[Vague Title] It continues... Josh Allen...

[YoloinOhio]

 

 

 

Well I was excited for the 1st day of camp!

[CommonCents]

3 minutes ago, YoloinOhio said:

 

 

 

Well I was excited for the 1st day of camp!

Is he trashing Allen or is he just linking what other people are saying? Or both?

[YoloinOhio]

Just now, Commonsense said:

Is he trashing Allen or is he just linking what other people are saying? Or both?

Having listened to him on the radio recently... both. 

[CommonCents]

Just now, YoloinOhio said:

Having listened to him on the radio recently... both. 

 They are spending the week trashing Belichick and harassing him about benching Butler, that’s a better listen than some blowhards torching Allen over and over again. 

 

Heres to hoping that Allen has a big set of balls and he is about to show everybody how they swing. 

[Billsman623]

9 minutes ago, Commonsense said:

Is he trashing Allen or is he just linking what other people are saying? Or both?

It is very obvious he has made his mind up on Allen and he is pouting because his golden boy Rosen was not the Bills selection. He has an agenda to smear Allen and when people call him out on it he gets all touchy and bent out of shape. It's gotten old and tiring. We get it, you don't like him. Some of us do and some of us are open minded. They are becoming very unlistenable already and it's day 1 of camp.

 

 

[Hapless Bills Fan]

59 minutes ago, YoloinOhio said:

 

 

 

Well I was excited for the 1st day of camp!

 

On the one hand, I hope Allen isn't paying any attention to this crap.

 

On the other hand, I hope he's making a scrapbook and it's lighting his fire.

 

 

[teef]

59 minutes ago, YoloinOhio said:

 

 

 

Well I was excited for the 1st day of camp!

ouch!

[GunnerBill]

19 hours ago, Bills Pimpin' said:

It's always more accurate to say a player is going to be a bust since 80% of them are (QB's are prolly closer to 90%). That's why most media D-bags and forum trolls claim bust. It is not because they are really smart or did any work to come up with their opinion. They are the bunch that cares only about claiming "I told ya so". In fact, in this specific case, even if Josh Allen is a serviceable QB for 10 years they will say " I told ya he wouldn't be a hall of famer" or "I told ya (QB X) would be better" or "I told ya we shouldn't have picked him 7th". It's really a miserable existence for them and especially kool aid drinking homers

 

Homers aren't all that great either (I tend to be one, I can't help it. I talked myself into Tyrod Taylor as a franchise QB and now I get to tell another fanbase he sucks), but it is a much better daily existence for sure.

 

I think realism is a much better existence. Don't come at any decision with a preconceived notion of them being right or wrong and just evaluate what you think of that decision. I promise not having to talk yourself into believing things that you don't deep down believe is a much less stressful way to live.  

[Hapless Bills Fan]

14 hours ago, Zerovotlz said:

The argument about Allen is this:  Given his physical tools and mental makeup, either A) he really does suck because we have a BIG SAMPLE size that says he does. or B) He has sucked so far because in his youth and college years, he never was around good coaching or other players that would push his development along and he won't suck anymore after he gets good coaching and has teammates and opponents who are better quality.

 

That's it.  I love metrics...I think metrics are really useful.....but it doesn't take advanced metrics to tell any reasonable person that Josh Allen's actual performance in football has been poor.  That includes the Bills staff.  

 

The Bills staff believes the answer is B....because you don't take him if you think it's A....and you also don't take him even if you believe it's B, but you aren't sure you can fix him.  They think he's undercoaced and underdeveloped.  

 

I don't think they can do it.  I would only say, in this case, where you may actually have a guy with a ton of talent, who just hasn't been developed properly for years...that could be something the analytics would miss because the analytics are analyzing high level football players, and makes no assumptions where they came from or how much football they played where, or against who.  This is where you'd not rely on a number spit out by a formula...but a human judgement.  Again..I don't think it will end well, but I can see a case here about why the analytic numbers may not apply.

 

This is a pretty good summary. 

 

The only thing I'll say is that the claim by Football Outsiders is that they do, in fact, factor in the talent level on the team and how much football they played where against whom.

And that's part of the problem some of us have with them.   They claim it's an objective formula - but let's look at what they say it is.

