy and whining about getting jobbed by an inferior team...or the tuck rule or the sorriest cornerback in the league mocking your LB's classless celebration dance.

It is the defenders of DVOA who are implying that those particular little things have some relevance over some other little things. Maybe they do; maybe they don't.

355
by hwc (not verified) :: Thu, 01/18/2007 - 8:28pm

the quality of New England’s team compared to the rest of the teams in the NFL.

BTW, I have made no claims about the quality of the 2006 Patriots. That is yet to be determined. All we know at this point is that they are one of the two best teams in the AFC. We won't know if they are the best team in the AFC until Sunday night.

356
by VarlosZ (not verified) :: Thu, 01/18/2007 - 8:37pm

#352: "I believe that tournament play and regular season play are qualitatively different due to the pressure of a single elimination format and the rarity of an NFL player ever getting to a Super Bowl.

It’s the same qualitative difference that you see between the Greater Hartford Open and the Masters. The guy who wins in Hartford can hit all the shots. But, can he hit them on the back nine on Sunday afternoon in Augusta? Different question. Usually a different answer.

I’m sure that is heresy in the hallowed halls of DVOA statistics, but that’s the way I see it."

Again, fair enough, but your opinion is not based in fact. As far as I know there has been no study of "clutch" performance ability in football, but it's a major source of debate among baseball enthusiasts. In that sport, study after study has shown objectively that clutch hitting doesn't exist (i.e., there is no statistically detectable clutch ability), even though the subjective impression of most fans and players is that it does. I see no reason why the same wouldn't be true in football.

That's not to say that postseason games are qualitatively identical to regular season games -- the weather is colder (which has a real, quantifiable effect), teams are at their most injury-plagued, and, most importantly, there are (usually) no bad teams left.

357
by Eddo (not verified) :: Thu, 01/18/2007 - 8:42pm

355:
I have made no claims about the quality of the 2006 Patriots. That is yet to be determined. All we know at this point is that they are one of the two best teams in the AFC.
Hypothetical situation: the Bears and Patriots meet in the Super Bowl. That means that we know they are the best teams in their respective conferences. Now let's say the Super Bowl is tied with 0:04 left, and the Patriots line up to attempt a 35 yard field goal. Gostowski's kick is blocked, and Brendan Ayanbadejo returns it for a touchdown.
Would you be willing to say that you know that the Bears are the best team in the league at that point?

358
by Pat (not verified) :: Thu, 01/18/2007 - 8:47pm

Ultimately, there’s on one indication of the quality of teams in the NFL. Who hoists the Vince Lombardi trophy.

By that argument, there's one good team in the league and 31 bad ones. I mean, you're welcome to it, but "SUCK/NOT SUCK" is not a very good rating system. Not tremendously informative.

BTW, I have made no claims about the quality of the 2006 Patriots

No, but what you did say is that Brown stripping that ball is an ability that will win them future games.

There's no evidence to support that.

359
by hwc (not verified) :: Thu, 01/18/2007 - 8:53pm

but it’s a major source of debate among baseball enthusiasts. In that sport,

What is baseball?

Would you be willing to say that you know that the Bears are the best team in the league at that point?

Yes, of course. By definition, the team hoisting the Vince Lombardi trophy is the best team in football. For the Patriots to be the best team in your hypothetical, they would have to figure out a way to beat the Bears in the SuperBowl.

All this other stuff is just an entertaining diversion. DVOA (or the similar kinds of statistical analysis that NFL teams do) probably have some value in coaching staffs prioritizing their goals and objectives. For example, if I know my team is below average on 3rd down conversions, I might focus some teaching resources on 3rd down conversions (if I believe that 3rd down conversions are important).

360
by hwc (not verified) :: Thu, 01/18/2007 - 8:56pm

By that argument, there’s one good team in the league and 31 bad ones. I mean, you’re welcome to it, but “SUCK/NOT SUCK� is not a very good rating system. Not tremendously informative.

No. The four teams playing this weekend are better than the four teams they beat last weekend. And, so forth.

At the end of the day, there is still one winner and 31 losers.

361
by VarlosZ (not verified) :: Thu, 01/18/2007 - 9:00pm

#354: "Ultimately, there’s on one indication of the quality of teams in the NFL. Who hoists the Vince Lombardi trophy."

That is one measure of team quality, but it's hardly the best one.

You could just use wins and losses to determine a team's true quality, but that is a seriously flawed approach in the face of so many better options. You can judge based on 16-20 results, each with only two possible outcomes, and thus get a pretty good idea of which teams are good and bad. But why focus on such a small and limited data set? Instead, you can examine thousands of plays for each team, each with countless possible outcomes. The latter approach should be -- and in fact is -- more predictive of future outcomes, so it only stands to reason that it is also better at describing past and current team quality.

362
by VarlosZ (not verified) :: Thu, 01/18/2007 - 9:11pm

hwc, I'm afraid I don't get the impression that you're seriously considering other peoples' points of view -- you're conceeding nothing, and, when an argument is put forth for which you apparently have no rebuttal, you have tended simply to drop the subject as if it never existed. Anyway, it's frustrating.

Good night.

363
by hwc (not verified) :: Thu, 01/18/2007 - 10:32pm

That is one measure of team quality, but it’s hardly the best one.

What other measure can there be except winning the Super Bowl? Seriously.

Let me pose a couple of questions to illustrate the point:

a) Which do you remember? The stats of the 1968 Baltimore Colts? Or Willie Joe Nameth and the NY Jets?

b) Which do you remember? The stats of the 1976 Minnesota Vikings? Or Kenny Stabler, Johnn Madden, and the Oakland Raiders?

c) Which do you remember? The stats of the 1990 Buffalo Bills (widely considered to be the best offense in the history of the NFL)? Or Bill Parcells, Lawrence Taylor, and the New York Football Giants?

d) Which do you remember? The stats of the 2001 Greatest Show on Turf St. Louis Rams? Or Bill Belichick and Tom Brady?

As interesting as stats may be, Lombardi trophies are what it's all about.

364
by hwc (not verified) :: Thu, 01/18/2007 - 10:41pm

The latter approach should be — and in fact is — more predictive of future outcomes, so it only stands to reason that it is also better at describing past and current team quality.

Does weighted DVOA really do all that?

Let's see. Weighted DVOA "predicted" that the Pats were better than the Chargers last week. On Sunday, the Pats defeated the Chargers on the road. This week, weighted DVOA "predicts" that the Chargers are better than the Patriots.

Sorry, but I remain skeptical of the DVOA's predictive powers. Certainly when it comes to accessing past performace, I'm sticking with good old fashioned, "who won the game?". Just as I will this Sunday. Just as I will on SuperBowl Sunday.

If your team doesn't win, it wasn't good enough to beat the other team. The DVOA of the Super Bowl game won't mean a darn thing beyond an interesting discussion topic.

365
by Eddo (not verified) :: Thu, 01/18/2007 - 10:43pm

hwc,
Obviously the ultimate goal is to win the Super Bowl. The fact that the Chargers outplayed New England on Sunday doesn't make their loss hurt any less or mean that they achieved their goal. However, this in an analytical website. What's the point of reading any analysis if you only believe that the wins and losses matter? All you need to figure out the best teams is a listing of the scores of the previous weekend's games.
The fact is, while the goal of every team is to "find a way to win," that doesn't necessarily mean they are decidedly better than the teams they beat. By your logic, if the Patriots are a better team than the Chargers because they won, what are we to make of the 14 regular season San Diego wins against the 12 regular season New England wins?

366
by hwc (not verified) :: Thu, 01/18/2007 - 10:56pm

The fact that the Chargers outplayed New England on Sunday doesn’t make their loss hurt any less or mean that they achieved their goal.

