Matthew Stafford

DVOA Week 1: NFC West is Best

The New Orleans Saints lead the Football Outsiders DVOA ratings after Week 1, thanks to their 38-3 dismantling of the Green Bay Packers. But the real headline is the hot start for the NFC West. The four teams from that division all finish in the top six for Week 1, with the Rams at No. 2, the Cardinals at No. 4, the 49ers at No. 5, and the Seahawks at No. 6.

The AFC West also swept its Week 1 games, but those teams are not as high in our ratings after playing closer games. We have Denver at No. 9, the Los Angeles Chargers at No. 15, Las Vegas at No. 15, and Kansas City down at No. 22. The Chiefs are the only team to win in Week 1 despite a lower DVOA rating than their opponent, the Cleveland Browns.

(Of course, though I'm calling our main metric DVOA here, it is actually VOA because there are no opponent adjustments right now. We do not apply opponent adjustments until after Week 4, so in Weeks 1-3 DVOA and VOA are the same thing.)

The close win over Cleveland doesn't hurt the Chiefs very much in our DAVE ratings. Those are the ratings that combine our preseason projection with the results of early games to give us a better prediction of how each team will rank at the end of the year. For those who don't know the story, this metric is called DAVE as a reaction to criticism that our stats are too much alphabet soup. I mean, who can argue with a guy named Dave? (Technically, it stands for "DVOA Adjusted for Variation Early.") In these week's DAVE, the preseason forecast still counts for 93% of the rating. Still, Week 1 did do some shifting to DAVE, with the Saints and Rams in particular moving up while the Packers and Titans fell.

Kansas City beating Cleveland despite the lower DVOA lets me introduce a new toy that I've been playing with over the last few weeks. It's called Post-Game Win Expectancy (PGWE). The idea is this: How often should we expect each team to win an NFL game given how the two teams played overall? We all know there are close games where the "wrong team wins," or at least it seems like the wrong team wins. This is a measurement of that.

This idea comes directly from Bill Connelly, who does something similar for college football. The Football Outsiders NFL PGWE is going to work a bit differently from what Connelly does. His PGWE consists of elements that go into his SP+ formula, such as success rates, marginal explosiveness, turnover chances, and average field position. Our PGWE consists specifically of DVOA splits. Well, technically they are actually VOA splits, because opponent adjustments are not included. (You don't get a bonus for playing Kansas City when it comes to the actual game on the field and turning your play into a win or a loss.) The PGWE formula uses both total VOA and VOA split into passing and rushing, with passing roughly three times as important as rushing in the formula. Yes, as you probably imagine, the team that passes better is more likely to win the game than the team that runs the ball better. Special teams DVOA is also part of the formula.

I tried to include some variables other than VOA splits in my PGWE formula, including penalties or penalty yards, total number of plays, and explosive plays (both plays over 10 yards and plays over 20 yards). None of those ended up improving the correlation with actual wins. One thing I didn't do which is part of Bill Connelly's formula is to account for the idea that interceptions are a somewhat random percentage of passes defensed in the same way that fumble recoveries are a somewhat random percentage of actual fumbles. Instead, DVOA only accounts for the actual interceptions. This is something I want to play around with in the future.

You're probably looking to see the PGWE for Week 1 of the 2021 season, so without further ado:

Post-Game Win Expectancy, Week 1 2021
Win Pts Loss Pts PGWE
NO 38 GB 3 99.97%
ARI 38 TEN 13 99.97%
PHI 32 ATL 6 99.92%
LAR 34 CHI 14 99.7%
SF 41 DET 33 99.6%
HOU 37 JAX 21 97%
SEA 28 IND 16 97%
CAR 19 NYJ 14 97%
PIT 23 BUF 16 86%
TB 31 DAL 29 80%
DEN 27 NYG 13 80%
MIA 17 NE 16 69%
LV 33 BAL 27 60%
KC 33 CLE 29 41%
LAC 20 WAS 16 40%
CIN 27 MIN 24 34%

As you can see, not all close games are created equal when it comes to PGWE. The formula believes that the Panthers (34.3% VOA) significantly outplayed the Jets (-40.5%) and the 49ers (39.0% VOA) completely outplayed the Lions (-54.6% VOA). On the other hand, you have three different games where PGWE has an unexpected winner. The specific splits of VOA matter here, because in only one of these games (Chiefs-Browns) did the losing team actually end up with a higher total VOA than the winning team.

