Jump to content

Football Outsiders DVOA Ratings


Coach Tuesday

Recommended Posts

8 minutes ago, ScottLaw said:

Fans forget this because they hate Tyrod for some strange reason. 

 

Nah, they forget because the offense was rarely exciting and the Bills were basically a running team.  And if we got behind we knew we weren't coming back. 

 

Nobody reminisces about the Rex Ryan Bills and says "wow, what a great offense that was."

 

 

Edited by eball
Link to comment
Share on other sites

14 hours ago, Coach Tuesday said:

https://www.footballoutsiders.com/dvoa-ratings/2020/week-3-dvoa-ratings
 

The Bills are a very weird, unexpected team so far this year. They have essentially been the exact opposite of what we expected. In Football Outsiders Almanac 2020, we wrote that the Bills might have the best roster in the league outside of the quarterback position. Their defense had our No. 1 projection for the preseason. So what has happened instead? Josh Allen has had a spectacular breakout, currently ranking second behind Russell Wilson in passing YAR. But the Bills rank 28th running the ball and a very surprising 24th on defense. They've scored a ton of points but keep letting opponents back into games. It's not just the blown 28-3 lead against the Rams; Miami was beating Buffalo until 6:03 of the fourth quarter last weekend. The rest of the team is not living up to what Josh Allen is doing so far.

Great, informative piece. Thanks for posting.

  • Like (+1) 1
Link to comment
Share on other sites

37 minutes ago, ScottLaw said:

That's revisionist history though because they got behind plenty. 

 

What's revisionist?  They scored a lot of points against bad teams and struggled vs. the good ones.  You really look back upon those years and yearn for that offense?

Link to comment
Share on other sites

7 hours ago, mannc said:

It was true when Rex was here, too.

The problem was we were screwed if we were behind and the opponent forced us to be one dimensional.  Roman's system and Tyrod's conservative style made it difficult to come back.

  • Like (+1) 1
Link to comment
Share on other sites

9 hours ago, Thurman#1 said:

 

 

Stats do indeed describe the past, not the future. So does the eye test. So does every possible form of intelligent looking at data and projection.

 

Predictions, forecasting, foretelling, prognostications and intelligent guesswork ... all of them share the problem that they can't look into the future and see what will happen before they predict it. So yeah, stats have the problem that they can only look at the past ... but they share that with every known form of data gathering. It's not a problem of stats, so much as a problem with the physics that prevent us from managing to travel in time.

 

You can guess about the future. That's actually all you can do about the future. You can make your guesswork as intelligent as possible, and that will absolutely mean looking at the past, including stats, trends, etc.

 

DVOA has proven itself as a good way of predicting. There is no perfect way of predicting and DVOA doesn't pretend to be one. And as Hapless pointed out, DVOA doesn't work as well early in the season (nor does any method of predicting, by the way), but it still has some predictive value from looking at preseason DVOA, which includes the last few games of the previous year, as they do. Put more specifically, it's their other stat, DAVE, which they use early in the season that has significant predictive value early. Like DVOA (and everything else), it's less predictive early than late. But it still has some value.

 

DVOA absolutely is a meaningful statistic, more meaningful than most. But it's far from perfect, and they're totally up front about that. The fact that they're willing to look back at times when DVOA has predicted more poorly reflects well on the openness of their minds and their unwillingness to be blinded by confirmation bias.

 

 

 

Thurman, you're an intelligent poster so let me ask...

 

Imagine that the top 50 posters at TBD predict the W-L record for every team and their results are combined.  

 

And then the statistical geeks at FO predicted the W-L record for every team.

 

Who would more accurately predict the season?  Personally, I think it would be roughly a toss-up.  But if I had to bet, I'd bet on TBD.  I think the TBD posters better appreciate the human element in the NFL.  Team X has a new offensive coordinator whose skill set matches the squad.  Team Y fired their lousy HC and replaced him with someone much more capable.  The Bills signed Stephon Diggs who was the missing piece for Allen & Daboll and creates mismatches that Allen can take advantage of.  

 

You say DVOA is meaningful but far from perfect.  DVOA can predict the outcome of games/seasons with some degree of accuracy that's better than chance.  But so can you and I - and everyone else here.    Given its imperfections, I'm not sure what DVOA brings to the table.  

 

Trivia.  Did you know that a 4 yard run on 1st and 10 is scored as a failure by FO while a 4 yard run on 2nd and 6 is scored as a success for DVOA purposes?  I'm not sure how many OCs would agree with this. 