If they had outplayed the other team, they would have won. Be that as it may, the Chargers should feel horrible. To squander the number one seed and home field advatage throughout the playoffs in a one-and-done home loss devastating. Shots like that don't come along very often.

By your logic, if the Patriots are a better team than the Chargers because they won, what are we to make of the 14 regular season San Diego wins against the 12 regular season New England wins?

The regular season exists only for two reasons for a championship contending team:

a) Position your team for the playoff run (anything but a bye seeding is a disappointing season because it makes it nearly impossible to win the SuperBowl).

b) Sixteen weeks of practice and games to build towards becoming a championship caliber team in January. By that measure, we really can't evaluate the regular season until the tournament games in January are played.

Nobody cares that the Chargers went 14-2 before losing their first playoff game. It's meaningless now. They obviously were not a championship caliber team or, fresh off a bye, they would have beaten a road team opponent.

367
by B (not verified) :: Thu, 01/18/2007 - 10:57pm

364: So the best team in 2004 was the Miami Dolphins? I think it should be rather obvious that the better team doesn't win every Sunday.

368
by hwc (not verified) :: Thu, 01/18/2007 - 11:08pm

However, this in an analytical website. What’s the point of reading any analysis if you only believe that the wins and losses matter? All you need to figure out the best teams is a listing of the scores of the previous weekend’s games.

Fair question.

For me, this analytical site is useful for four reasons:

a) It's entertaining

b) It helps me get a better handle on strength of schedule issues.

c) It helps me understand the strengths and weaknesses of potential Patriots opponents

d) It serves as a useful check and balance against the Sagarin rankings. For the most part, the two sites are in pretty close agreement once we are deep enough into the season to have meaningful data. Early in the season, all ranking systems are suspect.

Oh, and a fifth reason. Aaron and the boys have put together a really interesting statistical model and a great website. I enjoy it because there are lots of different viewpoints.

369
by hwc (not verified) :: Thu, 01/18/2007 - 11:21pm

So the best team in 2004 was the Miami Dolphins? I think it should be rather obvious that the better team doesn’t win every Sunday.

Are you talking about the final week victory of the Dolphins over the Pats? They won, so they were better that day.

Remember, I don't put any stock in the regular season beyond the fact that those games position a team for the real games in the tournament.

My own personal view is that Belichick instructed Matt Cassels to throw the ball in the stands and lose the game because that result positioned the Patriots to have best chance of getting to the SuperBowl.

Belichick had it right. Things fell into place for the Pats to host the AFC Championship game in Foxboro. But, the Pats blew it by stinking up the joint in Denver. They obviously were not a championship caliber team or they would have beaten Denver.

370
by Pat (not verified) :: Thu, 01/18/2007 - 11:41pm

It serves as a useful check and balance against the Sagarin rankings.

Sagarin's rankings really aren't that great. You'd do just as good to just say "team with higher pythagorean wins is going to win". (You'd probably do even higher to do "team with higher estimated wins wins"). Sagarin's got a slew of flaws - not the least of which is a bias towards teams which generate a high-paced offense. Sagarin uses a 1D game output function, so there's no difference between a 10-0 victory and a 40-30 victory. Sagarin's method of dealing with no (and few) loss teams is also very poor - Elo, in particular, was meant to provide scoring in a system that lacked scoring, not be on a par with a full maximum likelihood ranking.

Certainly when it comes to accessing past performace, I’m sticking with good old fashioned, “who won the game?�.

It is impossible to build a ranking system which satisfies all previous wins. The best you could do is something like beatpaths.com. Any ranking system - any one - allows for the possibility of ambiguous game outcomes.

As I've said, many, many times - I have no idea, whatsoever, why people put such a premium on "who won the game" for "who's better". It's a game. The team that wins the game won. Yay for them. It's the entire point of playing.

But being "better" - at least in the terms of position on a ranking - is something entirely different than "winning the game."

I do wonder if there's a language issue here that you're having issues with? If the "better" team always wins the game, maybe you should consider substituting "better" for "higher ranked"? It is impossible to build a ranking that does not allow for the possibility of a lower-ranked team beating a higher-ranked team. Utterly and completely impossible - because the game itself allows for that.

The statement that weighted DVOA's "predictive" power is diminished by a result where a lower ranked team beat a higher ranked team completely misses the point. See post #336. You can't predict NFL outcomes better than 65%. In the playoffs, you probably can't predict them better than 55%. That's just the average separation in strength of the teams.

If you think that predicting the outcome of games (won/loss) is the mark of a successful ranking system, I'm sorry - you're wrong.

371
by hwc (not verified) :: Fri, 01/19/2007 - 12:23am

I do wonder if there’s a language issue here that you’re having issues with?

Yeah. I'm definitely having a language issue. I'll try again.

a) The only games that really matter are the tournament games. The regular season is only interesting in so far as it positions teams for the tournament.

b) There are only four really good (successful, whatever word you want to use) teams in the NFL each year: the four teams playing in the championship games this weekend.

c) There is only one best (successful, winning, memorable, whatever word you want to use) team in the NFL each year: the team hoisting the Vince Lombardi trophy. All the others are losers, footnotes, sideshows, whatever you want to use.

d) I don't think it's possible to predict football games between roughly comparable teams. What better proof of that do any of us need than last week's Colts/Ravens matchup. No statistical model could have anticipated the way that game played out.

e) Beyond pure entertainment, I find the ranking systems useful to provide a broad brush sense of teams I don't watch regularly. I don't need to know if team is ranked #3 or #4 to have an idea that the team is probably a contender. But, if a team is ranked number #28 here and #30 in Sagarin, it's pretty good bet that they suck. I'm not looking for razor sharp granularity. I'm looking to see which teams are contenders, which teams are your standard-issue NFL midpackers, and which teams are truly wretched. A ranking system that placed teams in three tiers (plus the NFC!) would be fine with me.

f) I don't need rating systems to evaluate how "my" team is doing. When you watch every game for years, you know just by watching. My non-statistical evaluation of the Pats chances this weekend? They have a shot. Whichever team plays the best football for 60 minutes will win and it's impossible to know which team that will be. What has happened in the previous 18 games this season probably doesn't mean "diddly poo" one way or the other. The Sagarin ratings don't mean "diddly poo". The weighted DVOA doesn't mean "diddly poo". Anyone who thinks he can predict this game is kidding himself beyond the fact that you have to give a little edge to the home team.

372
by Peter Libero (not verified) :: Fri, 01/19/2007 - 1:36am

371: You really believe that the team that plays better always wins? I mean... really? There's no such thing as luck? I take it for granted that the best team doesn't always win the Superbowl. They did... which means they are the team that is rewarded, lauded, and remembered, but that doesn't mean that they were necessarily the best.

As long as football plays different schedules and plays a tournament, the best team will not always win. That's really why we LIKE tournaments... because the number of games is small, so anybody can get on a hot streak (or get lucky) and win the games that we arbitrarily declare to be more important than others. Really the most logical system is the European soccer leagues, where the team with the best regular season record (everyone in the league plays each other twice) wins the title, and the cup is just kind of for fun and so more than one team can claim to have won something (plus it makes money).

373
by PantsB (not verified) :: Fri, 01/19/2007 - 2:02am

"371: You really believe that the team that plays better always wins? I mean… really? There’s no such thing as luck?"
Two things:
The best and most simple definition of "the team that plays better" is "the team that won." OK, yes sometimes officiating changes things - but not often. Occasionally a ball bounce determines the game - but not often. The Patriots game in San Diego was not won on luck, but on execution - even the 'luckiest' play resulted from a smart play by Troy Brown and a less than smart play by the SD safety (McCree I believe his name is). Is that luck? Only in the "Luck is what happens when preparation meets opportunity" way.

