DVOA/DYAR

[Eyes Of Adam]

I see a lot of reference to DVOA and DYAR when it comes to Darnold vs other QB's. I think advanced stats are great, but they have to be used properly and you also have to understand their limitations their methodology.

I did some research to understand them a bit more, and right from the horses mouth comes the following:

ISSUES WITH DVOA/DYAR

DVOA is limited by what’s included in the official NFL play-by-play or tracked by the Football Outsiders game charting project. Because we need to have the entire play-by-play of a season in order to compute DVOA and DYAR, these metrics are not yet ready to compare players of today to players throughout the league’s history. As of this writing, we have processed 26 seasons, 1989 through 2014, and we add seasons at a rate of roughly two per year (the most recent season, plus one season back into history.)

Football is a game in which nearly every action requires the work of two or more teammates -- in fact, usually 11 teammates all working in unison. Unfortunately, when it comes to individual player ratings, we are still far from the point at which we can determine the value of a player independent from the performance of his teammates. That means that when we say, "In 2014, Marshawn Lynch had a DVOA of 23.1%, what we are really saying is “In 2014, Marshawn Lynch, playing in Darrell Bevell’s offensive system with the Seattle offensive line blocking for him and Russell Wilson selling the keeper when necessary, had a DVOA of 23.1%."

With fewer situations to measure, the numbers spread out a bit more, so you'll see more extreme DVOA ratings for part-time players and for measurements of teams in more specific situations (for example, passing on third downs). The charts listing players in order of DVOA have cut-offs for number of attempts, because players with just a handful of plays end up with absurd VOA and DVOA numbers. (In 2014, for instance, Johnny Manziel had a -73.2% passing DVOA in 38 plays.)

Passing statistics include sacks as well as fumbles on aborted snaps. Receiving statistics include all passes intended for the receiver in question, including those that are incomplete or intercepted. At some point, we hope to be able to determine just how much impact different receivers have on completes vs. incomplete passes, but various regression analyses make it clear that both quarterback and receiver have an impact on whether a pass is complete or not. The word passes refers to both complete and incomplete pass attempts.

Unless we say otherwise, all references to third down also include the handful of rushing and passing plays that take place on fourth down (primarily fourth-and-1).

 

The bold textz above is absolutely critical to understand when evaluating these advanced statistics. DVOA and DYAR do not isolate the player from the team or the system. They are team and system statistics far above being independent player statistics.

I would love to get some discussion going with those who use this stat as an individual statistic, so I can understand better their logic. Advanced stats are really cool and play a big part in sports.

[Untouchable]

Analytics are useful in baseball, not football.

Bill Belichick straight up dismisses them. That should tell you all you need to know.

I still don’t understand how the hell an analytics site like PFF grades players. 

I’ve said it numerous times, but 5-6 years ago they had Ben Hartsock graded as a Top 3 TE in the league despite not catching a single pass all season. I’d love for someone to explain that silly sh*t to me.

[Eyes Of Adam]

1 minute ago, Untouchable said:

Analytics are useful in baseball, not football.

Bill Belichick straight up dismisses them. That should tell you all you need to know.

I still don’t understand how the hell an analytics site like PFF grades players. 

I’ve said it numerous times, but 5-6 years ago they had Ben Hartsock graded as a Top 3 TE in the league despite not catching a single pass all season. I’d love for someone to explain that silly sh*t to me.

Belicheck uses statistics quite a bit, he just doesnt use individual player stats or metrics. But he uses teams tendencies quite a bit which are absolutely analytics.

Having played football most of my life, and coached it, I think there is a place for individual analytics, but I think its incredibly difficult to normalize and isolate the variables to do it properly.

[#27TheDominator]

Belichick was talking about going for 2 point conversions, when to kick a field goal or go on 4th down.  I think the Pats were actually on the cutting edge of using it for scouting purposes - which is what we are discussing here.  Also, Bill Belichick, the 2013 winner of the MIT Sloan Analytics conference Lifetime Achievement Award, is full of sh*t: 

Bill Belichick downplays analytics again, but Patriots know value of math

[CanadienJetsFan]

My head hurts reading this.

I’m sure it’s helpful though.

[dbatesman]

8 minutes ago, #27TheDominator said:

Belichick was talking about going for 2 point conversions, when to kick a field goal or go on 4th down.  I think the Pats were actually on the cutting edge of using it for scouting purposes - which is what we are discussing here.  Also, Bill Belichick, the 2013 winner of the MIT Sloan Analytics conference Lifetime Achievement Award, is full of sh*t: 

Bill Belichick downplays analytics again, but Patriots know value of math

It certainly is a weird coincidence that Belichick “straight-up dismisses” analytics when basically every major decision the Patriots make, from player acquisition to contracts to in-game decision-making, is exactly what analytics would tell you to do. 

