Projecting the Careers of the 2019 Quarterback Class

[AlNFL19]

In my spare time, I've worked to develop a regression model to predict the career success of college quarterbacks in the NFL. I took a couple of different statistics, including Games Started, Draft Position, and Adjusted YPA, to develop this model. The end result of each projection is in terms of Approximate Value, which is ProFootballReference's all-in-one evaluation statistic. The sample goes from 2006, the last class when all used data is available, to 2015. It only goes to 2015's draft because the final projection is in "average Approximate Value in Years 3 and 4". Players drafted later than 2015 don't have 4 years of NFL experience yet. I chose years 3 and 4 because they're more reliable than years 1 and 2 and are still on the rookie contract of every player.

My model projects an Approximate Value, a percent chance of being a bust (which I determined to be <5 AV) and percent chance of being a Pro Bowler (>10 AV) for quarterbacks drafted in the first three rounds only.

It's not a perfect model, obviously, but it's pretty okay. Having a high projection won't guarantee success, but the projected chances of reaching different thresholds have been relatively fair. Its percent chances, since 2006, have been relatively close to projected, as seen below:

Projection: >5.00 AV
Threshold	Reality	Projected Avg.
Pro Bowl	44.44%	40.63%
Hit	55.56%	49.60%
Bust	44.44%	50.40%

Projection: <5.00 AV
Threshold	Reality	Projected Avg.
Pro Bowl	4.76%	12.46%
Hit	9.52%	18.12%
Bust	90.48%	81.88%

Overall, the projected chances have been relatively close to the results. Particularly for those under a projection of 5.0 AV, there isn't much in the way of prospects for success. Of the 21 qualified passers since '06 to have a sub-5 projection, only Matt Ryan has surpassed 10.0 AV and only 2 of the 21 have even hit 5 (with the other being Chad Henne hitting exactly 5, so even calling him a "hit" is debatable).

 

Without further ado, the rankings (again, this model only works for quarterbacks drafted, or in this case projected to be drafted, within the first three rounds):

1. Kyler Murray, Oklahoma

Kyler Murray, QB, Oklahoma
Statistic	Figure
Projected AV	7.10
Bust Chance	29.24%
PB Chance	60.86%

Obviously, Murray's statistically-historic year results in a pretty high projection, giving him a Top 5 ranking since 2006 (along with Baker Mayfield, Marcus Mariota, Cam Newton, and Vince Young). Murray's projection ranks as one of just 6 quarterbacks since 2006 (Rounds 1-3) with a Pro Bowl chance of greater than 50%. Murray's 7.2 rushing YPA (a statistic that actually correlates greater than you'd expect to QB success) figures out to an 8.33 projection (rushing YPA is one of several stats that comprises the final weighted average AV projection).

2. Dwayne Haskins, Ohio State

Dwayne Haskins, QB, Ohio State
Statistic	Figure
Projected AV	5.56
Bust Chance	53.27%
PB Chance	37.89%

Haskins' 5.56 projection is right between Mark Sanchez (yikes) and Jared Goff, who hit 5.57 and 5.51, respectively. Haskins has the highest projection for any Big Ten quarterback since 2006, and is the only one above 5.0 projected AV. However, there's a decent gap between Murray and Haskins, according to the model, as Haskins is in the 50+% bust club (along with most other prospects).

3. Drew Lock, Missouri

Drew Lock, QB, Missouri
Statistic	Figure
Projected AV	5.42
Bust Chance	55.47%
PB Chance	35.79%

Lock's projection is helped by his class-best 46 career starts and his senior year completion rate, which finally climbed above 60%. Historically, Lock's projection falls right below Jameis Winston's 5.46, and Lock is the final quarterback from this class with a projection above 5.0 AV.

4. Ryan Finley, N.C. State

Ryan Finley, QB, N.C. State
Statistic	Figure
Projected AV	4.55
Bust Chance	68.97%
PB Chance	22.89%

There's a pretty significance drop-off from the Murray-Haskins-Lock trio, the consensus Round 1 quarterbacks, to the second tier in my model. Finley's projection falls below 5.0 AV, which is bad but not a guarantee of failure. However, just 2 of 21 qualified passers since 2006 have hit 5.0 AV despite a lower projection (Matt Ryan - who was Top-3 among the below-fivers - and Chad Henne). However, Finley hits all 7 of the Parcells Rules for drafting a quarterback.

5. Daniel Jones, Duke

Daniel Jones, QB, Duke
Statistic	Figure
Projected AV	3.95
Bust Chance	78.29%
PB Chance	13.99%

Jones' projection is right below Brian Brohm's 4.00. Jones was seen as a Day 1 quarterback, but has seen his stock drop as of late. Obviously, that lower expected draft position hurts his standing in the model. Quarterbacks need dominating statistics in the model to hit 5.0 as a projection despite a low (expected) draft position, and Jones didn't have them (Russell Wilson did!).