That game is a large part of the argument against the weight of yards as a defense measure. SD out gained New England. Thats not even that unusual. San Diego also had better field position than New England on top of their out gaining them. New England outscored San Diego (all on offense, none on defense or returns) and thats the only thing that matters in the end.

If this was an outlier that would be one thing.... Except its not. Its routine in Patriots games.

374
by hwc (not verified) :: Fri, 01/19/2007 - 2:26am

You really believe that the team that plays better always wins? I mean… really?

In the tournament, yes.

I understand your point. But, what meaning does "better" have without the win? I mean, sure...on paper, the St. Louis Rams were a "better" team than the Pats going into the 2001 SuperBowl. Belichick's comment to Ernie Adams after the game was "Can you believe we beat that team with these players?" Belichick called that game his favorite SuperBowl (he's now coached in six with five wins). So I guess if the Rams want to take some small solace in being the "better" team, good for them. But, at the end of the day, the Patriots had the Vince Lombardi trophy and the Rams were the losers.

If you look at it with no preconceived notions, the Pats beat the two best teams (on paper) in the AFC and the best team in the NFL (on paper) in three consecutive tournament games. So maybe the "paper" was wrong when you get right down to it. Only one team could get the job done. They are obviously better.

That’s really why we LIKE tournaments… because the number of games is small, so anybody can get on a hot streak (or get lucky) and win the games that we arbitrarily declare to be more important than others.

I don't think there's anything arbitrary at all. I see the NFL season as a pyramid. You start out in August with 32 teams and 16 weeks of relatively meaningless games as the teams jockey for position and attempt to improve to championship caliber teams (there's no such thing as a championship caliber team in August). At various stages along the way, teams are jettisoned.

The tournament games are played at an entirely different level. Winning tournament games is a seriously impressive accomplishment. Underdog, favorite, 1 point, 20 points, doesn't matter. These games are not like a six-footer for par in the Greater Hartford Open. These are Sunday afternoon putts at Augusta. Playing for immortality in the record books. Really stop and think about just how hard it is to get to the SuperBowl, even if you are a perennial playoff team.

At the top of the pyramid, you end up with the single best team in each division battling it out in the SuperBowl -- a game that is played with such tight collars the players can barely breathe. Ask Donovan McPuke about the pressure.

The tournament is when you find out who is really the better team. You'll never see me whining about the Pats being the "better" team after a tournament loss. I think that's the biggest pile of crap from a team or fans. If the team can't find a way to win, they obviously weren't better.

375
by hwc (not verified) :: Fri, 01/19/2007 - 2:41am

If this was an outlier that would be one thing….

It wasn't even a fluky game. The Pats put 24 points on the board and held the leagues highest scoring offense to, what, 10 points below their season long scoring average? That usually produces a win.

The only big surprise to me was that the Pats were able to win what became an all pass/all the time come from behind scoring game. I think the Pats are strong in a lot of areas, but that's the one facet that I didn't think this year's team could pull off. I didn't think the Pats could fall behind on the road to the Chargers and mount a comeback.

To their credit, San Diego effectively forced New England to try to beat them with their "worst" stuff...and they did. San Diego's gameplan was to take away the Pats ball control balanced offense. The Pats were never able to force them out of their base run-stopping, screen stopping defense. San Diego forced the Pats to go to a downfield passing game. Having watched the Pats this year, that was certainly not the game I wanted to see. It's not been the team's strength.

376
by B (not verified) :: Fri, 01/19/2007 - 12:01pm

369: You're thinking of the wrong year. I meant when the Patriots lost to the Dolphins in week 15 before going on to win the Superbowl. That was a game the Patriots needed in order to have a chance at the #1 seed.

377
by Pat (not verified) :: Fri, 01/19/2007 - 12:28pm

There are only four really good (successful, whatever word you want to use) teams in the NFL each year: the four teams playing in the championship games this weekend.

That's a stupid cutoff. The Colts have lost only to the Super Bowl winners for the past three years. In all likelihood, they will again this year (or win it all). Even though they only made the championship once, you can easily make a case (by your logic) that they're the second-best team in the league.

Anyone who thinks he can predict this game is kidding himself beyond the fact that you have to give a little edge to the home team.

I can't agree. Plain and simple. It's right in the FO FAQ, right above - you can explain a large portion of the variation seen in the NFL with statistics. Heck, the correlation coefficient in the preseason DVOA predictions are incredibly high for a game, and the necessary uncertainty due to lack of knowledge.

Predicting the outcome of games, though? Well, you can do significantly better than home field. Home field is what, 56%? You can predict 65% of the outcomes of games.

You can't do better than that because the game output function isn't a step function - it's a rather wide logistic curve. It has nothing to do with an "inability to predict games". It has everything to do with the fact that it's a game.

I don’t think it’s possible to predict football games between roughly comparable teams. What better proof of that do any of us need than last week’s Colts/Ravens matchup.

Cherrypicking. Just because one example appears unpredictable doesn't mean all are. The Saints/Eagles played out exactly as the first meeting. The Giants/Eagles game played out exactly as their last meeting.

More importantly, if you can predict that two teams are comparable in strength, then getting an unpredictable game is perfectly fine. It means your ranking system predicted the outcome of the game as unpredictable!

Saying "I can't predict this game" is a prediction in and of itself.

378
by Pat (not verified) :: Fri, 01/19/2007 - 12:28pm

Occasionally a ball bounce determines the game - but not often.

Had the muffed punt return bounced out of bounds, or the fumbled interception bounced out of bounds, the game would've been different.

Do you know why a football is shaped the way it is? There are two reasons. First: it was easier to hold to throw. Second: the players liked the fact that it bounced unpredictably. Fumbles were intended to be unpredictable.

379
by Pat (not verified) :: Fri, 01/19/2007 - 12:36pm

Actually, regarding "you can't predict games", let me state it in a clearer sense.

When I say that you can predict NFL games at about 65% or so consistently, that sounds... not so good. I mean, home field is 56%, so that's only a 9% improvement. Doesn't seem so good.

But you can calculate from the rankings how accurate you think you should have been. And the answer is... around 65-70%.

380
by Rich Conley (not verified) :: Fri, 01/19/2007 - 12:36pm

351

"It would be a different story if you could show that San Diego routinely made dumb mistakes over the course of the season, but given that had the best record in football that seems exceedingly unlikely."

Again, the regular season and playoff football are different. Generally, when a team like sandiego plays a regular season game, they are vastly superior to the opponent. In the playoffs, thats not necessarily the case anymore.

I'd say we've got PLENTY of evidense that Marty Shottenheimer teams will continue to lose playoff games to inferior opponents because of lack of discipline.

381
by Rich Conley (not verified) :: Fri, 01/19/2007 - 12:45pm

"Obviously the ultimate goal is to win the Super Bowl. The fact that the Chargers outplayed New England on Sunday doesn’t make their loss hurt any less or mean that they achieved their goal"

DVOA says the Chargers outplayed New England because the things the Chargers did generally lead to scoring point. But thats the thing, the chargers DIDNT score points. It doesnt matter that the Chargers moved the ball well, becaues they couldnt move it once they got to the Pats 40.

This wasnt a game where two flukey INT runbacks let an inferior opponent beat a superior one. This was a game where a team outplayed the other one in the first half, and couldnt score. If you can't score, it doesnt matter how well you move the ball.

Thats the problem here, the Chargers moved the ball really well between their endzone and the Pats 40, and then just couldnt move the ball anymore. That doesnt win games. DVOA seems to think it does.

