Another QB Evaluation Statistic | FinHeaven - Miami Dolphins Forums

Another QB Evaluation Statistic

S

Shouright

☠️ Banned ☠️
Joined
May 18, 2004
Messages
15,051
Reaction score
18
Age
52
After chatting with Brian Burke from the Advanced NFL Stats website and bouncing some ideas off of him, I've decided to alter a bit the approach to the QB evaluation statistic I came up with in this thread:

http://www.finheaven.com/showthread.php?351306-A-New-QB-Wins-Added-Statistic&highlight=

Instead of using win percentage as the dependent variable in the analysis, which confounds it with other areas of team play (running game, defense, etc.), I decided to use WPA per game for QBs, which extracts "winning" play from the QB's play more exclusively.

I then did a stepwise regression analysis with 40 different variables associated with QB play. The result was that only two variables -- 1) DVOA, and 2) percentage of pass dropbacks in which QBs are pressured -- could be used to produce a model that accounted for the most variance in WPA per game, 76%.

DVOA has a positive relationship with WPA per game, and is explained here: http://www.footballoutsiders.com/info/methods

The percentage of passing dropbacks in which QBs experience pressure of course has a negative relationship with WPA per game.

The model is fairly elegant theoretically because it consists of a statistic (DVOA) that constitutes a wide array of other variables in the game, and another statistic (pressure on the QB) that theoretically interferes with quarterback play, and which isn't included in DVOA.

Using the regression coefficients in the model produces the following data for the 2013 regular season:

Player
"Winning Play"
Peyton Manning
3.76
Aaron Rodgers
2.55
Philip Rivers
2.53
Drew Brees
2.53
Nick Foles
2.52
Andy Dalton
1.86
Colin Kaepernick
1.72
Matthew Stafford
1.67
Tom Brady
1.64
Ben Roethlisberger
1.47
Tony Romo
1.47
Sam Bradford
1.12
Matt Ryan
1.08
Andrew Luck
1
Russell Wilson
0.99
Jay Cutler
0.96
Matt Cassel
0.93
Cam Newton
0.93
Alex Smith
0.86
Ryan Fitzpatrick
0.84
Carson Palmer
0.72
Ryan Tannehill
0.7
EJ Manuel
0.41
Chad Henne
0.4
Jason Campbell
0.29
Robert Griffin III
0.23
Joe Flacco
0.22
Mike Glennon
0.05
Matt Schaub
-0.2
Eli Manning
-0.28
Geno Smith
-0.52
Case Keenum
-0.71
Brandon Weeden
-0.74
AVERAGE
1
SD
1.04
TANNEHILL PERCENTILE
34.3

EDIT: This is what we get when I correct for the error I made the first time of failing to remove the negative value for pressured dropback percentage:

Player "Winning Play"
Nick Foles 0.7808
Russell Wilson 0.7804
Philip Rivers 0.7684
Matt Ryan 0.75
Peyton Manning 0.7216
Drew Brees 0.7128
Colin Kaepernick 0.7108
Jay Cutler 0.7036
Carson Palmer 0.7
Aaron Rodgers 0.6968
Tony Romo 0.6928
Andrew Luck 0.684
Mike Glennon 0.6836
Sam Bradford 0.6772
Tom Brady 0.6704
Cam Newton 0.6684
Ben Roethlisberger 0.6516
Case Keenum 0.6436
Matt Schaub 0.6356
Ryan Fitzpatrick 0.6324
Alex Smith 0.6188
Jason Campbell 0.618
Robert Griffin III 0.618
Matthew Stafford 0.6164
Geno Smith 0.608
Eli Manning 0.6044
Matt Cassel 0.604
Ryan Tannehill 0.596
Chad Henne 0.5928
Joe Flacco 0.5808
Andy Dalton 0.578
EJ Manuel 0.5416
Brandon Weeden 0.528

AVERAGE 0.66
STANDARD DEVIATION 0.07
TANNEHILL PERCENTILE RANK 15.6
 
Last edited by a moderator:
Very cool!! Did you happen to try any other variable selection methods and see if those also produced the same two variables?
 
Very impressive outreach and idea but I think you realize you are setting yourself up to be attacked via any mention or reliance on WPA. That holds a lower reputation around here than YPA, or anything that places Tannehill below 7th in the league, the PFF brainstorm.

These numbers and particularly the ordering look impressive. I agree with what WV posted in the other thread, that it looks as close to a real world list of 2013 quarterback level as numbers can provide.