The Cincinnati-Minnesota game may be the most interesting one here. Both teams ended up with positive VOA overall, but the PGWE is based on the idea that passing is usually more important than rushing. Therefore, Minnesota ends up with the higher PGWE thanks to an advantage of 51% to 38% in pass offense DVOA, even though Cincinnati had a larger advantage of -9% to -45% in run offense DVOA. Of course, pass/run ratio matters as well, which is why the formula also incoporates total VOA (specifically, total defensive VOA) but even here the Vikings were slightly better, 2.4% to 5.3%. The big difference between PGWE and the result of this game was penalties, where Minnesota had 12 of them for 116 yards while Cincinnati had only three for 15 yards.

The Chargers-Washington game is also about the difference between passing and rushing. This may surprise you given how many accolades Justin Herbert got for some great throws in this game, but it was Washington that finished with the better pass offense VOA, 24% to 4%. The Chargers were the better rushing team, -19% to -37%.The Chargers were the better offense overall because passing is generally more efficient than rushing and was in this game, and the Chargers passed about two-thirds of the time while Washington passed only slightly more than half the time.

The Kansas City-Cleveland result should not surprise you, especially if you know that Cleveland outgained Kansas City on average 8.2 to 6.5 yards per play. As noted above, it wasn't just PGWE, Cleveland also has the higher VOA rating in total for the game.

Last year, there was a .80 correlation between PGWE and actual wins. (This is slightly higher than the correlation that Bill Connelly has for his college version of PGWE, which is .78.) To give an example of the kind of game where the final result is very different from the PGWE, here are the five most "unexpected" wins from the 2020 regular season, the games with the lowest PGWE for the team that took the win:

1) Week 9, New England 30 at New York Jets 27 (6%): This was a Monday night game which the Patriots won with a last-second field goal. Joe Flacco subbed for an injured Sam Darnold and was fantastic, 18-of-25 for 262 yards through the air. So the Jets had 115% pass offense VOA compared to 66% pass offense VOA for the Patriots. Overall, the Jets outgained the Patriots 7.3 to 5.7 yards per play, but the Patriots ran far more plays, 76 to 44. "The winning team ran many more plays despite being less efficient" is a general theme of "unexpected wins." As noted above, I tried to account for this by adding a variable based on the number of plays each team had, but it didn't do anything to improve the formula.

2) Week 12, Arizona 17 at New England 20 (7%): Here we go with the Patriots again. New England won this game despite -144% pass offense VOA. Cam Newton was 9-of-18 for just 84 yards with two interceptions and three sacks. Overall, the Cardinals had more yards per play (4.3 to 3.5) and ran more plays (70 to 51). The Cardinals won the turnover battle (+1) and the two teams had the same number of penalties. In retrospect, a mind-boggling loss for Arizona.

3) Week 2, New Orleans 24 at Las Vegas 34 (12%): You probably remember this game, a Monday night contest where the Saints committed roughly 11 billion penalties. OK, technically it was 10 for 129 yards compared to the Raiders having just three for 13 yards. The Saints outgained the Raiders in this game, 7.4 to just 5.0 yards per play, and the turnover battle was even, and the Saints lost anyway primarily because of all those penalties. Drew Brees looked bad despite gaining all those yards and everybody talked about how his arm as shot. I had to spend a lot of time explaining this one in the DVOA column the next day.

4) Week 5, Indianapolis 23 at Cleveland 32 (16%): This is an interesting one because the two teams were essentially even in offensive and defensive VOA although the Colts did outgain the Browns 5.9 to 5.3 yards per play. The big gap was in special teams, mostly because Isaiah Rodgers had a 101-yard kick return for a touchdown for the Colts. And yet the Colts lost the game anyway.