382
by Rich Conley (not verified) :: Fri, 01/19/2007 - 12:51pm

378

"Had the muffed punt return bounced out of bounds, or the fumbled interception bounced out of bounds, the game would’ve been different."

Give up Pat. Yeah, thats true, but essentially impossible. The ball would have had to have bounds 25 yards in both cases to go out of bounds.

383
by B (not verified) :: Fri, 01/19/2007 - 12:59pm

"Thats the problem here, the Chargers moved the ball really well between their endzone and the Pats 40, and then just couldn't move the ball anymore. That doesn't win games. DVOA seems to think it does."
The Chargers scores 24 points, though. So, they weren't having any trouble moving the ball beyond that point.
Anyways, let's say we have a team who starts every possession at their 20 yard line, and they consistently make it to the opponents' 40 and punt. They have good special teams, so the other team is starting at their 10 yard line, but they march down the field and score a field goal on each possession. DVOA would give the offense a good score and the defense a bad one, but I suppose you wouldn't see it that way.

384
by PantsB (not verified) :: Fri, 01/19/2007 - 1:32pm

"Had the muffed punt return bounced out of bounds, or the fumbled interception bounced out of bounds, the game would’ve been different."
Or if Brady had hit Watson when he was wide open, or if LT hadn't made Hawkins miss on his long run or ....

For one, neither of those were remotely possible. Both fumbles were 15+ yards from the sideline. For another, if we take that tract all football is luck. There is ALWAYS a couple of plays that teams can say if we have had done x we would have won. Thats execution. Thats what counts.

385
by PantsB (not verified) :: Fri, 01/19/2007 - 1:37pm

The Chargers scores 24 points, though. So, they weren’t having any trouble moving the ball beyond that point.
Anyways, let’s say we have a team who starts every possession at their 20 yard line, and they consistently make it to the opponents’ 40 and punt. They have good special teams, so the other team is starting at their 10 yard line, but they march down the field and score a field goal on each possession. DVOA would give the offense a good score and the defense a bad one, but I suppose you wouldn’t see it that way.
----
The Chargers had more yards, and better field position than the Patriots. They scored 21 points, or approximately 10 points below their season average. They had a better DVOA in both offense and defense - but the Patriots scored 24 points. Your example is flawed because while the Chargers did score on three drives, they didn't kick any FG - it was feast or famine. New England scored more with worse field position.

If a team wins, and the metric had them losing as bad as DVOA has them.....

386
by B (not verified) :: Fri, 01/19/2007 - 1:46pm

"If a team wins, and the metric had them losing as bad as DVOA has them….."
They were fortunate to recover all 5 fumbles. (Or 5 of 7 if you count the two other muffed punts), have the other team attempt a 4th and 11 conversion in field-goal range, and have the other team waste their timeouts in the second half of a close game. The Patriots capitalized on the other team's stupidity, but the were lucky that the other team was stupid.

387
by chris clark (not verified) :: Fri, 01/19/2007 - 2:44pm

: hwc "only wins count"

While I agree with you at one level, winning the SB is the goal (and everything else is secondary or tertiary--here tertiary meaning doesn't really count). It isn't true that we remember only the SB winners. I recall the Bengals from the year in the 80's when they got whalloped by the 49ers, and I don't remember which QB played for SF that year, and I'm not a CIN fan. I also remember many plays for individual regular season games, and couldn't recite the complete list of SB winners even if it was worth money for me to do so.

So, in that sense there isn't just one winner each year. In fact, I will remember this years Chargers as better than this years Patriots, unless something happens to make me change that position. Just like, I don't remember last years Steelers as the "best team" that year (I will remember the Colts as that (and I'm not a Colts fan either)), although I do remember them as an amazing team that accomplished unheard of feats on the way to their deserved SB victory.

That's because "best" is a judgment term. We each decide what we mean by best and apply that judgment to the facts we know.
DVOA doesn't measure best, except as it codifies a set of measurements that may reflect someone's judgment of best. Neither does winning the SB equate to being best, unless that is the way someone judges best. So, you can say "for you" that best is exactly equivalent to winning the SB, that is your judgment. However, it is not an absolute term, and it may not be someone else's judgment of best.

Curiously, Pat and I had this same argument (well not identical) over the Polian's S--t article. I one uses W/L records as the point of judgment, then substituting points/yards/DVOA/anything else isn't the same, because those aren't the same, even though they may appear to be similar to someone who has a different judgment system. So, at this level, your point stands, because that's the judgment you want to make. However, it can look irrational to someone who has a different judgment system--and many readers of FO definitely do not take W/L records as their metric for best.

388
by hwc (not verified) :: Fri, 01/19/2007 - 3:37pm

Saying “I can’t predict this game� is a prediction in and of itself.

Fair enough. BTW, don't you think the 65% prediction rate derives from the luxury of being able to predict a lot of mismatches during the regular season? San Diego versus Oakland, etc.

In other words, I agree that DVOA or PURE POINTS can predict those games efficiently. I guess the point I was making is that it is then a mistake to try to attribute more granularity to the rankings than is warranted.

For example, having confidence that the ranking system could truly differentiate between Indy and Baltimore in last week's game. #1, #2, #4. Whatever. Those games are probably too close to call considering the intangibles that no statistical model can capture from a limited dataset of 16 games and an inherently variable game like football.

BTW, the more I think about it, DVOA should, in reality, be a more sophisticated PURE POINTS model, correct? As I understand it, everything selected for inclusion in DVOA was to achieve the goal of correlating with points scored and allowed. What DVOA gives up in correlation with points, it gains with more sophisticated modelling of opponent strength and so forth.

One other point to address the Miami question regarding the Pats. I view the specific Miami/Patriots matchup as close to pick-em each year or at least much closer than any rankings system would suggest. Why? Specific matchups. The specific defense that Miami uses creates big problems for the Pats offense. So folk looking at rankings might chuckle at the thought of Miami beating the Pats, I am never surprised. Many of the Pats wins over the Dolphins, even during this recent run, have looked about like the Chargers game...or worse.

389
by Rich Conley (not verified) :: Fri, 01/19/2007 - 3:41pm

383
"Anyways, let’s say we have a team who starts every possession at their 20 yard line, and they consistently make it to the opponents’ 40 and punt. They have good special teams, so the other team is starting at their 10 yard line, but they march down the field and score a field goal on each possession. DVOA would give the offense a good score and the defense a bad one, but I suppose you wouldn’t see it that way."

And they'd lose. And they would consistently lose. Moving the ball is only important if it leads to points.

390
by Peter Libero (not verified) :: Fri, 01/19/2007 - 3:43pm

387: You're right, obviously... "best" is a vague term that we can define as we choose. hwc defines it purely as the team that wins the Superbowl, while Pants sees it as the leader in points scored - points against (although I'm betting he'll correct me and say the Superbowl as well).

I think those views assume that there is no variance or randomness in football, or if there is it is always for a reason. That is, the Colts didn't just have a bad game, because all good teams sometimes do not play up to potential, but they "choked" under the pressure (like at Augusta, is the favored analogy) or were poorly coached to "execute" the "little things." Honestly, it is hard for me to view that as logical, because it denies the basic reality of the fact that they are playing a game that has an unpredictable outcome. It is also impossible to disprove the notion that the Superbowl is the ultimate judge of goodness, because I would require a 7 (or 1000) game series or something to prove my point.