6. Jarrett Stidham, Auburn

Jarrett Stidham, QB, Auburn
Statistic	Figure
Projected AV	3.71
Bust Chance	82.08%
PB Chance	10.37%

Stidham did himself no favors in returning to school. He was liked by many (myself included) as a potential Round 1 guy this year coming into the 2018 season, and he tanked his stock. Had he not returned to school, his projected AV would have been 4.57 rather than 3.71. Stidham's two closest prospects historically aren't good signs for his NFL career, as both (Colt McCoy and Garrett Grayson) put up 0.0 Approximate Value in their third and fourth years in the league.

7. Will Grier, West Virginia

Will Grier, QB, West Virginia
Statistic	Figure
Projected AV	3.04
Bust Chance	92.57%
PB Chance	0.35%

Grier put up a ton of numbers at West Virginia, playing Air Raid just like Kyler Murray. So what gives? 

Well, Grier has:

• a much lower expected draft position
• negative yards per rushing attempt
• not as good AY/A (10.7 to Murray's record 13.0)

And that's just to name a few things. Grier is a quarterback on whom statistical models don't agree. FiveThirtyEight's model (results below) really likes Grier's chances, while FootballOutsiders does not. However, there is hope for Grier's NFL career. There is one quarterback below his projection to hit 5.0 AV in the NFL - Michigan's Chad Henne, who hit exactly 5.0 to beat his projection of 2.72 AV.

 

Other Projections

Football Outsiders' QBASE
Player	Proj. DYAR	Bust Chance	Elite Chance
K. Murray	595	47.00%	7.00%
D. Haskins	527	51.00%	5.00%
R. Finley	398	52.00%	6.00%
D. Lock	271	59.00%	3.00%
D. Jones	263	56.00%	5.00%
J. Stidham	-45	67.00%	1.00%
W. Grier	-151	71.00%	1.00%

FO's Functional Mobility Model
Player	Projected ANY/A
K. Murray	6.16
D. Lock	5.78
R. Finley	5.65
D. Haskins	5.58
W. Grier	5.36
D. Jones	5.15
J. Stidham	4.80

FTO's Completion Percentage Over Expected Model
Player	CPOE	7.1 Chance*
K. Murray	+9%	97.00%
W. Grier	+6%	90.00%
R. Finley	+4%	78.00%
D. Haskins	+9%	63.00%
D. Jones	-2%	17.00%
J. Stidham	+3%	3.00%
D. Lock	-1%	<1.00%
"7.1 Chance" is the % chance they hit 7.1 YPA.

And the average ranking:

1. Murray

2. Haskins

3. Finley

4. Jones

T-5. Lock

T-5. Grier

7. Stidham

[BayRaider]

Outside of your model, seems most sites really like Finley. I think he’ll be great tbh.

[aDandyGuyInSpace]

Do you have one for last years qbs?

[Fl0nkerton]

This is cool. Yeah, could you post some of the previous years?

[AlNFL19]

14 hours ago, aDandyGuyInSpace said:

Do you have one for last years qbs?

 

4 hours ago, Fl0nkerton said:

This is cool. Yeah, could you post some of the previous years?

Sure. Sorry for the delay.

2016 NFL Draft
Player	Projected AV	Bust Chance	PB Chance
J. Goff	5.51	54.01%	37.19%
P. Lynch	5.25	58.11%	33.27%
J. Brissett	3.52	85.12%	7.47%
C. Hackenberg	3.34	87.94%	4.77%
C. Kessler	2.82	95.98%	<1.00%

2017 NFL Draft
Player	Projected AV	Bust Chance	PB Chance
M. Trubisky	5.87	48.47%	42.48%
D. Watson	5.80	49.56%	41.44%
P. Mahomes	5.65	51.90%	39.20%
D. Kizer	4.52	69.51%	22.37%
D. Webb	2.61	99.25%	<1.00%
C.J. Beathard	2.31	>99.90%	<1.00%

2018 NFL Draft
Player	Projected AV	Bust Chance	PB Chance
B. Mayfield	6.75	34.68%	55.66%
L. Jackson	5.93	47.46%	43.45%
S. Darnold	5.35	56.60%	34.72%
J. Rosen	4.62	67.97%	23.85%
J. Allen	4.56	68.83%	23.03%
M. Rudolph	4.23	73.94%	18.14%

Some things to note:

• Mayfield's projection is 5th since 2006
• Beathard's projection is 3rd-worst since 2006
• Allen's projection is the 2nd-worst for a 1st-round pick since 2006 (Brandon Weeden)
• Jackson's projection is 12th since 2006

[aDandyGuyInSpace]

I was curious to see how Lamar stacked up and also to see if any qbs with a high AV would bust, it looks like Paxton lynch is really the only outlier, that bodes well for us Ravens fans

[AlNFL19]

24 minutes ago, aDandyGuyInSpace said:

I was curious to see how Lamar stacked up and also to see if any qbs with a high AV would bust, it looks like Paxton lynch is really the only outlier, that bodes well for us Ravens fans

• The current split for quarterbacks with a higher projection than Jackson (excluding Mayfield) is 6 "hits" and 4 "busts", which isn't a great rate but when the projected average bust rate for the 10 is 41.55%, 40% is right where you'd expect it to be. 
• The bucket directly below Lamar is 4-for-7 hitting 5.0 AV, despite an average projected bust rate of 50% (around 41% in reality).