5) Week 5, Tampa Bay 19 at Chicago 20 (17%): This was a Thursday night game, another one you probably remember. It was brought up for much of the season as evidence that the Buccaneers were struggling even though DVOA had them as one of the top two or three teams in the league. The Bucs outgained the Bears 5.3 to 4.1 yards per play and ran more plays as well, with the turnover battle even at one apiece, but the Bucs had issues finishing drives (four field goals) and then Tom Brady lost track of the downs when the Bucs were trying to come back in the final minute.

* * * * *

Football Outsiders playoff odds are updated through Week 1. The annual stats pages are now updated with 2021 data, although some of that data can be kind of sketchy after just one week (in particular the offensive line and defensive line pages). Snap counts and the FO+ DVOA database are also now fully updated through Week 1.

There's a technical issue with the player position stats pages that's keeping us from showing most player stats because of how the table minimums are set. We're working on fixing that issue.

A reminder that all our free stats pages, including DVOA and player position stats, now require registration to view. This is not a paywall! You only need to register (for free) and then log in to the site to view these pages. While you're at it, you can get a seven-day trial of FO+ and check out the FO+ features like a deeper DVOA database, weekly fantasy projections, fantasy football research tools, and picks against the spread.

* * * * *

These is the Football Outsiders Top 16 through one week of 2021, measured by our proprietary Defense-adjusted Value Over Average (DVOA) system that breaks down every single play and compares a team's performance to the league average based on situation in order to determine value over average. (Explained further here.) Please note that there are no opponent adjustments in DVOA until after Week 4, which is why it is listed here as VOA.

OFFENSE and DEFENSE VOA are adjusted for performance indoors and consider all fumbles, kept or lost, as equal value. SPECIAL TEAMS DVOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.

DAVE is a formula which combines our preseason forecast with current DVOA to get a more accurate projection of how a team will play the rest of the season. DAVE is currently 93% preseason forecast and 7% actual performance.

To save people some time, please use the following format for all complaints:

<team> is clearly ranked <too high/too low> because <reason unrelated to DVOA>. <subjective ranking system> is way better than this. <unrelated team-supporting or -denigrating comment, preferably with poor spelling and/or chat-acceptable spelling>

RK TEAM TOTAL
DVOA
TOTAL
DAVE
RANK W-L OFF.
VOA
OFF.
RANK
DEF.
VOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
1 NO 78.4% 9.8% 6 1-0 25.4% 7 -45.8% 1 7.1% 5
2 LAR 67.2% 12.1% 4 1-0 56.6% 2 -9.1% 11 1.5% 11
3 PHI 56.5% -5.8% 22 1-0 21.8% 9 -31.8% 4 2.8% 9
4 ARI 52.4% 2.7% 17 1-0 11.7% 13 -43.0% 2 -2.3% 20
5 SF 39.0% 9.6% 7 1-0 35.8% 3 -8.3% 13 -5.1% 25
6 SEA 35.5% 11.9% 5 1-0 29.5% 5 -7.2% 14 -1.2% 16
7 CAR 34.3% -7.8% 25 1-0 0.2% 15 -35.8% 3 -1.7% 17
8 HOU 30.2% -18.7% 31 1-0 -1.7% 17 -28.0% 6 3.9% 7
9 DEN 20.9% 4.3% 13 1-0 26.6% 6 5.8% 20 0.2% 14
10 MIA 17.7% 3.0% 16 1-0 5.0% 14 -9.8% 10 3.0% 8
11 CLE 16.8% 3.5% 15 0-1 57.1% 1 30.3% 28 -10.0% 29
12 CIN 15.6% -5.6% 21 1-0 12.8% 12 5.3% 19 8.1% 4
13 TB 14.9% 20.3% 1 1-0 17.5% 10 16.3% 24 13.7% 1
14 MIN 8.9% 5.3% 10 0-1 -1.1% 16 2.4% 18 12.4% 2
15 LAC 5.7% -7.1% 24 1-0 -4.9% 20 -14.6% 7 -4.0% 24
16 PIT 4.4% 4.6% 12 1-0 -13.9% 26 -28.7% 5 -10.4% 30

Click here for the full table.

