Josh Allen: Analytics

[Billsflyer12]

As a guy whose mind is wired for data and numbers, I certainly appreciate that not all minds think the same way.  There are times I wish my mind didn't.

 

Maybe the best example I can think of is buying a house.  You find a house you love, most people then look at price, comps, Sq Ft, school rankings and many others pieces of data that paints a picture.  At end of the day some use that data to form a risk analysis that tells them weather it is a smart investment.  Others will just look at the house they love and say who cares, we can fix the house up and help fix up the neighborhood, we can make it better.

 

I guess neither is right and wrong.  I prefer to use data, to form a picture that increases my probability of success.  It is not 100%, like any data there are outliers, I hope Josh Allen is one of them.  

[Stank_Nasty]

7 minutes ago, GoBills808 said:

He's not a world beater but for me you aren't giving him enough credit here. Flacco's never had much to rely on as far as weapons go, I think the average age of his #1 wideout over the course of his career is something pretty ridiculous, and the Ravens' philosophy has been run-heavy for some time now. I'd be very satisfied if we peg Allen to Flacco with upside.

in flacco's ten years he's averaged 3500 yds, 20 td and 13 int a season with a career 84 rating..... in todays nfl if that's what we get out of allen then we better pray we have a top 5 defense and run game every year.

 

flacco has been extremely average with the luxury of some consistently awesome defenses over his career.

[GoBills808]

3 minutes ago, Stank_Nasty said:

in flacco's ten years he's averaged 3500 yds, 20 td and 13 int a season with a career 84 rating..... in todays nfl if that's what we get out of allen then we better pray we have a top 5 defense and run game every year.

 

flacco has been extremely average with the luxury of some consistently awesome defenses over his career.

Yes, those are his numbers. I've also watched him play quite a bit and I'm saying he's better than you're giving him credit for. He hasn't had much to work with over the course of his career.

[Sky Diver]

What does the Ouija board have to say?

[GoBills808]

36 minutes ago, BuffaloHokie13 said:

A more mobile Flacco was one of the comps I arrived at.

The issue with Flacco is sometime over the last 2-3 seasons he stopped trusting his protection and got nervous in the pocket. 

[Stanley Lombardi]

Analytics is what Pro Football Focus does:  You watch a game and think someone played great, but analytics says they were bad.

Then you watch a game and think someone played bad, but analytics says they performed well.  

 

Analytics is the quack science which teaches us how to doubt our lyin' eyes.

[Mat68]

3 hours ago, Fadingpain said:

I just started a thread on this topic; listen to Eric Galko from yesterday on WGR-550.  Maybe that was what you were thinking of.

 

For example, since 1999, no college QB with fewer than 30 starts and a completion % in college below 60% has ever amounted to jack at the NFL level.

 

Allen had 27 starts.  Completion percentage was 56.2% for his career.

 

There are mathematical models for predicting college QB success in the NFL.  The people who actually study that stuff/compute that stuff/know of that stuff almost universally do not think Allen will amount to anything in the NFL.

The people who are prone to ignoring it are the people who are ignorant of it to begin with, or place little value in it, i.e., ex football players.

 

Allen COULD be a success, of course, but the data/historical record predicts he will not be successful.

 

He's an unbelievably high risk/reward pick.

 

 

 

 

6th rounder and one of the greatest statistical anomalies in sports history.

 

You basing your draft strategy off crazy 1 in 10 million lighting strikes?  Let me know how that works out for you.

 

Much smarter, more predictable, and more reliable to go where the data leads you.

 

 

The problem is their are outliers.  Guys with crazy arm talent breaks alot of those constructs.  Stafford, Cutler, Newton, Marino, and Farve are all guys with crazy natural abilities who all fall out of those peremiters.  Completion % is one aspect of the overall analytic analysis.  Allen falls to Farve and Marino in thats aspect.  Now here is where party lines form.  You cant compare a player in 2018 to a player in 1992 and 83.  Allen is closer to those players in college systems, supporting casts, and development.  

 

If you want to dive deep into the analytics.  Allen was one of the best prospect when throwing from a clean pocket.  He also saw the least amounts of them.  Production wise, Allen had the highest percent of air yardage.  Over 60% of his yards were through the air.  In comparison mayfield was 39%.  

[Stank_Nasty]

On ‎5‎/‎3‎/‎2018 at 5:38 PM, GoBills808 said:

Yes, those are his numbers. I've also watched him play quite a bit and I'm saying he's better than you're giving him credit for. He hasn't had much to work with over the course of his career.

agree to disagree I guess.... I've watched him play plenty and saw a middling qb that could make some timely clutch throws on occasion but usually didn't win if he didn't have his defense clicking on all cylinders. 