Basically, the numbers are slightly on Jackson's side. 

This is probably because, interestingly enough, one of the statistics I tested with the highest correlation to NFL Approximate Value for quarterbacks was Rushing Yards per Attempt. It sounds crazy, but it makes some sense. Check this out if you feel like it, I think it explains it a little: https://www.footballoutsiders.com/stat-analysis/2019/qb-functional-mobility-model-2019. Jackson's 6.9 final season YPA rushing is second since 2006 to just Murray at 7.2, which obviously projects highly.

[Scoundrel]

Some of us didn’t need this fancy system to know Baker was going to be good. Just saying... 

[aDandyGuyInSpace]

1 hour ago, Counselor said:

Some of us didn’t need this fancy system to know Baker was going to be good. Just saying... 

Hey he's a stud and yall should be stoked. I hope Lamar makes strides and we have a good rivalry moving forward  bro

[Scoundrel]

8 minutes ago, aDandyGuyInSpace said:

Hey he's a stud and yall should be stoked. I hope Lamar makes strides and we have a good rivalry moving forward  bro

I’m a Broncos fan. I’d be stoked if he was a Bronco 

[aDandyGuyInSpace]

2 hours ago, AlNFL19 said:

• The current split for quarterbacks with a higher projection than Jackson (excluding Mayfield) is 6 "hits" and 4 "busts", which isn't a great rate but when the projected average bust rate for the 10 is 41.55%, 40% is right where you'd expect it to be. 
• The bucket directly below Lamar is 4-for-7 hitting 5.0 AV, despite an average projected bust rate of 50% (around 41% in reality).

Basically, the numbers are slightly on Jackson's side. 

This is probably because, interestingly enough, one of the statistics I tested with the highest correlation to NFL Approximate Value for quarterbacks was Rushing Yards per Attempt. It sounds crazy, but it makes some sense. Check this out if you feel like it, I think it explains it a little: https://www.footballoutsiders.com/stat-analysis/2019/qb-functional-mobility-model-2019. Jackson's 6.9 final season YPA rushing is second since 2006 to just Murray at 7.2, which obviously projects highly.

Great read. Interesting how Lamar's numbers were over Darnolds and Allens. His rushing really does make him another level threat if his passing improves

[aDandyGuyInSpace]

Just now, Counselor said:

I’m a Broncos fan. I’d be stoked if he was a Bronco 

Sorry for assuming. I should do a profile check before claiming you as a Brownie.

[AlNFL19]

Okay, so I've been working the last few days on updating the model. I came up with an adjustment factor for Experience. It correlates just as well to success as the original model, so I'll just show the experience-adjusted projections here:

Experience-Adjusted (2016-19 Classes)

2016-2019 Draft Classes: Experience-Adjusted Projections
Year	Player	Projected AV	Bust Chance	PB Chance
2019	K. Murray	6.93	31.91%	58.31%
2018	B. Mayfield	6.80	33.93%	56.37%
2018	L. Jackson	5.94	47.34%	43.57%
2017	D. Watson	5.78	49.83%	41.18%
2017	M. Trubisky	5.75	50.30%	40.74%
2017	P. Mahomes	5.62	52.32%	38.80%
2016	J. Goff	5.54	53.57%	37.61%
2019	D. Lock	5.46	54.82%	36.42%
2019	D. Haskins	5.43	55.28%	35.98%
2018	S. Darnold	5.31	57.15%	34.18%
2016	P. Lynch	5.28	57.62%	33.74%
2018	J. Rosen	4.60	68.22%	23.61%
2019	R. Finley	4.59	68.38%	23.46%
2018	J. Allen	4.53	69.31%	22.57%
2017	D. Kizer	4.48	70.09%	21.82%
2018	M. Rudolph	4.27	73.36%	18.69%
2019	D. Jones	3.97	78.04%	14.23%
2019	J. Stidham	3.68	82.56%	9.91%
2016	J. Brissett	3.49	85.52%	7.08%
2016	C. Hackenberg	3.36	87.55%	5.14%
2019	W. Grier	3.00	93.16%	<1.00%
2016	C. Kessler	2.86	95.34%	<1.00%
2017	D. Webb	2.60	99.39%	<1.00%
2017	C.J. Beathard	2.30	>99.90%	<1.00%