But none of that matters to the happy adjustment crew, the ones who are currently debating if Marino would have broken 6000 or 7000 yards under today's rules.

Maybe you can add a separate column, on all time quarterback ratings via a backfitted version of this criteria. Best of all, you don't have to actually do the work. Just make up the digits and they'll never notice or care, as long as Marino is 40% clear of second place.

BTW, you are violating my suggestion of spacing these threads a week or so apart. It might have worked other than WPA. I didn't know whether to laugh or cringe once I saw that.
 
Awsi Dooger said:
Very impressive outreach and idea but I think you realize you are setting yourself up to be attacked via any mention or reliance on WPA. That holds a lower reputation around here than YPA, or anything that places Tannehill below 7th in the league, the PFF brainstorm.

These numbers and particularly the ordering look impressive. I agree with what WV posted in the other thread, that it looks as close to a real world list of 2013 quarterback level as numbers can provide.

But none of that matters to the happy adjustment crew, the ones who are currently debating if Marino would have broken 6000 or 7000 yards under today's rules.

Maybe you can add a separate column, on all time quarterback ratings via a backfitted version of this criteria. Best of all, you don't have to actually do the work. Just make up the digits and they'll never notice or care, as long as Marino is 40% clear of second place.

BTW, you are violating my suggestion of spacing these threads a week or so apart. It might have worked other than WPA. I didn't know whether to laugh or cringe once I saw that.
Click to expand...

Wow, quickly attacking and assuming your cleverly named "happy adjustment crew". Always know you're in for a good thread when this is the first post I read from a well credited poster...a little pretentious, in my humble opinion.

Just a question for you Grav, I want to make sure I understand correctly. If the SD is 1, then he is, statistically speaking, no worse than Kaep, Stafford, Brady, Big Ben...and no better than Shaub, Eli, and Glennon?

I could be completely off but I am trying to grasp this the more I see it on the forums.

Surprised to see Glennon so low, and also 2 very recent Superbowl winning QBs in the bottom 7.
 
We've already established that WPA stands for weed puffing addict, so I find this libelous diatribe against Peyton Manning to be especially galling.
 
Awsi Dooger said:
Very impressive outreach and idea but I think you realize you are setting yourself up to be attacked via any mention or reliance on WPA. That holds a lower reputation around here than YPA, or anything that places Tannehill below 7th in the league, the PFF brainstorm.
Click to expand...
And that says more about the forum than it does about WPA, unfortunately.

---------- Post added at 05:36 PM ---------- Previous post was at 05:35 PM ----------
russianbear said:
Wow, quickly attacking and assuming your cleverly named "happy adjustment crew". Always know you're in for a good thread when this is the first post I read from a well credited poster...a little pretentious, in my humble opinion.

Just a question for you Grav, I want to make sure I understand correctly. If the SD is 1, then he is, statistically speaking, no worse than Kaep, Stafford, Brady, Big Ben...and no better than Shaub, Eli, and Glennon?

I could be completely off but I am trying to grasp this the more I see it on the forums.

Surprised to see Glennon so low, and also 2 very recent Superbowl winning QBs in the bottom 7.
Click to expand...
Check the edit of the original post.
 
The thing with "pressures" is that not only is it debatable what may constitute a pressure or what is deemed as a pressure on a specific play, some QBs, such as Peyton Manning for example, simply do not allow themselves to be pressured very often. I think the stats you just rolled out demonstrate that fact. In other words, a pressure is not only subjective, but pressures are variables that sometimes the QBs themselves have a good deal of control over.

But, these things are really interesting you do with the numbers. As someone who has worked with mathematics as a profession, Im impressed.
 
Just curious, where would Josh McCown rank on this chart?
 
Because thats just what we needed, ANOTHER evaluation statistic


ANOTHER

ANOTHER


statistic

:lol:
 
WVDolphan said:
The thing with "pressures" is that not only is it debatable what may constitute a pressure or what is deemed as a pressure on a specific play, some QBs, such as Peyton Manning for example, simply do not allow themselves to be pressured very often. I think the stats you just rolled out demonstrate that fact. In other words, a pressure is not only subjective, but pressures are variables that sometimes the QBs themselves have a good deal of control over.

But, these things are really interesting you do with the numbers. As someone who has worked with mathematics as a profession, Im impressed.
Click to expand...
True, and I farted around a bit with the variable "time to attempt," which PFF tracks and which I believe indicates the duration of time it takes a QB to begin to throw after the snap. Peyton Manning leads the league in that category, which you would think would decrease the frequency with which he's pressured. I'm still farting with it a bit to see if I can factor that in somehow.
 