 

Here's another scenario.  Your team is down by 2 points.   It's 3rd and 32 and time is running out.  You need 20 yards to get into FG position.  Things seem hopeless.  Then your QB fires a perfect 31 yard pass to a toe-tapping receiver.  Clock is stopped with plenty of time for the winning kick.  That pass is scored as a failure because by FO rules, any 3rd or 4th down play that doesn't pick up a first down is a failure  ("on third or fourth down, only gaining a new first down is considered success").

 

You say the FO folks are open-minded about their mistakes.  Maybe as they fine-tune their scoring system and algorithms, they'll get better at this.  

 

 

Link to comment
Share on other sites

38 minutes ago, hondo in seattle said:

 

Thurman, you're an intelligent poster so let me ask...

 

Imagine that the top 50 posters at TBD predict the W-L record for every team and their results are combined.  

 

And then the statistical geeks at FO predicted the W-L record for every team.

 

Who would more accurately predict the season?  Personally, I think it would be roughly a toss-up.  But if I had to bet, I'd bet on TBD.  I think the TBD posters better appreciate the human element in the NFL.  Team X has a new offensive coordinator whose skill set matches the squad.  Team Y fired their lousy HC and replaced him with someone much more capable.  The Bills signed Stephon Diggs who was the missing piece for Allen & Daboll and creates mismatches that Allen can take advantage of.  

 

You say DVOA is meaningful but far from perfect.  DVOA can predict the outcome of games/seasons with some degree of accuracy that's better than chance.  But so can you and I - and everyone else here.    Given its imperfections, I'm not sure what DVOA brings to the table.  

 

Trivia.  Did you know that a 4 yard run on 1st and 10 is scored as a failure by FO while a 4 yard run on 2nd and 6 is scored as a success for DVOA purposes?  I'm not sure how many OCs would agree with this. 

 

Here's another scenario.  Your team is down by 2 points.   It's 3rd and 32 and time is running out.  You need 20 yards to get into FG position.  Things seem hopeless.  Then your QB fires a perfect 31 yard pass to a toe-tapping receiver.  Clock is stopped with plenty of time for the winning kick.  That pass is scored as a failure because by FO rules, any 3rd or 4th down play that doesn't pick up a first down is a failure  ("on third or fourth down, only gaining a new first down is considered success").

 

You say the FO folks are open-minded about their mistakes.  Maybe as they fine-tune their scoring system and algorithms, they'll get better at this.  

 

 

 

I think you're misunderstanding what FO and DVOA does.  They admit that their preseason W/L predictions based on DVOA aren't very reliable, nor are the first few weeks of DVOA ratings.  This is because they don't - because they can't - take into account quality of opponent.  DVOA becomes much more accurate and useful as the season progresses, once the opponent adjustments are incorporated into the model.

 

And even then - DVOA isn't really designed to predict wins and losses.  To be sure, it projects a range of outcomes for wins and losses and is a pretty good model for doing so.  But where it's really useful is in betting and fantasy football.  It can help tell you whether to start a certain player against a certain defense, or whether a point spread is too high or low, etc.  If you use the model for those purposes you can reduce your risk and increase your odds of success.  Other than that, DVOA is good for evaluating whether a W/L record is a true measure of a team's success, for whatever that's worth.  You might have a 10 win season and would like to know, as a fan, whether it was a success.  DVOA can help you look behind the wins and losses - did you get lucky?  Should you really have won 14 games?  Etc.

Link to comment
Share on other sites

10 minutes ago, Coach Tuesday said:

 

I think you're misunderstanding what FO and DVOA does.  They admit that their preseason W/L predictions based on DVOA aren't very reliable, nor are the first few weeks of DVOA ratings.  This is because they don't - because they can't - take into account quality of opponent.  DVOA becomes much more accurate and useful as the season progresses, once the opponent adjustments are incorporated into the model.

 

And even then - DVOA isn't really designed to predict wins and losses.  To be sure, it projects a range of outcomes for wins and losses and is a pretty good model for doing so.  But where it's really useful is in betting and fantasy football.  It can help tell you whether to start a certain player against a certain defense, or whether a point spread is too high or low, etc.  If you use the model for those purposes you can reduce your risk and increase your odds of success.  Other than that, DVOA is good for evaluating whether a W/L record is a true measure of a team's success, for whatever that's worth.  You might have a 10 win season and would like to know, as a fan, whether it was a success.  DVOA can help you look behind the wins and losses - did you get lucky?  Should you really have won 14 games?  Etc.

 

Interesting point.