Comments

75 comments, Last at 17 Sep 2021, 11:21pm

27 That's why

In reply to by Raiderfan

Guys, the entire DVOA table is for paying members. Umm, the staff here does have to make a living, you know.

30 Actually...

In reply to by BigRichie

Registered members, actually! We just need you to register for the site. It's not behind a paywall.

51 Dynamically included

In reply to by Aaron Schatz

If one were so inclined, they could have the article check to see if the user is logged in and show the full table in the article. If not logged in, then only show half.

31 DON'T HAVE TO PAY

In reply to by BigRichie

You don't have to pay to see the whole table. Just be registered. I still find it annoying to have to make another click. If I'm on a desktop/laptop where split screen/multi screen is easy it's still easy to reference the whole thing for a comment. But on a mobile device I can't just scroll back up to see the whole table. It would actually make more sense if it was behind a paywall and not just a registration wall. Though I guess a bot scraping data would have a harder time getting it with just the registration wall. Maybe there is another reason, maybe registration walls have some proven link to engagement or something. To me it's just annoying.

32 No, anyone can access the…

In reply to by BigRichie

No, anyone can access the entire DVOA table, simply by clicking through -- paid member or not.

The only thing it requires is to have an account -- not even an FO+ account, just a regular, free account!

 

(Though, yes, I find I do enjoy both food and shelter.)

38 Accessing the full table

I was also interested in seeing the full table.  So I spent the energy necessary to use my mouse to guide the cursor to the link.  But that's not all.  I then was forced to use my dominant index finger to press lightly upon a button on the mouse to actually activate the link.  It was only then that I was able to view the full table but, needless to say, was far to taxed to comprehend the contents.  Who are these monsters?

43 Wow. I had no idea that the…

Wow. I had no idea that the complaints of all the above were totally invalid. Good thing BroncosGuyAgain has figured out that people are complaining for literally no reason. Literally no reason at all. 

Q: Is it annoying for mobile device users that don’t support split screen to have to switch back and forth tabs?

A: Trick question, because there is literally no reason why anyone could possibly complain about this obnoxious site feature.

Q: Does having to open said new tab, read the contents, then go back to the article break up the flow of the article?

A: Trick question, because there is literally no reason why anyone could possibly complain about this obnoxious site feature.

And so forth.

45 You know, companies spend…

You know, companies spend millions of dollars paying teams to carefully internalize feedback from users. Yes, that includes among other things, the placement of a button, the location of a table, or the flow of a webpage.

For example, a data scientist will often design a very careful ab test with thousands of lines of code, hours spent internalizing the results, and then present those findings to stakeholders.

If warranted, this will then kick off a redesign at which point a ui/ux designer will make a design change and a front end developer will rewrite the JavaScript to the desired result. 

At a minimum, assuming this is a bare bones startup with no dev ops, security, or legal considerations or formal code and commit process, that's three tech salaries and hours of work. All for a useless button.

Man, how dumb are those idiots to fret over something sooo frivolous, when they could have just read the wisdom of luminaries here on these message boards and they too might have the same "ah ha" epiphany. 

 

57 As someone who has worked in…

As someone who has worked in software design for many years, I completely concur with "theslothook". There's a lot of analysis and information gathering that goes into making even slight modifications to the user experience of a website or product. The complaints above are valid complaints though, especially for long-timers like myself who just get used to things working a certain way. I'm sure FO is paying attention to them (as the responses here from them confirm).

That said, I think the suggestion above to simply check on page load if the user is signed in or not, and if so display the whole table, if not don't, is a pretty simple solution to this feedback. We'll see what happens!

2 "The Chargers-Washington…

"The Chargers-Washington game is also about the difference between passing and rushing. This may surprise you given how many accolades Justin Herbert got for some great throws in this game, but it was Washington that finished with the better pass offense VOA, 24% to 4%. The Chargers were the better rushing team, -19% to -37%. And in an example of Simpson's paradox, the Chargers were the better offense overall because passing is generally more efficient than rushing and was in this game, and the Chargers passed about two-thirds of the time while Washington passed only slightly more than half the time."