 

i'm not big on that going forward if that's what we are getting.

 

EDIT: This convo here is about flacco. 

Edited May 11, 2018 by Stank_Nasty

[BuffaloHokie13]

18 minutes ago, GoBills808 said:

The issue with Flacco is sometime over the last 2-3 seasons he stopped trusting his protection and got nervous in the pocket. 

Well, and he's always been a better playoff QB than regular season QB. More mobile Flacco is my 'realistic expectation' of Josh Allen. If that's what I get I'm not upset. If he's better that's great! If he's not then he's a bust.

[OrtonHearsaWho]

5 minutes ago, Stank_Nasty said:

agree to disagree I guess.... I've watched him play plenty and saw a middling qb that could make some timely clutch throws on occasion but usually didn't win if he didn't have his defense clicking on all cylinders. 

 

i'm not big on that going forward if that's what we are getting.

 

I'm with you.  He as entirely too many games that he looks absolutely awful.  I know you can't take away the four game playoff stretch (should have been two games if Denver could just not give up a 70 yard TD)...but IF you could, then who is he? 

[HappyDays]

39 minutes ago, GoBills808 said:

Yes, those are his numbers. I've also watched him play quite a bit and I'm saying he's better than you're giving him credit for. He hasn't had much to work with over the course of his career.

 

Flacco is not a positive comparison. Ravens fans despise him and his ridiculous contract. He caught fire at exactly the right time which is cool and all but I'm aiming for more. They just traded back into the 1st round for a QB! They are trying to do better there.

[Wayne Arnold]

The 4:00 minute mark is the epitome of Josh Allen's 2017 season at Wyoming.

 

 

[GoBills808]

6 minutes ago, HappyDays said:

 

Flacco is not a positive comparison. Ravens fans despise him and his ridiculous contract. He caught fire at exactly the right time which is cool and all but I'm aiming for more. They just traded back into the 1st round for a QB! They are trying to do better there.

Sure. He's also 33 and coming off injury concerns. I don't dispute that you should be looking for better, but if you have him as Allen's baseline I think it's a positive.

[Stank_Nasty]

50 minutes ago, Wayne Arnold said:

 

Took over a 2-10 team and led Wyoming to back-to-back 8 wins seasons for the first time in 20 years and just the sixth time in the history of the 126 year old program.

 

Middling, tho.

uh... dude. 808 and I were discussing flacco... i'm cool with allen.

37 minutes ago, HappyDays said:

 

Flacco is not a positive comparison. Ravens fans despise him and his ridiculous contract. He caught fire at exactly the right time which is cool and all but I'm aiming for more. They just traded back into the 1st round for a QB! They are trying to do better there.

seriously.... did anyone actually think that was a good deal? I feel like even ozzie newsome was like "crap. this guy just screwed around and won us a super bowl and I know he's not worth this sort of money but i'll look like a complete jerk if I don't pay him"

29 minutes ago, GoBills808 said:

Sure. He's also 33 and coming off injury concerns. I don't dispute that you should be looking for better, but if you have him as Allen's baseline I think it's a positive.

I get you on the bolded. fair enough.... and honestly he had some nice years from like 2009-2012. but I would even argue his best seasons weren't even top ten type of years.

 

the guy got really really really lucky against Denver when Rahim moore went full pop warner on a horribly thrown pass back in 2012.

Edited May 3, 2018 by Stank_Nasty

[Wayne Arnold]

12 minutes ago, Stank_Nasty said:

uh... dude. 808 and I were discussing flacco... i'm cool with allen.

 

[joevan]

I will take strong, big, smart over weaker, smaller dumber players.

[Hapless Bills Fan]

 

4 hours ago, Fadingpain said:

For example, since 1999, no college QB with fewer than 30 starts and a completion % in college below 60% has ever amounted to jack at the NFL level.

(...)

There are mathematical models for predicting college QB success in the NFL.  The people who actually study that stuff/compute that stuff/know of that stuff almost universally do not think Allen will amount to anything in the NFL.

The people who are prone to ignoring it are the people who are ignorant of it to begin with, or place little value in it, i.e., ex football players.

 

4 hours ago, TigerJ said:

It's statistical analysis.  Of course, it's a lot more detailed than college completion average, but it's still an analysis of a current player based on his college statistics as compared with the college statistics of past QBs.  That is what gives you the overwhelming odds that Allen will bust.

