🦅🤠 NFC East 🤷‍♂️🆘 | Celebrating the Cowboys' 28 Years of Sadness

[HTTRDynasty]

Just now, Gmen said:

I'm just presenting the data.  So far you've disregarded CPOE, EPA, On-Target%, QBR, TDs, and picked the only statistic that favors Haskins - Total Points per play. 

As for the WRs not winning enough, I recall seeing a statistic that showed a big discrepancy in Mclaurin's production when thrown to by Keenum vs Haskins.  Which is ironic considering they played together in college too.

What you call "disregarding" is actually emphasizing the metrics (Total Points, PFF) that have the greatest explanatory power, as they adjust for context much more than the others.  Not coincidentally, they also have the highest year-to-year correlation.

[Gmen]

Just now, HTTRDynasty said:

What you call "disregarding" is actually emphasizing the metrics (Total Points, PFF) that have the greatest explanatory power, as they adjust for context much more than the others.  Not coincidentally, they also have the highest year-to-year correlation.

I recall CPOE + EPA being higher that Total points on that tweet you posted by Ben Baldwin.

And Just like with EPA, Jones' Total points is negatively affected by the fumbles.  As an actual thrower (measured by CPOE and On-Target%), Jones is head and shoulders above Haskins.

[EagleBlueDon]

Meh, I've seen or heard enough execs and higher ups state that they'd take Wentz over Dak with the same stable of weapons. However, I've never seen or heard anybody say they'd take Dak over Wentz with the same stable of weapons.

[HTTRDynasty]

2 minutes ago, Gmen said:

I recall CPOE + EPA being higher that Total points on that tweet you posted by Ben Baldwin.

And Just like with EPA, Jones' Total points is negatively affected by the fumbles.  As an actual thrower (measured by CPOE and On-Target%), Jones is head and shoulders above Haskins.

Ignore the title, as this is from 2016-2019:

 

 

Again, CPOE doesn't account for drops or pressure on the QB, forcing an errant throw.  Total Points and PFF do.

And if you're trying to wave away Jones' fumbles as if avoiding them isn't an integral part of playing QB, I may as well wave away Haskins' sacks, which significantly improves his Total Points efficiency:

Quote

 

Furthermore, sacks really hampered Haskins’ efficiency from a Total Points perspective. He endured the second-highest sack rate (11.9%) among quarterbacks with at least 200 dropbacks. Although quarterbacks do contribute to their own sack numbers, Haskins looks much better when you examine just his attempts.

After removing sacks, his 17.9 passing Total Points Added/100 Dropbacks ranks 18th. That put him just below Kyler Murray (18.3, 17th) and above all of the following: Tom Brady (16.5, 22nd), Gardner Minshew (16.1, 23rd), Josh Allen (15.3, 24th), Daniel Jones (14.0, 25th), Sam Darnold (13.6, 26th), and Baker Mayfield (10.0, 31st). It was also better than Drew Lock’s 13.6, but Lock was a nonqualifier.

 

 

 

[EagleBlueDon]

Honestly, which collection of talent would you rather have on your team?

Group A: Gerald McCoy, Dontari Poe, Haha Clinton-Dix, Maurice Canady, Blake Bell, Greg Zeurlein

Group B: Jeff Heath, Travis Frederick, Byron Jones, Robert Quinn, Maliek Collins, Randall Cobb

Let's revisit this

[Gmen]

10 minutes ago, HTTRDynasty said:

Again, CPOE doesn't account for drops or pressure on the QB, forcing an errant throw.

Next Gen Stats CPOE does account for it.  So does my model.  And so will the new model Ben Baldwin is working on.

[HTTRDynasty]

8 minutes ago, Gmen said:

Next Gen Stats CPOE does account for it.  So does my model.  And so will the new model Ben Baldwin is working on.

How so?  

Regardless, I've acknowledged multiple times on this site that Haskins needs to clean up his footwork, which leads to too many inaccurate throws, which is the main reason we hired Ken Zampese this offseason.

But I still contend that Total Points and PFF have more explanatory power as it relates to a QB's overall performance, which is what matters most when determining who the better QB was.

[Matts4313]

4 hours ago, Danger said:

It's too bad Dak came up short in every big moment last year, then maybe he'd have an argument for the money he's seeking.

You arent really pinning those losses all on Dak, are you? 

1 hour ago, EagleBlueDon said:

Too bad weaponry can't be included in these studies for the sake of the Wentz v Dak argument. That sticks out like a sore thumb in that debate. It's impossible though.