In essence, complaining about DVOA because the stats come out lopsided in the direction of the loser (and thus using "wins" as your ONLY indication of team quality) is declaring that football is a perfect exercise, which is patently absurd. That is what Pat means when he repeatedly says that "the game" is not that predictive; it is an imperfect exercise, where playing the game best does not always mean you win. I am truly not sure how this point can be debatable, given the complexity and randomness inherent to the game.

What more can we say for any metric but that it means the things the players did lead to points in the long run? The things the Pats did on Sunday do not lead to points, they lead to losing, which they almost did.

391
by Rich Conley (not verified) :: Fri, 01/19/2007 - 3:47pm

386

"They were fortunate to recover all 5 fumbles. (Or 5 of 7 if you count the two other muffed punts),"

So they recoverd 5 of 7, and one of them actually hurt them (the strip sack of Rivers on 4th down, where the ball went forward), and one of the fumbles they "recovered" was a fumble out of bounds that was NOT going to be recovered by anyone. If you're going to argue that fumbles are not random, but fumble bounces are, you have to accept that ball was going out of bounds, as it never actually touched the ground.

So basically, they recovered 3 of 5 actual fumbles. Thats by no means out of the ordinary.

392
by Peter Libero (not verified) :: Fri, 01/19/2007 - 3:53pm

I forgot to address 374:
I see your point, and I think that's how most people view the NFL season and the justification we give to it. The trouble with football (unlike basketball or baseball) is that you only CAN play so many games, so the tyranny of small numbers is inevitable. That is why I say it is arbitrary that we value the tournament; it is! There is no schedule adjustment, and everyone plays a different schedule. The number of games is TINY compared to what we would call "significant," so good teams can be excluded for poor luck. Even once they get to the tournament, the number of games is even smaller, meaning that victory has even LESS significance to how good the team that wins is. The reality is that the NFL does a very poor job of discovering the "best" team that season, for the simple reason that it cannot stage enough contests to figure that out. So it just has a knock-out system to figure out who can put on a short, lucky streak to win, and crowns that team champion.

Pat is also completely correct to point out that it is wrong to assume that teams that make it farther are better than those who don't; in general that is the case, but the team that gets knocked out in the wildcard round by the eventual superbowl winner could almost equally claim to be the #2 team as the Superbowl loser.

Oh, and I also dispute that superbowls are the only thing that's remembered... I suppose no one remembers the late 80's/early 90's Bills? I think even if Peyton never wins, we'll remember his Colts.

393
by Rich Conley (not verified) :: Fri, 01/19/2007 - 3:54pm

"What more can we say for any metric but that it means the things the players did lead to points in the long run? The things the Pats did on Sunday do not lead to points, they lead to losing, which they almost did."

They way they played leads to losing. The things they did lead to winning.

The Patriots entire gameplan in every game, theyre whole philosophy, is to maximize chances at winning, and minimize chances of losing, no matter how the game play actually goes. The best way to do that is to minimize an opponents points when they get in the maroon/red zone, and maximize your points in the maroon/red zone. Thats what the patriots do.

"where playing the game best does not always mean you win."

Correct, playing the game best does not always lead to winning. The thing is, there are certain strategical game plans that maximize your chance of winning when you play the game best, and certain strategies that maximize your chances of winning when you get outplayed. DVOA completely ignores that.

Preventing long plays at the sacrafise of short yardage maximizes your chance to prevent opponent scoring.

394
by chris clark (not verified) :: Fri, 01/19/2007 - 5:51pm

: 393

Yes, there is some truth to your assertion that certain strategies may lower DVOA (or VOA) but actually lead to winning games (for certain teams in certain situations). DVOA (and its kin) being a statistical measure need to emphasize teams that work in the aggregate case. If your average team would lose by playing that strategy, then DVOA needs to rate those plays worse to get a better overall correlation. (That's not quite completely true, but it is close enough for this argument.) However, if you have an outlier team, then perhaps said strategy could actually be better. There is a wonder example where the probability of a person being bald is low and probability of having a beard is low, but the probability of being both bald and with a beard is high, if you have either. Statistics can be counter intuitive that way.

The problem with that argument is that it's hard to "prove" that given team is truly an outlier. One of the facets of human intelligence is that we see patterns (even patterns that aren't there, but are artifacts of the data we are looking at). The human mind is also particularly sharp at picking out exceptions--the suggestion is that it is/was evolutionarily beneficial to be able to do so, especially in the savanah environment where a lion might be lurking in that grass. Thus, we can tend to see teams as having qualities that seem to go against the norm (and for making up new norms that support that as a general pattern). A lot of statistical analysis is designed to compensate for those tendancies so that one doesn't jump to unsound conclusions not supported by sufficient evidence. That means that FO will always have to deal with "fumbles aren't random" type arguments, because those are clearly intuitively correct to some people--that is that people will actually see and see clearly those connections. However, it doesn't necessarily make the arguments right (or wrong). One can examine the data dispassionately and then determine whether the evidence suggests the conclusion or not (and that will be a judgment), and thus two people can come to different conclusions.

Of course, with something passionate like football, it isn't clear how one gets to the dispassionate evidence sifting, unless perhaps one is BB.

395
by B (not verified) :: Fri, 01/19/2007 - 6:08pm

The strategy of giving up yards to prevent long scoring plays makes it harder for the other offense to score, but it also makes it harder for your offense to score. It's not a bad strategy. In fact, both teams in the AFC title game use the strategy. However, since the offense is pulling more weight, we should recognize their increased performance, which is why the Patriots Offense is rated higher than casual observations might put them, just as their defense is rated lower.

396
by PantsB (not verified) :: Fri, 01/19/2007 - 9:03pm

That is what Pat means when he repeatedly says that “the game� is not that predictive; it is an imperfect exercise, where playing the game best does not always mean you win. I am truly not sure how this point can be debatable, given the complexity and randomness inherent to the game.
------
This presumes a more valid metric for "playing the game best" than "winning." That is not the case. What happens in the game is irrelevant except for the ultimate outcome. As a certain coach would say "You play to win the game!" You don't play to play well, or pretty or execute a complex drive with elegance and wit.

NFL is a fairly predictable game. Pythagorean wins went 6-2 in the playoffs so far (although New England was only very marginally better than SD if you didn't include the NYJ game), missing Indy-Balt and Dal-Sea (a game in which a minor point of execution - not luck - turned the game).

Marty Schottenheimer routinely loses games his team "should" have won. The opposite is true of Bill Belichick teams. Its not a coincidence. Occam's Razor correctly states we should not invent any entities unnecessary to explain a facet of nature. Its possible that New England is just incredibly lucky in those situations and Marty is incredibly unlucky. A more probable explanation is that the Patriots know how to win better than metric that say the other team played "better."

397
by PantsB (not verified) :: Fri, 01/19/2007 - 9:14pm

Furthermore, re: the Pat's Defensive strategy makes it more difficult for their offense to score.

If this is the case, DVOA should rate the Patriots higher than their actual results in terms of points scored. It does not. Every yearly Conference Championship level New England offensive DVOA rating (01, 03, 04, 06) has aligned itself within one place of New England's points per game rating (and when its not exact, DVOA rates them 1 spot lower).

398
by Peter Libero (not verified) :: Fri, 01/19/2007 - 9:36pm

Yes... it DOES presume that, because playing well and winning are not absolutely linked. That's EXACTLY what I said in my post, that you assume that the game is a flawless way to judge the better team.

I suppose you could look at it that way; it's one of the fundamental bases for Football Outsiders that there is a better way to measure "goodness" than wins, which is a philosophical decision. That's not to say there isn't reason to think so, however; for instance, wins do a poor job of predicting future wins, which suggests (but not proves) that the best team doesn't always win.