Gravity said:
True, and I farted around a bit with the variable "time to attempt," which PFF tracks and which I believe indicates the duration of time it takes a QB to begin to throw after the snap. Peyton Manning leads the league in that category, which you would think would decrease the frequency with which he's pressured. I'm still farting with it a bit to see if I can factor that in somehow.
Click to expand...

There are a ton of variables that go into that. If you have a PHd and work in applied mathematics you could probably come up with some interesting stuff. But, even then I would think its simply impossible to account for everything while being able to pin down everything that is a result of the QB only in a statistic. Still fascinating and worth the time for someone skilled enough to play around with it IMHO.

Using Manning again, its easy to say that he gets the ball out so quickly for a variety of reasons, the biggest being his skill level and how he manipulates things at the LOS before the snap. His preparation is remarkable and that allows him to know exactly where he is going with the ball more often than not as soon as it is snapped. Few QBs can even come close to what he does. However there are still variables at play that which Manning simply dosent control. How well the opposing defender plays the man he feels he is going to with the ball when it is snapped is a huge factor. There are just so many things one would have to account for when trying to measure a QBs individual skill statistically.

Having said all of that, IMHO anyone who knows what they are talking about would say that the pass blocking skill of Denver's OL rarely comes into play. Of course that is the case with most really good QBs. When the opposing defense actually finds a way to confuse Manning, the pass defenders actually cause him to have to look at a 2nd option, or both, are really the only times the pass protection is going to have to do a nice job to help Manning out.
 
Another approach, based on factor analyzing WPA per game, EPA per play, DVOA, completion percentage, completion percentage on throws made at 2.5 seconds after the snap and sooner, adjusted net YPA, QB rating, QB rating on throws made at 2.5 seconds after the snap and sooner, and interception rate:

Player "Win Factor"
Peyton Manning 2.123
Nick Foles 1.944
Aaron Rodgers 1.683
Philip Rivers 1.612
Drew Brees 1.559
Russell Wilson 0.832
Tony Romo 0.634
Tom Brady 0.483
Colin Kaepernick 0.408
Matt Ryan 0.394
Ben Roethlisberger 0.304
Andrew Luck 0.234
Jay Cutler 0.225
Cam Newton 0.215
Alex Smith 0.080
Andy Dalton 0.078
Sam Bradford -0.025
Matthew Stafford -0.159
Carson Palmer -0.216
Ryan Fitzpatrick -0.286
Matt Cassel -0.353
Robert Griffin III -0.402
Ryan Tannehill -0.458
Jason Campbell -0.682
Chad Henne -0.688
Mike Glennon -0.762
EJ Manuel -0.827
Joe Flacco -0.876
Case Keenum -1.074
Matt Schaub -1.148
Eli Manning -1.599
Brandon Weeden -1.613
Geno Smith -1.640

AVERAGE 0.000
SD 1.000
TANNEHILL PERCENTILE 31.200


EPA per play alone:

Player EPA Per Play
Peyton Manning 0.33
Nick Foles 0.30
Aaron Rodgers 0.29
Philip Rivers 0.23
Drew Brees 0.22
Russell Wilson 0.16
Colin Kaepernick 0.13
Andrew Luck 0.13
Tom Brady 0.13
Cam Newton 0.13
Tony Romo 0.12
Matt Ryan 0.11
Ryan Fitzpatrick 0.11
Andy Dalton 0.11
Ben Roethlisberger 0.10
Matthew Stafford 0.10
Carson Palmer 0.08
Alex Smith 0.07
Jay Cutler 0.05
Sam Bradford 0.04
Robert Griffin III 0.03
Matt Cassel 0.03
Ryan Tannehill 0.03
Joe Flacco 0.03
Mike Glennon -0.01
Jason Campbell -0.03
Case Keenum -0.05
Eli Manning -0.05
Chad Henne -0.05
Geno Smith -0.06
EJ Manuel -0.07
Matt Schaub -0.08
Brandon Weeden -0.08

AVERAGE 0.08
STANDARD DEVIATION 0.11
TANNEHILL PERCENTILE RANK 28.1
 
Dont forget he also led the nfl in qbs who picked their ass up after getting knocked down.. Lets not pick and choose stats ;)
 
I'd like to see this figured in with nagging injuries, poorly run routes, and bad play calling. After all, there's nothing quite like defining intangibles to suit our agenda, then presenting them as statistics.
 
You must log in or register to reply here.
Back
Top Bottom