 

"QBASE favors quarterbacks expected to go high in the draft who also have a relatively long resume of college success according to the stats. Those stats include completion percentage, yards per attempt, and team passing efficiency. These numbers are adjusted both for the quality of the defenses that a prospect had to face as well as the quality of his offensive teammates. "

 

So let's see what goes into it.  https://www.footballoutsiders.com/stat-analysis/2015/introducing-qbase

-Completion percentage

-YPA

-Team Passing efficiency.  The NCAA formula is: [ { (8.4 * yards) + (330 * touchdowns) - (200 * interceptions) + (100 * completions) } / attempts ].
-Quality of defenses  "The strength of opposing defenses was measured in the same way as Pro-Football-Reference's Simple Rating System. "

-Quality of his offensive teammates "We measured teammate quality based on the draft value of offensive teammates in both the player's draft year and the following year."

(OK - but there's "not drafted" and there's "truly abysmal" and can this method distinguish?)

 

As stated, they are doubling down on completion percentage and YPA, and counting them twice.  As someone who once dealt with statistics in daily life, I don't like that.  Quality of defenses and Quality of offensive teammates are not objective stats kept by the NCAA, and it's not clear what they're doing.  Good luck if you try to figure it out.  The link above to "Simple Rating System" doesn't take me to anything that explains. 

 

Then apparently they're using the QBase number along with "College Experience" and "Projected draft slot" to conduct some sort of regression - again, details not explained.

So it sounds all objective and stuff (50,000 simulations), but that all depends upon the quailty of what they're putting in there.

 

Anyway, the real question: does it work?  If you go to the link above, they show how their model applies to a bunch of QB drafted between 1997 and 2010.

(Nothing I can find between 2011 and 2014, if anyone does, LMK...).  Note that their color coding is inconsistent - they flag Peyton Manning as someone their model under-predicted, but they don't flag RGIII as someone their model over-predicted even though he's overpredicted to the same degree as John Beck (another model failure).  Anyway, their data is there, chew it up if you like.

 

My initial chewing is as follows:

43% of the time their model correctly predicted the QB's career

40% of the time their model over-predicted the QB's success

22% of the time their model under-predicted the QB's success. 

 

Under-predictions included Peyton Manning, Aaron Rogers, Matt Stafford, Chad Pennington,Daunte Culpepper, Drew Brees, Matt Schaub, Matt Ryan, Brian Griese, and Josh McCown (for some of these, they did predict decent careers, just not Franchise Success)

 

Correctly predicting players gives them credit for Mike Vick and Alex Smith (they say bad, Vick had some good years esp Philly and Smith just got the Big Bucks), Vince Young and Bradford (they say good, Young as we know flamed out and Bradfordwas meh before getting sidelined by injuries)

 

If someone offered me a statistical package for my field that was correct 43% of the time, over-predicted 40% of the time, and under-predicted 22% of the time, I would not purchase their product or make business decisions based upon it, no matter how much hyperbole about "zero chance of success" and "nigerian princes" they laded it with.

 

 

 

 

[Figster]

42 minutes ago, Hapless Bills Fan said:

 

This is a pretty good summary. 

 

The only thing I'll say is that the claim by Football Outsiders is that they do, in fact, factor in the talent level on the team and how much football they played where against whom.

And that's part of the problem some of us have with them.   They claim it's an objective formula - but let's look at what they say it is.

 

"QBASE favors quarterbacks expected to go high in the draft who also have a relatively long resume of college success according to the stats. Those stats include completion percentage, yards per attempt, and team passing efficiency. These numbers are adjusted both for the quality of the defenses that a prospect had to face as well as the quality of his offensive teammates. "

 

So let's see what goes into it.  https://www.footballoutsiders.com/stat-analysis/2015/introducing-qbase

-Completion percentage

-YPA

-Team Passing efficiency.  The NCAA formula is: [ { (8.4 * yards) + (330 * touchdowns) - (200 * interceptions) + (100 * completions) } / attempts ].
-Quality of defenses  "The strength of opposing defenses was measured in the same way as Pro-Football-Reference's Simple Rating System. "

-Quality of his offensive teammates "We measured teammate quality based on the draft value of offensive teammates in both the player's draft year and the following year."

(OK - but there's "not drafted" and there's "truly abysmal" and can this method distinguish?)

 

As stated, they are doubling down on completion percentage and YPA, and counting them twice.  As someone who once dealt with statistics in daily life, I don't like that.  Quality of defenses and Quality of offensive teammates are not objective stats kept by the NCAA, and it's not clear what they're doing.  Good luck if you try to figure it out.  The link above to "Simple Rating System" doesn't take me to anything that explains. 