[The Voice of Reason]

22 minutes ago, Untouchable said:

Analytics are useful in baseball, not football.

Bill Belichick straight up dismisses them. That should tell you all you need to know.

I still don’t understand how the hell an analytics site like PFF grades players. 

I’ve said it numerous times, but 5-6 years ago they had Ben Hartsock graded as a Top 3 TE in the league despite not catching a single pass all season. I’d love for someone to explain that silly sh*t to me.

Ernie Adams, the man behind BB's success uses a ton of analytics so don't use that as a stat to die on 

[CTM]

Dont let perfect be the enemy of good.

[themeangreenkillingmachine]

I still don’t know what DVOA or DYAR stand for. 
how about at least telling us that?

[Eyes Of Adam]

12 minutes ago, CTM said:

Dont let perfect be the enemy of good.

The question is, if they cannot isolate an individuals performance, are they good for measuring the player? Im not sure we are going for perfect here, I am questioning are they even good.

No doubt they are a fantastic measure of the offense as a whole, but that is made up of lineman, WR's, backs, QB's and TE's, and all managed by an OC.

 

[Eyes Of Adam]

5 minutes ago, themeangreenkillingmachine said:

I still don’t know what DVOA or DYAR stand for. 
how about at least telling us that?

Defense-adjusted value over average

Defense-adjusted yards above replacement

[themeangreenkillingmachine]

3 minutes ago, Eyes Of Adam said:

Defense-adjusted value over average

Defense-adjusted yards above replacement

What’s replacement and how is everything adjusted? Sounds like a bunch of gibberish 

[Sonny Werblin]

In football, You evaluate with your eyes, not stats (traditional or advanced). Too many variables i.e. level of competition, talent of teammates, and play calling. 

[Eyes Of Adam]

10 minutes ago, themeangreenkillingmachine said:

What’s replacement and how is everything adjusted? Sounds like a bunch of gibberish 

I think its a very limited measure of an individual player, but a good measure of an offense overall. That is why I made this post, to have this type of discussion. Some are using the statistics as a way to measure the QB, but I don't feel it isolates the QB properly to do so. Just trying to get some intellectual discussion going

[CTM]

49 minutes ago, Eyes Of Adam said:

The question is, if they cannot isolate an individuals performance, are they good for measuring the player? Im not sure we are going for perfect here, I am questioning are they even good.

No doubt they are a fantastic measure of the offense as a whole, but that is made up of lineman, WR's, backs, QB's and TE's, and all managed by an OC.

 

Yes they are good and most importantly much better than traditional stats.

 

 

[ASH1962]

1 hour ago, CanadienJetsFan said:

My head hurts reading this.

I’m sure it’s helpful though.

Mine too

[SAM SAM HE'S OUR MAN]

1 hour ago, Eyes Of Adam said:

I see a lot of reference to DVOA and DYAR when it comes to Darnold vs other QB's. I think advanced stats are great, but they have to be used properly and you also have to understand their limitations their methodology.

I did some research to understand them a bit more, and right from the horses mouth comes the following:

ISSUES WITH DVOA/DYAR

DVOA is limited by what’s included in the official NFL play-by-play or tracked by the Football Outsiders game charting project. Because we need to have the entire play-by-play of a season in order to compute DVOA and DYAR, these metrics are not yet ready to compare players of today to players throughout the league’s history. As of this writing, we have processed 26 seasons, 1989 through 2014, and we add seasons at a rate of roughly two per year (the most recent season, plus one season back into history.)

Football is a game in which nearly every action requires the work of two or more teammates -- in fact, usually 11 teammates all working in unison. Unfortunately, when it comes to individual player ratings, we are still far from the point at which we can determine the value of a player independent from the performance of his teammates. That means that when we say, "In 2014, Marshawn Lynch had a DVOA of 23.1%, what we are really saying is “In 2014, Marshawn Lynch, playing in Darrell Bevell’s offensive system with the Seattle offensive line blocking for him and Russell Wilson selling the keeper when necessary, had a DVOA of 23.1%."

With fewer situations to measure, the numbers spread out a bit more, so you'll see more extreme DVOA ratings for part-time players and for measurements of teams in more specific situations (for example, passing on third downs). The charts listing players in order of DVOA have cut-offs for number of attempts, because players with just a handful of plays end up with absurd VOA and DVOA numbers. (In 2014, for instance, Johnny Manziel had a -73.2% passing DVOA in 38 plays.)