I will say, however, that football is NOT a very predictable game; if it were, Vegas would have to take down the moneylines. The game is VERY difficult to predict. Even if we accept the idea that the best team always wins, we have to lean heavily on variance if we're to be logical at all about surprising game results, and the entire point of variance is that it's variable, so we won't know how well each team will play.

399
by Peter Libero (not verified) :: Fri, 01/19/2007 - 10:00pm

OK, here's a quick point that I forgot to make earlier... when it was claimed (for some reason) that the Patriots' improvement on defense in the red zone was disproportionate to other teams, so much that it invalidates the DVOA system.

The Pats went from the 8th overall defense to 3rd in the red zone, a net improvement of 27%. The Bears go from 4th to 2nd, improving 29.7% The Cardinals go from 23rd to 4th, improving 20%. The Saints go from 19th to 6th, improving 19%. Even the Redskins go from 32nd to 15th, improving 15%.

So the Pats don't even have the largest raw number improvement, and other teams make absolutely absurd ranking diferences. The Cardinals and Redskins are particularly hilarious for the claim that this jump is a big deal for your chance of winning.

Just as a quick counter-example, the Bears go from 2nd to 12th, a 19% drop, and they're favorites to make the Superbowl. San Diego was 14th overall, and actually 32nd in red zone, a 43% drop, and had the best record in the NFL and had to have a pretty crazy game to lose in the playoffs.

400
by Peter Libero (not verified) :: Fri, 01/19/2007 - 10:03pm

Sorry, sorry, it was Minnesota, not the Bears, who went from 4th to 2nd and had the greatest overall improvement on defense. The Bears are still the counter-example.

401
by PantsB (not verified) :: Fri, 01/19/2007 - 10:16pm

Yes… it DOES presume that, because playing well and winning are not absolutely linked. That’s EXACTLY what I said in my post, that you assume that the game is a flawless way to judge the better team.
------
The only way one can reasonably argue that the best team won't always win is because of variance and sample size. However, part of this argument is based on the fact that not only is DVOA rating New England's defense lower than their results (in points, turnovers and even resultant field position) not only this year but in several previous years. DVOA is very complex compared to most statistical measures in sports but fails to predict the estimated number of wins better than a simple application of the Pythagorean theorem with points for and against. Weighted DVOA did no better than the Pythagorean wins based predictions so far this year in the postseason (both 6-2) and last year Pythagorean wins were a better predictor IIRC.

Clearly DVOA has validity, but just as clearly it is not as strong as it could be. An area that seems to call for improvement is the defense DVOA that value one style of D over another.

BTW Money lines are essentially automatically profitable. By building the vig into a payout and adjusting as time goes by (while ensuring that you don't get let both sides win like moving the line can conceivably do) a bookie/casino can manipulate the payouts to ensure they'll get their ~10%. The difference between payouts is not set and thus can be more easily manipulated to essentially guarantee human behavior will get a payout on each side less than the bets on the other. There are many people who make a living balancing the money lines between Boston and NYC and I'm sure the same can be said between most Northeastern major cities.(However, money lines are potentially the most profitable thing for a wagerer to bet on. Much of the money in it is from people who get greedy and convince themselves the team getting +240 odds is going to win because they want that payout. Also, small lines are often inconsequential so betting the money line instead of a +2 can be smart.) Baseball - now thats a game thats tough to be a bookie on. Football is extremely easy by comparison.

402
by PantsB (not verified) :: Fri, 01/19/2007 - 10:30pm

OK, here’s a quick point that I forgot to make earlier… when it was claimed (for some reason) that the Patriots’ improvement on defense in the red zone was disproportionate to other teams, so much that it invalidates the DVOA system.
---
The claim is between the number of points allowed and the DVOA. New England's points allowed on D was statistically indistinguishable from Baltimore and was 2nd in Interceptions. The DVOA ratings have New England 8th. The DVOA has some ridiculous results (the aforementioned GB is the big one) and seems to clearly correlate with yards rather than points/turnovers as defensive metric. Since the former is vastly inferior to the latter in terms of translating to wins.... Field position does not provide a suitable excuse, nor does hindering the offense in the case of New England because the results have not born out and the DVOA offense ratings indicate that New England had a completely normal level of difficulty in producing those results. Sample size does not work either since this has occurred over several seasons.

403
by PantsB (not verified) :: Fri, 01/19/2007 - 10:39pm

BTW - what I have and others are pointing out in now way violates the underlying assumptions of DVOA. Red Zone plays are now valued 20% more than other plays presumably because its realized that scoring or preventing a score is the underlying "success" that each play is striving toward. I believe FO even found that defense in your own 10-40 through part of the 03 season but never followed up and as far as I can tell never adjusted DVOA accordingly. This is the portion of the field I'd say New England is strongest in (although I don't have stats to back that up).

404
by Peter Libero (not verified) :: Fri, 01/19/2007 - 11:39pm

401: I'm going to give you the benefit of the doubt and assume you didn't read Pat's posts when he proved how ridiculous that statement is. The Pats allow the 3rd fewest points/drive, and are 3rd in VOA. They played a crappy schedule, however, so the opponent adjustments move them down to 8th in DVOA.

Also, here are the quotes: "it still seems to me that the fact that the NE D improves so disproportionately in the red zone compared to other D’s deserves a more specific explanation than just “the short field masks their deficiencies�."

"Notice the teams that have great Red Zone DVOA seem to be really good teams?"

"The best way to do that [win] is to minimize an opponents points when they get in the maroon/red zone"

None of them are from you, but I didn't say they were. I think it's pretty clear that red zone defense is important, and correlates with wins/points allowed... but not better than red zone defense AND the defense on the rest of the field, which is what DVOA considers. DVOA also weights the areas according to the best correlation with wins, so basically it's just wrong to claim that it doesn't consider it enough. If counting them more led to a better correlation with wins, Aaron would do it.

"BTW - what I have and others are pointing out in now way violates the underlying assumptions of DVOA."
That is true, as far as it relates to the red zone defense/points allowed thing. But the discussion that was also had about wins being the only measure of a team IS counter to DVOA, and betrays a different philosophy from FO.

Also, at the top of the page there's a link to a spreadsheet showing team quality (in DVOA) at different parts of the field, so you can make your point that way. I think you forgot a word or two in your third sentence, so I don't know what you were trying to say.

Your post in 400 straw-mans DVOA; it isn't just yards. It considers a variety of factors, not just yards. Maybe DVOA is closer to yards than points, but that doesn't make it less accurate, and you have no basis for saying so. Again, how high do you expect the Pats to be rated? Do you just not like opponent adjustments?

405
by hwc (not verified) :: Sat, 01/20/2007 - 12:27am

They played a crappy schedule, however, so the opponent adjustments move them down to 8th in DVOA.

That is at odds with the Sagarin schedule difficulties. The ELO Chess and PURE POINTS systems both had the AFC East teams playing the most difficult schedules in the NFL this season.

Does DVOA go back and adjust opponent difficulty as the season progresses and more data comes in? Or is it taking a snapshot at the time of the game?

If it takes a simple snapshot, that could be problematic. There are many examples of teams that look "good" early in the season and end up being obviously "bad" teams as the data accumulates.

406
by Pat (not verified) :: Sat, 01/20/2007 - 12:49am

That is at odds with the Sagarin schedule difficulties.

This would matter if Sagarin actually measured offenses rather than total teams.

New England played crappy offenses. By any measure.

407
by Pat (not verified) :: Sat, 01/20/2007 - 1:05am

This presumes a more valid metric for “playing the game best� than “winning.� That is not the case.

You're crossing wordings here: suppose team A plays team B. Team A wins. When you say "team A played better than team B", I think you mean "for that game, in that instance." What ranking systems mean is by "better" is "does team A have a true strength exceeding that of team B"?