Some notes:

• The experience adjustment knocks Murray's projection, but not by much: from 7.10 to 6.93, still an elite projection.
• With the experience adjustment, Murray and Mayfield both rank Top 3, behind Mariota's 6.95.
• The experience adjustment hits Trubisky with a -0.12 net, knocking him below Deshaun Watson.
• The experience adjustment makes a big difference in this year's rankings, as it puts Drew Lock ahead of Dwayne Haskins

And the 2019 rankings, adjusted for college experience:

1. Kyler Murray

2. Drew Lock

3. Dwayne Haskins

4. Ryan Finley

5. Daniel Jones

6. Jarrett Stidham

7. Will Grier

Edited April 21, 2019 by AlNFL19

[HoboRocket]

34 minutes ago, AlNFL19 said:

Okay, so I've been working the last few days on updating the model. I came up with an adjustment factor for Experience. It correlates just as well to success as the original model, so I'll just show the experience-adjusted projections here:

Experience-Adjusted (2016-19 Classes)

2016-2019 Draft Classes: Experience-Adjusted Projections
Year	Player	Projected AV	Bust Chance	PB Chance
2019	K. Murray	6.93	31.91%	58.31%
2018	B. Mayfield	6.80	33.93%	56.37%
2018	L. Jackson	5.94	47.34%	43.57%
2017	D. Watson	5.78	49.83%	41.18%
2017	M. Trubisky	5.75	50.30%	40.74%
2017	P. Mahomes	5.62	52.32%	38.80%
2016	J. Goff	5.54	53.57%	37.61%
2019	D. Lock	5.46	54.82%	36.42%
2019	D. Haskins	5.43	55.28%	35.98%
2018	S. Darnold	5.31	57.15%	34.18%
2016	P. Lynch	5.28	57.62%	33.74%
2018	J. Rosen	4.60	68.22%	23.61%
2019	R. Finley	4.59	68.38%	23.46%
2018	J. Allen	4.53	69.31%	22.57%
2017	D. Kizer	4.48	70.09%	21.82%
2018	M. Rudolph	4.27	73.36%	18.69%
2019	D. Jones	3.97	78.04%	14.23%
2019	J. Stidham	3.68	82.56%	9.91%
2016	J. Brissett	3.49	85.52%	7.08%
2016	C. Hackenberg	3.36	87.55%	5.14%
2019	W. Grier	3.00	93.16%	<1.00%
2016	C. Kessler	2.86	95.34%	<1.00%
2017	D. Webb	2.60	99.39%	<1.00%
2017	C.J. Beathard	2.30	>99.90%	<1.00%

Some notes:

• The experience adjustment knocks Murray's projection, but not by much: from 7.10 to 6.93, still an elite projection.
• With the experience adjustment, Murray and Mayfield both rank Top 3, behind Mariota's 6.95.
• The experience adjustment hits Trubisky with a -0.12 net, knocking him below Deshaun Watson.
• The experience adjustment makes a big difference in this year's rankings, as it puts Drew Lock ahead of Dwayne Haskins

And the 2019 rankings, adjusted for college experience:

1. Kyler Murray

2. Drew Lock

3. Dwayne Haskins

4. Ryan Finley

5. Daniel Jones

6. Jarrett Stidham

7. Will Grier

With the experience adjustment, Clayton Thorson's projection shoots up to 7.03 (just kidding). Though, I am curious - what about Thorson and Tyree Jackson?

[AlNFL19]

20 minutes ago, HoboRocket said:

With the experience adjustment, Clayton Thorson's projection shoots up to 7.03 (just kidding). Though, I am curious - what about Thorson and Tyree Jackson?

Neither is projected by Scouts Inc. to be a Day 1-2 prospect, so I didn't include them, but here are their projections:

Regular Projection:

T. Jackson, QB, Buffalo
Statistic	Figure
Projected AV	3.22
Bust Chance	89.81%
PB Chance	2.98%

C. Thorson, QB, Northwestern
Statistic	Figure
Projected AV	1.36
Bust Chance	>99.90%
PB Chance	<1.00%

Experience-Adjusted Projection:

T. Jackson, QB, Buffalo
Statistic	Figure
Projected AV	3.21
Bust Chance	89.88%
PB Chance	2.91%

C. Thorson, QB, Northwestern
Statistic	Figure
Projected AV	1.39
Bust Chance	>99.90%
PB Chance	<1.00%

Either way, the model hates both of them. Starting a million games wasn't enough for Thorson to overcome some dreadful statistics.

**Keep in mind, statistics for quarterbacks drafted after Round 3 are even more unreliable than before (which is certainly saying something), so these projections might be more off than the other ones will be.