 

Then apparently they're using the QBase number along with "College Experience" and "Projected draft slot" to conduct some sort of regression - again, details not explained.

So it sounds all objective and stuff (50,000 simulations), but that all depends upon the quailty of what they're putting in there.

 

Anyway, the real question: does it work?  If you go to the link above, they show how their model applies to a bunch of QB drafted between 1997 and 2010.

(Nothing I can find between 2011 and 2014, if anyone does, LMK...).  Note that their color coding is inconsistent - they flag Peyton Manning as someone their model under-predicted, but they don't flag RGIII as someone their model over-predicted even though he's overpredicted to the same degree as John Beck (another model failure).  Anyway, their data is there, chew it up if you like.

 

My initial chewing is as follows:

43% of the time their model correctly predicted the QB's career

40% of the time their model over-predicted the QB's success

22% of the time their model under-predicted the QB's success. 

 

Under-predictions included Peyton Manning, Aaron Rogers, Matt Stafford, Chad Pennington,Daunte Culpepper, Drew Brees, Matt Schaub, Matt Ryan, Brian Griese, and Josh McCown (for some of these, they did predict decent careers, just not Franchise Success)

 

Correctly predicting players gives them credit for Mike Vick and Alex Smith (they say bad, Vick had some good years esp Philly and Smith just got the Big Bucks), Vince Young and Bradford (they say good, Young as we know flamed out and Bradfordwas meh before getting sidelined by injuries)

 

If someone offered me a statistical package for my field that was correct 43% of the time, over-predicted 40% of the time, and under-predicted 22% of the time, I would not purchase their product or make business decisions based upon it, no matter how much hyperbole about "zero chance of success" and "nigerian princes" they laded it with.

 

 

 

 

Nice post Hapless,

 

guesswork anyone?

[Cripple Creek]

4 minutes ago, Figster said:

Nice post Hapless,

You haven’t read that many words since 4th grade. Who are you trying to kid?

[Buddy Hix]

3 hours ago, YoloinOhio said:

 

 

 

Well I was excited for the 1st day of camp!

The needy Nigerian prince line....????

 

[Figster]

9 minutes ago, Cripple Creek said:

You haven’t read that many words since 4th grade. Who are you trying to kid?

Hapless brings out the best in me.

[Mikey152]

17 hours ago, Zerovotlz said:

 

...um...North Dakota State plays in an indoor stadium.  

 

The argument about Allen is this:  Given his physical tools and mental makeup, either A) he really does suck because we have a BIG SAMPLE size that says he does. or B) He has sucked so far because in his youth and college years, he never was around good coaching or other players that would push his development along and he won't suck anymore after he gets good coaching and has teammates and opponents who are better quality.

 

That's it.  I love metrics...I think metrics are really useful.....but it doesn't take advanced metrics to tell any reasonable person that Josh Allen's actual performance in football has been poor.  That includes the Bills staff.  

 

The Bills staff believes the answer is B....because you don't take him if you think it's A....and you also don't take him even if you believe it's B, but you aren't sure you can fix him.  They think he's undercoaced and underdeveloped.  

 

I don't think they can do it.  I would only say, in this case, where you may actually have a guy with a ton of talent, who just hasn't been developed properly for years...that could be something the analytics would miss because the analytics are analyzing high level football players, and makes no assumptions where they came from or how much football they played where, or against who.  This is where you'd not rely on a number spit out by a formula...but a human judgement.  Again..I don't think it will end well, but I can see a case here about why the analytic numbers may not apply.

 

This is a good post, and I generally agree with your points...but I think you are a little off (and it's not just you) when you say he was poor in college.  Inefficient, maybe, but not poor...it was clear that he was carrying his team, though...they were really bad before he got there and when he was hurt, and an 8 win bowl team when he played. 

 

Serious question:

 

Would you take Tyrod Taylor right now over Brett Favre in his prime?  Do you know who would?  Analytics...because their model rewards "efficiency" and has no measure for things like plays left on the field.  The year the Packers won the SB, Brett Favre completed 59.9% of his passes.

Edited July 26, 2018 by Mikey152

[oldmanfan]

1 hour ago, Hapless Bills Fan said:

 

This is a pretty good summary. 

 

The only thing I'll say is that the claim by Football Outsiders is that they do, in fact, factor in the talent level on the team and how much football they played where against whom.