 

Hokay, I'll try this on.  What all these groups of number cruncher are doing, at heart, is trying to build a predictive model for NFL QB success.  And I have my "number cruncher" DNA, so I appreciate what they're trying to do.

 

BUT, the very first thing you have to do when you're looking at a predictive model, is ensure that the situation where you apply your model, matches the test dataset from which you derived it.  A simple example would be clinical trials for the safety of a new medicine conducted exclusively on young men 18-25 y.o., but now you want to apply these results to children <12, women, and men >50.  The safety results may or may not be valid in these groups, because they have some significant metabolic differences and they weren't part of the dataset you modeled from.

 

So here's the $21,481,462 question: does Josh Allen in fact match the test dataset from which all these analytics on QB success or failure were derived?

 

3 hours ago, billspro said:

Allen is a huge risk, no doubt. I had trouble evaluating him in the predraft process because he is the most talented QB I have ever seen. His 2017 tape was not good, but was also tough to evaluate because his WRs were not open and his line was brutal. Off of the tape I had him as the 6th ranked QB in this class. So I would not be shocked if he busts, but I would also not be surprised if he is a top 5 QB in the NFL. I don’t think there are any prospects comparable to him in the last 25 years. He will be an interesting case to see if someone can develop at the NFL level. It is rarely done, but if anyone is going to do it Allen will be the guy.

 

And this captures the essence of my point.  If in fact, the talent level of the conference he played in, the offense his team used, the minimal amount of QB coaching he's had so far, etc etc mean he's outside the model, the model predictions are relatively meaningless.  Maybe Beane and the coaches did a deep dive and narrowed the dataset down and did their own predictive model that says

 

Or maybe they didn't.  Maybe they fell in love with his "face of the franchise" blue-collar well-spoken persona and howitzer arm, and said "damn the statistics".

 

In the latter case of course, I would like to play poker with Brandon Beane.  Which is not a good thing, because a GM should be master of playing the odds.

 

[cba fan]

6 hours ago, Billsflyer12 said:

If true, he is terrible at it.

That shows how far the play went. NOT how far the QB threw the ball in the air.

 

My point is Allen throws the ball much much further in the air.

 

Also the following graph taken off same site your comp % was on shows Rosen and Allen are about the same. Rosen does better on 4 yrd passes and that is about it.

Edited May 4, 2018 by cba fan
add

[1st&ten]

6 hours ago, Soda Popinski said:

I'll just defer to this FO knowing a lot more about this than I ever will.   Daboll was with the Patriots and Alabama, he's worked under Saban/BB.   Beane and McDermott have had success before through drafting.  Last year's draft was pretty good considering we got a LT in the 2nd round, got a runner up candidate for DROY and hopefully a WR who can make a major contribution this season.   I love the Edmunds/Phillips picks.   In fact i love pretty much every pick other than Allen, so I think the issue is me not knowing more than them picking the wrong guy.   

I remember you from the former Buffalo Bills board. Welcome aboard.

[billspro]

2 hours ago, Hapless Bills Fan said:

 

 

 

Hokay, I'll try this on.  What all these groups of number cruncher are doing, at heart, is trying to build a predictive model for NFL QB success.  And I have my "number cruncher" DNA, so I appreciate what they're trying to do.

 

BUT, the very first thing you have to do when you're looking at a predictive model, is ensure that the situation where you apply your model, matches the test dataset from which you derived it.  A simple example would be clinical trials for the safety of a new medicine conducted exclusively on young men 18-25 y.o., but now you want to apply these results to children <12, women, and men >50.  The safety results may or may not be valid in these groups, because they have some significant metabolic differences and they weren't part of the dataset you modeled from.

 

So here's the $21,481,462 question: does Josh Allen in fact match the test dataset from which all these analytics on QB success or failure were derived?

 

 

And this captures the essence of my point.  If in fact, the talent level of the conference he played in, the offense his team used, the minimal amount of QB coaching he's had so far, etc etc mean he's outside the model, the model predictions are relatively meaningless.  Maybe Beane and the coaches did a deep dive and narrowed the dataset down and did their own predictive model that says

 

Or maybe they didn't.  Maybe they fell in love with his "face of the franchise" blue-collar well-spoken persona and howitzer arm, and said "damn the statistics".

 

In the latter case of course, I would like to play poker with Brandon Beane.  Which is not a good thing, because a GM should be master of playing the odds.

 

 

Exactly let’s hope we have that special player that is an outlier. 

 

That is all we can do at this point.

Edited May 4, 2018 by billspro