 

You say that as a season progresses, DVOA becomes more accurate.  But so do you and I.

 

Has their statistical modeling really surpassed the intellect of an informed fan?   

 

Maybe because I'm not into betting or fantasy football, I don't see their utility yet. 

Link to comment
Share on other sites

Just now, hondo in seattle said:

 

Interesting point.

 

You say that as a season progresses, DVOA becomes more accurate.  But so do you and I.

 

Has their statistical modeling really surpassed the intellect of an informed fan?   

 

Maybe because I'm not into betting or fantasy football, I don't see their utility yet. 


It has yes.  The average fan has trouble looking behind the stats and results to determine how well or poorly a team or individual player actually performed.  Your QB threw for 359 yds and a couple of TDs - great.   But DVOA, which takes into account the opponent as well as the game situations, can tell you how that QB did compared to an average replacement-level player, based on the various down and distances and overall opponent adjustments.  Your average fan may have a vague idea about that but DVOA can tell you within a much higher degree of reliability.

Link to comment
Share on other sites

4 hours ago, Coach Tuesday said:


It has yes.  The average fan has trouble looking behind the stats and results to determine how well or poorly a team or individual player actually performed.  Your QB threw for 359 yds and a couple of TDs - great.   But DVOA, which takes into account the opponent as well as the game situations, can tell you how that QB did compared to an average replacement-level player, based on the various down and distances and overall opponent adjustments.  Your average fan may have a vague idea about that but DVOA can tell you within a much higher degree of reliability.

 

I think you're providing a good defense of DVOA.  But I remain skeptical.  Scenarios like the ones I mention above still  bother me.   

 

And from what I understand, Shatz and FO don't actually break down game tape.  They rely entirely on box score data.  They take those numbers and contextualize them: down and distance, game clock, and so on.  There are advantages to that approach but issues as well.  The full context can't be quantified.  There's scheme, playcalls, officiating, full & partial injuries, momentum, leadership, and all sorts of other human variables that effect the (statistical) performance of a player and the outcome of a game.

 

While I think examining DVOA can provide some insight, I think the human brain is still more capable at evaluating the totality of a team's performance in all it's complexity.  

 

To put it another way, if I think Josh Allen has taken an important leap forward and Schatz still insists that he's a subpar QB, I'll trust my eyes before I trust Schatz's metrics.  

Link to comment
Share on other sites

On 9/29/2020 at 6:39 PM, Coach Tuesday said:

https://www.footballoutsiders.com/dvoa-ratings/2020/week-3-dvoa-ratings
 

The Bills are a very weird, unexpected team so far this year. They have essentially been the exact opposite of what we expected. In Football Outsiders Almanac 2020, we wrote that the Bills might have the best roster in the league outside of the quarterback position. Their defense had our No. 1 projection for the preseason. So what has happened instead? Josh Allen has had a spectacular breakout, currently ranking second behind Russell Wilson in passing YAR. But the Bills rank 28th running the ball and a very surprising 24th on defense. They've scored a ton of points but keep letting opponents back into games. It's not just the blown 28-3 lead against the Rams; Miami was beating Buffalo until 6:03 of the fourth quarter last weekend. The rest of the team is not living up to what Josh Allen is doing so far.

Also, according to our resident Jests representative @Jimmy Spagnola, they kicked our butts after the 1st quarter. Sar1 would be proud!

Link to comment
Share on other sites

8 hours ago, Coach Tuesday said:

 

I think you're misunderstanding what FO and DVOA does.  They admit that their preseason W/L predictions based on DVOA aren't very reliable, nor are the first few weeks of DVOA ratings.  This is because they don't - because they can't - take into account quality of opponent.  DVOA becomes much more accurate and useful as the season progresses, once the opponent adjustments are incorporated into the model.

 

And even then - DVOA isn't really designed to predict wins and losses.  To be sure, it projects a range of outcomes for wins and losses and is a pretty good model for doing so.  But where it's really useful is in betting and fantasy football.  It can help tell you whether to start a certain player against a certain defense, or whether a point spread is too high or low, etc.  If you use the model for those purposes you can reduce your risk and increase your odds of success.  Other than that, DVOA is good for evaluating whether a W/L record is a true measure of a team's success, for whatever that's worth.  You might have a 10 win season and would like to know, as a fan, whether it was a success.  DVOA can help you look behind the wins and losses - did you get lucky?  Should you really have won 14 games?  Etc.

I can tell you playing strictly DVOA against the spread is losing fwiw

Link to comment
Share on other sites

×
×
  • Create New...