And that's part of the problem some of us have with them.   They claim it's an objective formula - but let's look at what they say it is.

 

"QBASE favors quarterbacks expected to go high in the draft who also have a relatively long resume of college success according to the stats. Those stats include completion percentage, yards per attempt, and team passing efficiency. These numbers are adjusted both for the quality of the defenses that a prospect had to face as well as the quality of his offensive teammates. "

 

So let's see what goes into it.  https://www.footballoutsiders.com/stat-analysis/2015/introducing-qbase

-Completion percentage

-YPA

-Team Passing efficiency.  The NCAA formula is: [ { (8.4 * yards) + (330 * touchdowns) - (200 * interceptions) + (100 * completions) } / attempts ].
-Quality of defenses  "The strength of opposing defenses was measured in the same way as Pro-Football-Reference's Simple Rating System. "

-Quality of his offensive teammates "We measured teammate quality based on the draft value of offensive teammates in both the player's draft year and the following year."

(OK - but there's "not drafted" and there's "truly abysmal" and can this method distinguish?)

 

As stated, they are doubling down on completion percentage and YPA, and counting them twice.  As someone who once dealt with statistics in daily life, I don't like that.  Quality of defenses and Quality of offensive teammates are not objective stats kept by the NCAA, and it's not clear what they're doing.  Good luck if you try to figure it out.  The link above to "Simple Rating System" doesn't take me to anything that explains. 

 

Then apparently they're using the QBase number along with "College Experience" and "Projected draft slot" to conduct some sort of regression - again, details not explained.

So it sounds all objective and stuff (50,000 simulations), but that all depends upon the quailty of what they're putting in there.

 

Anyway, the real question: does it work?  If you go to the link above, they show how their model applies to a bunch of QB drafted between 1997 and 2010.

(Nothing I can find between 2011 and 2014, if anyone does, LMK...).  Note that their color coding is inconsistent - they flag Peyton Manning as someone their model under-predicted, but they don't flag RGIII as someone their model over-predicted even though he's overpredicted to the same degree as John Beck (another model failure).  Anyway, their data is there, chew it up if you like.

 

My initial chewing is as follows:

43% of the time their model correctly predicted the QB's career

40% of the time their model over-predicted the QB's success

22% of the time their model under-predicted the QB's success. 

 

Under-predictions included Peyton Manning, Aaron Rogers, Matt Stafford, Chad Pennington,Daunte Culpepper, Drew Brees, Matt Schaub, Matt Ryan, Brian Griese, and Josh McCown (for some of these, they did predict decent careers, just not Franchise Success)

 

Correctly predicting players gives them credit for Mike Vick and Alex Smith (they say bad, Vick had some good years esp Philly and Smith just got the Big Bucks), Vince Young and Bradford (they say good, Young as we know flamed out and Bradfordwas meh before getting sidelined by injuries)

 

If someone offered me a statistical package for my field that was correct 43% of the time, over-predicted 40% of the time, and under-predicted 22% of the time, I would not purchase their product or make business decisions based upon it, no matter how much hyperbole about "zero chance of success" and "nigerian princes" they laded it with.

 

 

 

 

Couldn't agree more.  It appears there may be some bias here; they think certain measures should be more important and thus put them into their model first, as oppose to actually assessing various factors to determine prospectively what factors have a more relevant basis for use. 

[TheElectricCompany]

Are there any CFB performance stats out there where J.A. was in the top 3 among 2018 prospects? 

I don't care if it's " 2/2 - 50+ yard throws off his back leg with 35 mph gusts". 

There has to be something. 

 

Edited July 26, 2018 by TheElectricCompany

[ToGoGo]

On 7/24/2018 at 1:48 PM, Hapless Bills Fan said:

 

Meh.  It's accurate reporting on the Football Outsiders take.

 

But all we really need to know about how Football Outsiders sees things is encapsulated in this sentence:

" Aaron Schatz of Football Outsiders said he'd rather have Tyrod Taylor. ".  Then turn to Pro Football Focus, which lists Taylor as the #12 NFL QB for 2017, above Marcus Mariota, Matt Stafford, and Dak Prescott, and far above Jared Goff, Kirk Cousins, Derek Carr. 

 

Any one here prefer to have one of those 6 as our QB over Tyrod Taylor?  I know how I vote.