True strength is kind of the Platonic ideal of a team. Rankings are estimates of a true strength.

You can actually show (and it has been shown) that football (and most sports) behaves like a game with teams distributed in true strength, where the outcome of the game is determined purely by a Bernoulli trial with probability given by a game output function of the input true strengths of the two teams. This is a very important fact - it means that football does not act like a sport where matchups have significant importance. Yes. Believe it or not, there's no evidence (and you would be able to see it) that Bill Belichick can defeat Peyton Manning.

If you want to stick to your definition, feel free. But it's somewhat pointless. If the team that plays best always wins, why even have the phrase "play best"? Why not just say that they won?

408
by Pat (not verified) :: Sat, 01/20/2007 - 1:08am

For one, neither of those were remotely possible. Both fumbles were 15+ yards from the sideline.

Of course. You're right. I've never seen a fumble move 15 yards from its original point. I mean, then you might have something like a wide receiver catching a ball short of the end zone, and the ball fumbling into the end zone only to be recovered by the original team for a touchdown!

Man, I've never seen something like that.

409
by Pat (not verified) :: Sat, 01/20/2007 - 1:21am

Preventing long plays at the sacrafise of short yardage maximizes your chance to prevent opponent scoring.

Unless you can do both. In which case you're awesome, and live in Baltimore or Chicago.

Honestly, the whole "DVOA penalizes".. blah stuff is really silly. There's no evidence that DVOA penalizes New England's defense at all because of its "strategic decisions". If you really believe there's evidence for it, email Aaron. But no one's given any evidence for it, just anecdotal guesses, or arguments that are explained by one of these three categories:

1) DVOA is pace-free, points per game is not. New England's games have been shorter than basically everyone above them (and a whole lot shorter than Chicago's).
2) DVOA is opponent adjusted, points per game is not.
3) The Patriots defense has had very good starting field position.

Feel free to argue with any one of those three points, but none of those points have anything to do with DVOA.

They have to do with "should we use a pace-free statistic", "should we use opponent adjustments", and... well, there's no reason whatsoever to ignore starting field position on defense, considering there's a clear correlation between starting field position and average points per drive.

410
by Peter Libero (not verified) :: Sat, 01/20/2007 - 3:30am

405: Well, it could be that their schedule is not particularly poor, but their opposing OFFENSES were. If they played a lot of teams with good defenses, that would bring those teams' overall DVOA (and probably the other rankings) up even if their offenses weren't so great. So they play Buffalo twice (24th), the Dolphins twice (28th), the Jets twice (15th), Denver (23rd, though they had Plummer and I think they might've been playing a little better at that point... although they only put up 17), Minnesota (29th), Green Bay (20th), Chicago (18th), Detroit (26th), Houston (16th), and Tennessee (25th).

They did also have to play Cincinnati (4th) and Indianapolis (1st), and Jacksonville (12th... were they better or worse with Garrard? I don't know.).

That's not exactly an inspiring group, it has to be below average, and thus the adjustment downward. Note several of those teams weren't so bad overall, but had weak offenses (Tennessee, Minnesota, Denver, etc.). Anyway, that's my instinct. I guess I could be wrong about it, but I have some faith in the opponent adjustments. No matter what it seems like it has to adjust downward.

411
by PantsB (not verified) :: Sat, 01/20/2007 - 4:48am

1) DVOA is pace-free, points per game is not. New England’s games have been shorter than basically everyone above them (and a whole lot shorter than Chicago’s).
2) DVOA is opponent adjusted, points per game is not.
3) The Patriots defense has had very good starting field position.
---
1 - Doesn't fly for two reasons. a) The Patriots gave up the 3rd fewest points and turnovers per drive. b) The Patriots offensive DVOA and ppg line up essentially exactly. If shortening the game justified penalizing the defense it would logically reward the offense for scoring the points it did in the 'shortened' game.
2 - By points scored, their opposition was average (20.1 ppg). Baltimore's was very similar 19.6. DVOA had NE=-2.29 offensive DVOA and Baltimore=-2.40.
3- Baltimore's LoS/drive is 28.17 (1st), New England's LoS/drive is 28.76 (5).

In other words, you're making things up to defend your preconceived idea.

412
by Peter Libero (not verified) :: Sat, 01/20/2007 - 3:52pm

Okay, for real PantsB, how can you not respond to the fact that VOA ranks the Pats 3rd, the same as their points allowed per drive? At least acknowledge you have read the words. What are you complaining about? What do you want?

I think you must not like FO's opponent adjustments. I'm not sure how to defend them, except to say that I'm pretty sure any system will downgrade the Pats. They played half their games against teams in the bottom quarter of offenses!

As for the Ravens, they had overlapping opponents of Tennessee, Denver, Indy, one game against Buffalo, and Cinci. They did get to play two games against Cleveland, who were 31st, Oakland, who were 32nd, and Tampa Bay, who were 30th (although I think they had Simms at the time). Then they played San Diego (2nd), New Orleans (5th), ATL and Carolina (17th and 19th), Pittsburgh twice (11th), KC (10th), and then another game against Cinci (4th).

So lets compare... the Pats played the Jets (#15) twice, Houston (#16) once, and Jacksonville (#12) once, so four of their unique games were against teams that were average or above (and three of them were right at average). The Ravens had unique games against #4, #11 twice, #10, #2, and #5. Which strikes you as more impressive?

The ravens did get to play 4 games against teams in the bottom 3, but New England played 3 games against 28/29.

413
by Pat (not verified) :: Sat, 01/20/2007 - 5:17pm

#411: I have to agree with #412: it really seems like at this point, you're not even trying to read. You compared New England to Baltimore's. Who cares about Baltimore? Baltimore's better in every defensive ranking. The team that's faced better offenses than New England is Jacksonville. The team that's had worst starting field position than New England is Minnesota and Miami.

Want to get a #8 rating for the Patriots defense, with just conventional stats? I'll spell it out for you.

1) Start with a pace-free statistic - now they move from #2 to #3. Chicago moves above them.
2) Add in opponent adjustments - now they go from #3 to #4. Jacksonville moves above them. (At this point, Minnesota, Miami, and even Carolina are also very close)
3) Add in adjusting for starting field position. Now they move from #4 to #6, with Miami and Minnesota going ahead, and Carolina's extremely close at this point, too.

The only two remaining are Carolina and Green Bay, and that's just because they evolved so heavily over the year, and the two rankings don't average the same way (there's not a linear relation between the two - holding someone to 10 points vs 10 is harder than holding someone to 30 points vs 20).

414
by PantsB (not verified) :: Sat, 01/20/2007 - 11:42pm

Okay, for real PantsB, how can you not respond to the fact that VOA ranks the Pats 3rd, the same as their points allowed per drive? At least acknowledge you have read the words. What are you complaining about? What do you want?

I think you must not like FO’s opponent adjustments. I’m not sure how to defend them, except to say that I’m pretty sure any system will downgrade the Pats. They played half their games against teams in the bottom quarter of offenses!
----
Have you been reading? This is not a one year phenomenon - if that was the case I'd dismiss it as a fluke. As I stated, using ppg or DVOA Baltimore and New England faced essentially identical quality of opponent's offense, produced essentially equivalent results but have wildly different results in terms of DVOA. Several teams with obviously (in essentially every category) inferior defenses have been rated higher than New England's. DVOA is not tracking with actual performance here. It seems it is rewarding defenses that quickly ending drives regardless of whether it actually benefits the offense (in New England's case offensive ppg and DVOA match up pretty perfectly) or results in less points being scored.