This isn't an example of Simpson's Paradox.  It would be if the Chargers had worse passing DVOA and worse rushing DVOA, but higher overall DVOA.

Here's how it could work (ignoring penalites):

- LAC runs 50 pass plays at 10% DVOA, 20 rush plays at -10% DVOA
- WAS runs 20 pass plays at 15% DVOA, 50 rush plays at -5% DVOA

LAC's overall DVOA would be 4.3%(*), while WAS's would be 0.7%(**), because of the play distribution.

(*) (50 * 10 + 20 * -10) / 70
(**) (20 * 15 + 50 * -5) / 70

4 DVOA Scale

These DVOA numbers seem way larger (both for the high DVOA and low DVOA teams) than they have been in the past. Did the calculation change?

8 Nope

In reply to by splutch

Nothing has changed. You're just reacting to seeing just one week of DVOA instead of several weeks. Once teams have played several games, the numbers will get closer to 0%.

15 78.4% actually seems kinda…

In reply to by splutch

78.4% actually seems kinda low for a weekly leader, based on my recollections of past Week 1s. I think there's usually someone at 90+%.

5 Packers

I'm curious how teams that have had Week 1 VOAs as bad as the Packers (-94.6) have fared historically in terms of final DVOA rankings or making the playoffs. More broadly, are Week 1 results more/less/the same correlation to full season results compared to any other individual week of the season?

34 Let's run some numbers

In reply to by nickqp

For 1983-2020, there are 44 teams that had a DVOA of -90% or worse after Week 1. Eight of those teams made the playoffs, including last year's Cleveland Browns. An additional team (1990 Seahawks) had a winning record but missed the playoffs. However, only four of those teams ranked in the DVOA top 10 by the end of the year: 1985 Jets, 1997 Steelers, 2003 Patriots, and 2011 Steelers.

35 How often does a great team play that bad?

Do teams that finish in the top-5 in DVOA (or thereabouts) lose games like that very often? That feels more useful than just looking at an isolated week, at least in terms of whether to panic about the Packers or not.

36 They hate their coach

I think that 2003 Pats game was the Lawyer Milloy game, where they got absolutely destroyed by the Bills.. 35-3 or something like that.   And then went on to go 14-2 and win the SB.   Not sure how those other teams fared though...

37 They hate their coach

I think that 2003 Pats game was the Lawyer Milloy game, where they got absolutely destroyed by the Bills.. 35-3 or something like that.   And then went on to go 14-2 and win the SB.   Not sure how those other teams fared though...

6 Updated DAVE

Where does one find the updated DAVE for offense, defense and special teams by team?  Used to be easy to find on the old site, but I can't seem to find it.

11 Free stats pages

In reply to by colunchbox

These stats are in the same place they've always been, the free stats pages. Hover over the Statistics dropdown menu above, then click on "Offense" or "Defense" under "Overall DVOA" and you'll find the DAVE stats on those tables.

9 Simpson

" And in an example of Simpson's paradox..."

It would only be Simpson's paradox if WAS had better DVOA in both rushing and passing, but LAC had better offensive DVOA overall.

Oh, I see Eddo already said this.

OK, it's only Simpson's paradox if Dr. Hibbert and Nick Riviera are involved.

Wait...that's a pair o' docs...

10 NFC West

Looked to be the toughest division in football before the year. Priors: confirmed!

28 Another insane tidbit about…

In reply to by Sportszilla

Another insane tidbit about the NFC West teams. Every single team held a 3 score lead at some point in the second half. 3 of those were on the road in 10 am PST games. That is an insane stat because of how much tougher those early start road games usually are for west coast teams. This division is going to be absolutely nuts. 

21 I have to say, as a fan of…

I have to say, as a fan of the Rams, I am desperately hoping the DAVE rankings for the NFCW is more accurate. 4th (Rams) 5th (Hawks) 7th (9ers) and 17th (Cards) is a division that, while the best in football by a mile, is one you can reasonably hope to go 4-2 in. 