Over their careers Wentz has had a better OL/TEs and near same WRs (until last year). 

38 minutes ago, EagleBlueDon said:

Meh, I've seen or heard enough execs and higher ups state that they'd take Wentz over Dak with the same stable of weapons. However, I've never seen or heard anybody say they'd take Dak over Wentz with the same stable of weapons.

lol, quit lying

[Gmen]

16 minutes ago, HTTRDynasty said:

How so?  

Regardless, I've acknowledged multiple times on this site that Haskins needs to clean up his footwork, which leads to too many inaccurate throws, which is the main reason we hired Ken Zampese this offseason.

But I still contend that Total Points and PFF have more explanatory power as it relates to a QB's overall performance, which is what matters most when determining who the better QB was.

Next Gen Stats uses tracking data to see how close a defender is to the QB at the time of throw.  Here is a link if you want to read more

https://www.nfl.com/news/next-gen-stats-introduction-to-completion-probability-0ap3000000964655

nflfastR (the database made by Baldwin that I used for my CPOE model), lists whether the QB was hit on a throw.  When a QB is hit on a throw, the probability of him completing the pass is much lower, so that is used as a feature in the model.  In fact, aside from throw distance, it's the most important feature for CPOE modeling.

 

Edited May 28, 2020 by Gmen

[HTTRDynasty]

1 minute ago, Gmen said:

Next Gen Stats uses tracking data to see how close a defender is to the QB at the time of throw.  Here is a link if you want to read more

https://www.nfl.com/news/next-gen-stats-introduction-to-completion-probability-0ap3000000964655

nflfastR (the database made by Baldwin that I used for my CPOE model), lists whether the QB was hit on a throw.  When a QB is hit on a throw, the probability of him completing the pass is much lower, so that is used as a feature in the model.  In fact, aside from throw distance, it's the most important feature for CPOE modeling.

Does it account for drops too?

[Gmen]

Just now, HTTRDynasty said:

Does it account for drops too?

nflfastR doesn't have drops information, so my model does not.  I don't know about Next Gen Stats. 

But I feel like drops can be subjective.  There are blatant drops, which are easy.  But if a receiver is stretched out and gets 7 fingers on a ball, is that a drop? 

[HTTRDynasty]

1 minute ago, Gmen said:

nflfastR doesn't have drops information, so my model does not.  I don't know about Next Gen Stats. 

But I feel like drops can be subjective.  There are blatant drops, which are easy.  But if a receiver is stretched out and gets 7 fingers on a ball, is that a drop? 

I just read the article you linked.  NextGenStats doesn't include drops.

Drops can be subjective, but they have a significant impact on QB stats, especially when a QB's receivers drop passes at a high rate.  So they can't be ignored just because they can be subjective.  And there are exceptions, but most drops are obvious. 

That's part of the reason Total Points and PFF's metrics are superior in explaining what an individual player can actually be blamed for or given credit for over the course of a season.  These other metrics don't do nearly as good of a job with that.

[Gmen]

3 minutes ago, HTTRDynasty said:

I just read the article you linked.  NextGenStats doesn't include drops.

Drops can be subjective, but they have a significant impact on QB stats, especially when a QB's receivers drop passes at a high rate.  So they can't be ignored just because they can be subjective.  And there are exceptions, but most drops are obvious. 

That's part of the reason Total Points and PFF's metrics are superior in explaining what an individual player can actually be blamed for or given credit for over the course of a season.  These other metrics don't do nearly as good of a job with that.

The point of Next Gen Stats and nflfastR is to keep human subjectivity out of the loop.  It's all automated.  And there is no easy way to automatically determine a drop.

[HTTRDynasty]

1 minute ago, Gmen said:

The point of Next Gen Stats and nflfastR is to keep human subjectivity out of the loop.  It's all automated.  And there is no easy way to automatically determine a drop.

When two completely different "subjective" sources agree Haskins was much better last year than the "objective" stats indicate (and better than Jones), that's a red flag on the "objective" stats.

[Gmen]

4 minutes ago, HTTRDynasty said:

When two completely different "subjective" sources agree Haskins was much better last year than the "objective" stats indicate (and better than Jones), that's a red flag on the "objective" stats.

But even SIS On-Target% and QBDataMine, which are subjective stats that include drops, rank Haskins super low in terms of accuracy and agree with the CPOE models in that regard.