 

The funny thing about all this to me is that I'm 100% a stats geek and here I am critiquing stats geeks.  But as a stats geek, I know that stats tell you something - not always what you think they're telling you.  You need to look carefully at which ones are meaningful, and which ones are correlated to winning.  All these groups try to roll and tumble a bunch of different stuff to boil complicated factors down into a single number, which means they've become critically dependent on the weighting of different factors (in the case of Football Outsiders, I believe they over-weight low INT and underweight passing yards per game, or at least don't penalize falling below a certain threshold there).

 

 

 

I have a data analysis background and I've been annoyed by the aura around advanced stats for many years now. Stats + intuition is key in any kind of data analysis you do otherwise you come up with more useless conclusions than true. These guys forgot about intuition a long time ago, and in many cases don't believe in intangibles at all. That's how you come up with a quote like that about Josh Allen. I'm willing to bet they didn't watch a second of game film. 

[Figster]

1 hour ago, Mikey152 said:

 

This is a good post, and I generally agree with your points...but I think you are a little off (and it's not just you) when you say he was poor in college.  Inefficient, maybe, but not poor...it was clear that he was carrying his team, though...they were really bad before he got there and when he was hurt, and an 8 win bowl team when he played. 

 

Serious question:

 

Would you take Tyrod Taylor right now over Brett Favre in his prime?  Do you know who would?  Analytics...because their model rewards "efficiency" and has no measure for things like plays left on the field.  The year the Packers won the SB, Brett Favre completed 59.9% of his passes.

The quality of a QB/teams Defense is a variable that plays a big part in everything the team does on Offense in my humble opinion.

 

At the college level for instance and I'll use Alabama by way of example. It probably wouldn't matter how good the QB that starts for the tide looks in every possible category because when all is said and done he's a product of the system/s. Playing with a lead vs coming from behind changes how a QB and teams O operates and how the opposition defends them. 

 

Taylor played well with a lead, but not so well from behind when Buffalo was forced out of their comfort zone running the football.

 

 

Edited July 26, 2018 by Figster

[Scorp83]

On 7/24/2018 at 9:29 PM, Bill_with_it said:

Wrong. Maybe to ignite you. Lamar has no more potential than Allen. Do you recall in OTAs they were lining him up at we? What legitimate QB does that? I mean a rookie should be lining up at quarterback  regularly, not like some veteran in a gadget play. You don’t really believe what you type do you? 

Yes I do! 

 

They line him up at WR ...quite sure they have some plays installed where he's playing decoy. If people paid attention, you'll know Lamar Jackson can't catch. But, people dont pay attention...which means teams are going to probably fall for the Jackson Wideout decoy play about 5 times this year. 

 

He's a QB...& all reports out of Baltimore is he has a great chance to win the job.

On 7/24/2018 at 1:03 PM, Gugny said:

 

Please try to stay on topic. 

Huh....yea...smh ... It's my topic to begin with...?

[Hapless Bills Fan]

2 hours ago, oldmanfan said:

Couldn't agree more.  It appears there may be some bias here; they think certain measures should be more important and thus put them into their model first, as oppose to actually assessing various factors to determine prospectively what factors have a more relevant basis for use. 

 

I mean, that's what everyone building a predictive multivariate model does. 

You start with an idea of what factors you think are important, based on your initial analysis.  It works, kind of, but not as well as you like.

So you slice and dice and roll in some other stuff, and come up with your chef d'oeuvre (not to be confused with the famous TBD rettata)  and see how that works.

 

Pretty much everyone has an inkling completion percentage matters for a QB, INTs matter, passing TDs matter, the quality of the competition matters, the quality of the team matters, so it's not as though they're using factors anyone would consider irrelevant.  It's more a matter of whether they've really found a winning formula.

 

My first roll through their actual results says "not too impressed here".  They indeed called a bunch correctly, but their failure rate is along the lines of the "bust probability" they came up with for Allen.

 

Oh - and one thing that struck me.  They were apparently using "predicted draft slot" as part of their regression model.  This is what they say: " We used mock drafts to project draft slot for this year's quarterbacks. "  OK - whose mock drafts?  Theirs?  Kipers?  Average of 10 different mock drafts? 

 

Unless it's their own mock draft, here's the paradox -  Football Outsiders is all in for predicting the demise of traditional, oldschool scouting as inferior and outdated relative to their wonderproduct of modern analytics.  But they are actually, in small print, encorporating it into their model.