In contrast, the DVOA of 2nd, 4th, 5th, and 6th would make the NFCW easily the best division in all NFL history. And it’s not just the rankings, every single team crushed their opponents last week, with the 9ers looking like the worst team, only for DVOA to explain that the game was less close than the score.

And it’s really hard to see why the DVOA is so far off. The Rams have essentially the exact same team, with a strong upgrade at QB. The Cards have a third year QB getting better, plus other roster additions. The 9ers have all players returning from injury, with the possibility of upgrading from Garoppolo. The Hawks look to remain a ~20% DVOA team, because why wouldn’t they? 

I really don’t see what the cause for hope would be for one of these teams ending up even as merely average, besides horrible injury luck. Maybe the Cards won’t see Murray take the next step forwards, but he sure seemed to yesterday when they were dismantling what everyone thought was a decent Titans squad. 

As always, by the end of the season I’m sure something will happen to make this take look extremely foolish.

26 Depth or lack thereof

Every QB in the division except Wilson has injury concerns that range from moderate-to-serious, and of course Wilson is just one play away from disaster himself.

Beyond that, all of these teams are relying on (very talented) older players are crucial positions: the Rams, 49ers, and Seahawks at left tackle, the Cardinals on the D-line, and all of them have real question marks in the secondary.

Week One with basically everyone at full strength is one thing, but who can stay healthy or overcome injuries will go a long, long way to determining the division, as unfun as that is to mention.

39 The best division in NFL…

The best division in NFL history is probably the 2013 NFC West, which saw the teams finish 1, 6, 10 and 12 with an average DVOA of 18.7%, so that's the target. Record-wise I think they have a chance of beating that division's winning percentage of 65.6%, as their main AFC matchup is the horrible AFC South, and their main NFC matchup is the mediocre NFC North. Right now the Playoff Odds page has every team finishing above .500, and expects 2.49 teams to make the playoffs; I wonder how many of those simulations had all four teams making it.

42 Like a commentator above…

Like a commentator above pointed out, injuries probably cap the projection somewhat. One of the four teams could easily have a serious QB injury, or the 9ers 2020 problem of tons of star player injuries.

Still, the top 10% projection, of all four teams stay healthy and all four teams keep playing at such a high level, how wild would that be?

EDIT: And as you said, the very weak AFCS, combined with the somewhat mediocre NFCN, definitely gives these teams back some wins that they will steal from each other.

52 I was thinking about this...

In the unlikely event that all NFCW teams stay healthy, it wouldn't be crazy, considering matchups, for more than one to go 11-0 outside the division.

Especially if GBs performance holds up.

It helps that TB will play the Rams rather than Seattle which a history of imploding badly against good d-lines.

68 The 49ers have already,…

The 49ers have already, after one game, lost their starting RB for the season, their best CB for the season, and one of their starting LBs for 6-8 weeks.  Their other starting CB is questionable, their best DT has an inflamed knee that won't calm down and so is out indefinitely, and their #1 WR has hamstring problems such that he received 0 targets last week.

What the ding-dang gosh darned dash blast the blankety heck?

53 Best to not overreact to one…

Best to not overreact to one game, especially keeping in mind that opponent adjustments haven’t started yet.  The Lions are definitely bad, and the Dalton Bears and Wentz Colts are probably bad.  The Titans maybe just a had a bad week, but who knows?

22 Small sample size theater!

Despite still having to play the entire AFCW, the Bengals now have the 2nd easiest remaining schedule.  Thanks, GB!  (and the rest of the NFCN and Jax, I guess).  Clearly it's all smooth sailing from here!   I mean "Cincinnati is clearly ranked too low because everyone knows the first week sets the tone for the rest of the year!  Desperate straw-grasping is way better than this!" 

29 One of the most interesting…

One of the most interesting aspects of the 2021 season is the 5 new first round rookie QBs. My impression of them, watching only the ~10 minute YouTube highlights from the games, then combining that with what others are saying.

1) Trevor Lawrence made a few very nice throws, combined with a few terrible mistakes. Kind of middle of the line.