Since DVOA is based on guesswork that includes many ad hoc adjustments anyway rather than from a central thesis (and has produced a metric that overall is pretty good from this) suggesting an adjustment might be called for it not blasphemous.
======
Want to get a #8 rating for the Patriots defense, with just conventional stats? I’ll spell it out for you.

1) Start with a pace-free statistic - now they move from #2 to #3. Chicago moves above them.
2) Add in opponent adjustments - now they go from #3 to #4. Jacksonville moves above them. (At this point, Minnesota, Miami, and even Carolina are also very close)
3) Add in adjusting for starting field position. Now they move from #4 to #6, with Miami and Minnesota going ahead, and Carolina’s extremely close at this point, too.
------
This is just ridiculous. You're starting with the conclusion and making crap up to try and justify it. New England had the 2nd most INT and the 2nd most ppg. They gave up an identical number of offensive TD over the season - only 2 FG separated the two Defenses. They faced opponents of almost identical offensive abilities. Thats why I've made the comparison - their DVOA should be similar because of their similar results.

Your reasoning just doesn't even have internal coherence. New England is in the top 5 of every one of the statistics you list and top 3 in the most important ones, but you argue that using them they should be #8. Does that really make sense to you?

415
by PantsB (not verified) :: Sat, 01/20/2007 - 11:56pm

The argument has gotten muddled with many red herrings, etc. Let me present the central thesis of why I think an adjustment is needed:
-New England has given up the 2nd least number of points per game, (6 points more than league leader Baltimore in terms of non-return TDs). If pts/drive is for some reason considered superior, it has given up the 3rd fewest pts/drive (.08 pts less than Chicago). They have the 2nd most interceptions.
-In terms of opponents, New England has faced superior offensive teams than Chicago, and nearly identical opposition to Baltimore.
-Despite this, New England is rated 3rd in team defense. Additionally, similar phenomena has occurred in 2001, 2003 and 2004 when defensive DVOA ranking was lower than points/game(or drive) and turnovers would indicate and specifically lower than teams with inferior points/game (or turnovers) results that play a different style of defense.
-The New England offensive DVOA and points per game line up perfectly. There is no evidence in terms of field position that New England's offense has a more difficult time scoring than the average offense (or the offense of an excellent D like Baltimore, where offensive ppg and DVOA also essentially match). There is no disadvantage evident in DVOA to the New England style of defense.
-Since DVOA already has (arbitrary) adjustments in certain circumstances in order to attempt to create greater correlation to points and wins (the justification of DVOA in the first place), making an adjustment to create a greater correlation in terms of points is not an extreme suggestion.

416
by PantsB (not verified) :: Sat, 01/20/2007 - 11:57pm

Above, That should be .08 more than Chicago and "-Despite this, New England is rated 8th in team defense."

417
by Peter Libero (not verified) :: Sun, 01/21/2007 - 5:48am

Okay, well I think I've definitely made the case that the Patriots have not played better offenses than the Ravens, in terms of DVOA. The only good offenses that the Pats faced were also played by the Ravens, and then Baltimore had another bunch of games against good offenses.

No one has said your suggestions are "extreme." But you have not actually suggested an adjustment, nor have you even suggested what you think is a fair ranking! If New England is third in VOA, how high do you think it should be? #1? Is that really a huge deal to you?

If the problem is DVOA, then clearly it's the "D" part that's getting to you. Have the Patriots really played even an average slate of offenses, let alone ABOVE average? I would have to say they didn't, so their score gets adjusted downwards, meaning they are lower than #3. Shucks.

I also think that your comment that DVOA is based on "guesswork" and and "arbitrary adjustments" is off-base. I'm actually kind of shocked that Aaron hasn't contributed anything to this discussion by now, but DVOA's adjustments are made because they make it a more effective metric, and are always done rationally. Additional things have been considered when they were suggested, and are either incorporated (usually making a very small difference) or rejected because they do not improve correlation with future wins.

You have also claimed that this is not a "fluke," but that DVOA has consistently underrated the Patriots' defense. Again... they were 6th in 04 and 2nd in 03... how high do they need to be? You've yet to even declare what the goal is, or how good you (based on whatever factors) imagine the Patriots to be. They are THIRD in VOA!

418
by PantsB (not verified) :: Sun, 01/21/2007 - 2:54pm

Okay, well I think I’ve definitely made the case that the Patriots have not played better offenses than the Ravens, in terms of DVOA.
---No you haven't. DVOA is mathematical so it can be objectively combined. The mean offensive DVOA for Baltimore is worse than New England's and Chicago is worse than that (NE=-2.29 offensive DVOA, Baltimore=-2.40, Chicago (and Minny since its almost an identical schedule) I forgot to write down but its ~ -5.1 IIRC).
======
I also think that your comment that DVOA is based on “guesswork� and and “arbitrary adjustments� is off-base. I’m actually kind of shocked that Aaron hasn’t contributed anything to this discussion by now, but DVOA’s adjustments are made because they make it a more effective metric, and are always done rationally.
-----
Currently, red zone defensive DVOA is afforded 20% greater value. A study in DVOA from 2003 showed that red zone D was not really more important in determining a correlation with points - but did find that from the 10-40 was. Unless the information was hidden, that was never taken into account. DVOA is a good metric because it correlates pretty well with points. Here it is consistently not based on defensive style.

I say guess work because what constitutes "success" on a play or the value of a big play was selected and then tested. It wasn't generated by an underlying formula as I understand it.

And defensive DVOA has the Patriots a distant 3rd when unweighted yes. However, there's a huge gap there (New England is "half" of Baltimore or Chicago) that points/turnover based stats don't indicate. New England is passed by teams with inferior points/turnover based stats, who have inferior opponents because - it seems - style of play. This also occurred in the past: In 01, the Patriots were 6th in points (6th in drive adjusted) and 13th in Def DVOA (10th non opponent adjusted). In 03, the Patriots were much better than any D in the league first in points by a great deal, first in points/drive, marginally 2nd in TO/drive (to St Louis 7th points/drive) but was a fairly distant 2nd to Baltimore (2nd in points, 8th in turnovers/drive) in DVOA. In 2004, New England was 4th in pt/drive, 2nd in turnovers/drive and 6th in DVOA.

DVOA is correlating almost exactly at the top with yards/drive. Thats not a good metric that doesn't correlate with wins or points very well.

419
by Peter Libero (not verified) :: Tue, 01/23/2007 - 3:53am

I wonder if this is an FOMBC situation? Does the opposition have to be irrational, or is any criticism of the DVOA system fair game? I haven't been around long enough to know this kind of trivia.

420
by mactbone (not verified) :: Tue, 01/23/2007 - 1:51pm

Re 419:
Not strictly speaking no. The first case was the Falcons and that was a horde of angry trolls from other message boards. There have been few indiviuals as determined as PantsB though and that might be the difference - however, there was, kind of, an argument there and correct spelling. Basically I see this as a wash situation. Not a good enough troll however determined PantsB was.

421
by Pat (not verified) :: Thu, 01/25/2007 - 9:02pm

This is just ridiculous. You’re starting with the conclusion and making crap up to try and justify it.

No. I'm showing how the rating is reached.

Which part did you have a problem with?

1) Pace-free stat
2) Opponent adjustments
3) Starting field position correction

None of those were chosen to bias for or against the Patriots. It just so happens that they fall exactly along the wrong side of all of those stats.

For a similar situation, look at the Colts offense, which is underrated in normal scoring stats for exactly the same reason (few drives, better defenses, worse starting field position). Indy similarly gets screwed in every single one of those stats, surprisingly.

You're actually the one who was cherrypicking random statistics which prove your point.