Passing statistics include sacks as well as fumbles on aborted snaps. Receiving statistics include all passes intended for the receiver in question, including those that are incomplete or intercepted. At some point, we hope to be able to determine just how much impact different receivers have on completes vs. incomplete passes, but various regression analyses make it clear that both quarterback and receiver have an impact on whether a pass is complete or not. The word passes refers to both complete and incomplete pass attempts.

Unless we say otherwise, all references to third down also include the handful of rushing and passing plays that take place on fourth down (primarily fourth-and-1).

 

The bold textz above is absolutely critical to understand when evaluating these advanced statistics. DVOA and DYAR do not isolate the player from the team or the system. They are team and system statistics far above being independent player statistics.

I would love to get some discussion going with those who use this stat as an individual statistic, so I can understand better their logic. Advanced stats are really cool and play a big part in sports.

[#27TheDominator]

1 minute ago, CTM said:

Yes they are good and most importantly much better than traditional stats.

 

 

I am no expert on these stats and don't even look most up, but I think one of the issues is the +/- way they grade.  I think PFF gives a grade between +2/-2.  If you do your job you get a positive grade. I think that is how a guy like Hartsock got a  good grade.  It is also how he hung around the league.  They didn't ask him to do much - but he always did it.  The issue is where guys like Robby Anderson or John Brown may miss blocks or run routes poorly, but if they get behind the D it is 6.  That is still only worth 2 points?  I think that is why they start dividing the grading to pass plays, under pressure, when targeted, etc. 

[Miss Lonelyhearts]

5 minutes ago, #27TheDominator said:

I am no expert on these stats and don't even look most up, but I think one of the issues is the +/- way they grade.  I think PFF gives a grade between +2/-2.  If you do your job you get a positive grade. I think that is how a guy like Hartsock got a  good grade.  It is also how he hung around the league.  They didn't ask him to do much - but he always did it.  The issue is where guys like Robby Anderson or John Brown may miss blocks or run routes poorly, but if they get behind the D it is 6.  That is still only worth 2 points?  I think that is why they start dividing the grading to pass plays, under pressure, when targeted, etc. 

And even that isn't a fix. Adjusting for situation works when the inputs are stats but not when it's charting data. If anything it only exacerbates the fact that the numbers are pseudo-quantitative in the first place.

[Eyes Of Adam]

29 minutes ago, CTM said:

Yes they are good and most importantly much better than traditional stats.

 

 

Traditional stats being not good, don't make DVOA or DYAR good. So in your opinion, do they isolate the QB's play from the rest of the team/system effectively?

[Apache 51]

1st downs, and touchdowns. The only things that matter. 

[CTM]

Just now, Eyes Of Adam said:

Traditional stats being not good, don't make DVOA or DYAR good. So in your opinion, do they isolate the QB's play from the rest of the team/system effectively?

I said it in my first post. If your goal here is to point out that it's not perfect, I already conceded that. Just because it's not perfect doesn't mean it's not worthwhile.

 It's better than what we had specifically because it adjusts for opponent and has a more appropriate definition of success as well as context.

It's also better than the "eye test" and is probably the best measurement we have today.

[talisaynon]

Bill Bellichick is a troll and would never let anyone in on his secret sauce for evaluating players. Lets be real they're a Boston sports team so they're going to be heavy into analytics.

[Losmeister]

1 hour ago, themeangreenkillingmachine said:

I still don’t know what DVOA or DYAR stand for. 
how about at least telling us that?

dont be lazy. go to football oustiders and read the damned glossary. 

[Losmeister]

1 hour ago, Eyes Of Adam said:

I think its a very limited measure of an individual player, but a good measure of an offense overall. That is why I made this post, to have this type of discussion. Some are using the statistics as a way to measure the QB, but I don't feel it isolates the QB properly to do so. Just trying to get some intellectual discussion going

it's imperfect...   but tries to be non-biased. fairly useful, but limited.

[T0mShane]

DVOA DYAR

DVED DY ARD

DWEDG TY ERD

DWEDGI TH NERD

WEDGIE THIS NERD

[Warfish]

3 hours ago, Untouchable said:

Analytics are useful in baseball, not football.

Remind everyone of this come draft time when the usual workout warriors get picked way ahead of legit performers based purely on their non-football testing analytics.

3 hours ago, Untouchable said:

Bill Belichick straight up dismisses them. That should tell you all you need to know.

It tells me you are naive and will believe anything you hear.

3 hours ago, Untouchable said:

I still don’t understand how the hell an analytics site like PFF grades players.

I bet you don't understand how a nuclear reactor works either.  Doesn't mean the electricity generated is fake, does it?

3 hours ago, Untouchable said:

I’ve said it numerous times, but 5-6 years ago they had Ben Hartsock graded as a Top 3 TE in the league despite not catching a single pass all season. I’d love for someone to explain that silly sh*t to me.

I hear TE's do more than just catch passes......