2) Zach Wilson I’ve heard mixed things. One person said he didn’t belong. Another said he made some very nice throws. I’ve also heard he was hit a lot, and that’s a bit worrying.

3) Trey Lance. Just a few plays, but he looked good when doing his package.

4) Justin Fields. See above. I did watch this entire game as a rams fan, and I thought he looked fine doing his package plays, if not really noteworthy other than being a rookie.

5) Mac Jones. Many think he had the best week, and I saw a lot of very accurate passes, but even extended highlights can skew opinion.

Curious to see what others here think, especially if you saw the full games these guys played.

40 Zach Wilson

Terrible first half. Deer in the headlights look.  Arm strength looked weak. Got more comfortable as game went on. Feeble throw for the much needed late two point conversion.

44 correlations

Last year, there was a .80 correlation between PGWE and actual wins.

What is the correlation between VOA difference and actual wins?

73 The answer:

In reply to by zenbitz

The correlation between (HOME VOA - AWAY VOA) and Wins was .71.

46 Mia-NE

I'm scratching my head trying to understand how a 1 pt loss by the Pats ended with such a difference with Miami. It's weird because the Pats defense and offense felt like they were stronger units all game. I guess those fumbles are really weighed heavily on them and for Miami's defense. Not sure why that would hurt the Pats D, though. 

47 new metrics

So now there's a metric called PGWE.  Seems to me you could have tweaked the terminology enough to get the acronym PAWG.  Might not attract the audience you want, though. 

 

48 Is DAVE only available for the top 16 DVOA teams?

Would be nice to have Kansas City, who surely rates highly by DAVE but is absent because they didn't crack the top 16 in DVOA in week 1.

 

Feels like previous years DAVE was available for everyone on the team total page.

49 Calibration of PGWE

There were 256 games last year + playoffs, right (so 267 total)?  Seems like you should have very low probability events occurring more frequently than you have them (a team winning when they had only a 6% chance of winning is not that surprising over 267 samples).  I haven't constructed models for any of this stuff, but my guess is that this is more of a question of scaling than anything else.  Would think about it, though - KenPom's game predictions tend to break down at the margins as well.  Just seems like the margins should be a bit further towards the tails.

(Big picture, that's a nit - love the writing / analysis.  Just trying to give something back for all the entertainment Aaron + team have given over the years.)

60 hmm...

The problem is that the win probabilities observed depend a lot more on the relative strengths of the teams (i.e., the distribution of the PGWE stat) than on the outcome of the given game. There are a lot of games here with PGWE > 95%.  But in all of last year, we didn't see a single game with PGWE < 5%.  Testing the probability estimates of rare events is really a pain, because n has to be quite large.

 

59 Estimated wins

I am curious how the Saints rate the best after this week, and yet are only credited with .8 estimated wins; while the Rams at #2 are credited with a full estimated win. 

61 opponent (?)

In reply to by Joseph

The Packers are considered to be a much better team than the Bears.  (I suspect DAVE is used, not VOA.) 

74 Actually

In reply to by RickD

It's a quirk of the Estimated Wins formula. The Rams had 100% DVOA in the second half of close games. The Saints had 0% because... it was never a close game. So the Rams end up higher.

Estimated Wins wasn't really meant to be used for single games -- I would fix this issue if it was.

63 Accessing DAVE

Hello 

I’m a paid member and can’t access the full DAVE numbers. When it asks me to log on to see the full table it doesn’t show DAVE any more. Please advise. 
thanks,

Alex

 

 

65 DAVE

I see it now, thanks! 
 

71 It seems to me that by…

It seems to me that by constructing PGWE from pass and rush VOA you're begging the question of whether either is actually "more important" than the other. VOA was built to evaluate units/players so it's inherently pegged to position and play tipe, but it seems to me that if you want to ascertain team performance and nothing else (which seems to be the goal here), it would be better to consider each play in terms of end state only, i.e., (points, yards, time).

This may explain why in 4/5 of the games listed, the team with the higher seasonal rush YPC won (quick and dirty analysis).