I'm not going to defend every half-ass analysis, but ignoring analytics completely is just so lolworthy.

[Larz]

Was he the guy on Tatooine ?

[nycdan]

They try to account for as many variables as they can.  I believe they watch every play and score it based on what the player did, what his teammates did to help or not help, what the opponent did, and what the play call was.  So if your WR drops an easy pass, the QB doesn't get dinged for it the same way he would if it was a Luke Falk special.  Conversely, if the RB matador's you right into the QB, you don't get the same DVOA juice you would get if you put a sick spin-move on him and earned your way to the QB.  Stuff like that.

It's probably got a lot of opportunity to criticize because there is subjective mixed in with the objective, but on a broad scope over a season, it's probably decent at valuing players.  

However, you are grading players based on the system they were in, the team they were on, the coach they had calling plays, the teams they played, the weather for each play, and his injuries or lack thereof.  That's a lot of variables that aren't the same from year to year, so a great player last year on the DVAR metric might be very average this year.  But on the whole, the top-20 players are probably going to perform better than the next 20 players.  

Take it for what it is.  It's a tool (yes, created by tools).  Useful, but you can't use it to replace your own judgment and knowledge of the team and system you are trying to find players for.  

[Matt39]

Any of the QB's who are ahead of Darnold in DVOA/DYAR not passing the eye test? His ranking to date is pretty in line with his performance while weighted against the other 31 QB's. He's had 3 meltdown games essentially (one huge meltdown) which he's now making up for against soft opponents. The guys ahead of Darnold have been better quarterbacks. Jets fans seem to have a tendency to not watch other teams it seems.

[CTM]

1 minute ago, Matt39 said:

Any of the QB's who are ahead of Darnold in DVOA/DYAR not passing the eye test? His ranking to date is pretty in line with his performance while weighted against the other 31 QB's. He's had 3 meltdown games essentially (one huge meltdown) which he's now making up for against soft opponents. The guys ahead of Darnold have been better quarterbacks. Jets fans seem to have a tendency to not watch other teams it seems.

Not watching other teams intently causes you to not realze the entire league has seen a passing frenzy. The frame of reference for many fans is Sanchez, Geno, etc.. so Darnold looks like a stud in comparison.  +some recency bias obv

[Miss Lonelyhearts]

17 minutes ago, nycdan said:

They try to account for as many variables as they can.  I believe they watch every play and score it based on what the player did, what his teammates did to help or not help, what the opponent did, and what the play call was.  So if your WR drops an easy pass, the QB doesn't get dinged for it the same way he would if it was a Luke Falk special.  Conversely, if the RB matador's you right into the QB, you don't get the same DVOA juice you would get if you put a sick spin-move on him and earned your way to the QB.  Stuff like that.

It's probably got a lot of opportunity to criticize because there is subjective mixed in with the objective, but on a broad scope over a season, it's probably decent at valuing players.  

However, you are grading players based on the system they were in, the team they were on, the coach they had calling plays, the teams they played, the weather for each play, and his injuries or lack thereof.  That's a lot of variables that aren't the same from year to year, so a great player last year on the DVAR metric might be very average this year.  But on the whole, the top-20 players are probably going to perform better than the next 20 players.  

Take it for what it is.  It's a tool (yes, created by tools).  Useful, but you can't use it to replace your own judgment and knowledge of the team and system you are trying to find players for.  

There's no subjectivity in FO numbers. The adjustments for situation and opposition all come from stats. They do keep track of things like dropped interceptions and do separate analysis on those but they don't factor in DVOA.

[Untouchable]

1 hour ago, Warfish said:

 

I hear TE's do more than just catch passes......

I'm not going to defend every half-ass analysis, but ignoring analytics completely is just so lolworthy.

Yeah, like blocking...

And I don’t give two sh*ts how good of an inline blocker you are at the position, there isn’t a metric in existence that should equate to a player being a Top 3 TE in the league without catching a single pass in a 16 game season.

None

I don’t need to break down some mathematical equation to realize that’s completely nonsensical.

[Losmeister]

6 minutes ago, Untouchable said:

there isn’t a metric in existence that should equate to a player being a Top 3 TE in the league without catching a single pass in a 16 game season.

it's a legit argument,

i still wonder how Maye gets such elevated ratings w/o INts and only a few PDs...   shrug

[CTM]

18 minutes ago, Miss Lonelyhearts said:

There's no subjectivity in FO numbers. The adjustments for situation and opposition all come from stats. They do keep track of things like dropped interceptions and do separate analysis on those but they don't factor in DVOA.

80 was saying they adjust for strength of teammates. I know they do that in Qbase but I don't think they do anything like that in DVOA and was too lazy to look. Maybe you know off top of head?