My local paper's lead sports columnist had a piece today on whether or not Washington should sign Kirk Cousins to a long-term contract. His argument was that the team should, and he argued that "it’s a lot more likely that he stays the same or gets better than he regresses." So I wondered if that was true. Now Thomas Boswell was talking about Cousins' NFL performance, but this is a fantasy football site, so I'm going to answer the question from a fantasy perspective.

I started to look at this issue in general a couple of years ago and came up with this graph:

This chart was for all QBs from 1978 (the year rules changes made passing easier) to 2013 who had 1500+ career pass attempts. The scoring is 25 pass yards = 1 FP, pass TDs = 4 FP, 10 rush yards = 1 FP, rush TDs = 6 FP, no penalty for turnovers. The curves are differentiated by the QB's age as a rookie, so the blue line is for all QBs in the sample years who first played in the NFL at age 21. And they are 2nd-order polynomial regression curves to smooth out the data.

The curves indicate that QBs peak somewhere between 7 and 10 years in the league, with QBs who are younger as rookies having higher peaks and slower aging declines. The data is affected by survivorship bias – the QBs who last 10+ years are better than average NFL players and so the declines We've seen Cousins be a red-hot commodity for fantasy, but how long will that last?don't exactly reflect the effects of aging on average QBs. And the QBs who were rookies at 21 (Drew Bledsoe, Michael Vick, Matthew Stafford, Alex Smith, and Josh Freeman) and 25 are small samples.

Kirk Cousins was 24 in his first year and has been in the league 5 years. The bad news is QBs who start their careers that late tend to have lower performance curves overall. The good news, and supporting Boswell's contention, is that they have a rough peak when they have 7-10 years in the league (ages 30-33). So Cousins can reasonably expect to get better over the next few years.

There are all sorts of limitations on this graph. It doesn't account for supporting cast, coaching changes, the change in the game between the late 1970s and the 21st Century teens, etc. It lumps together a lot of different QBs of varying quality. So I'm going to look at the question from a slightly different perspective this time.

Cousins has basically been the starter for two full seasons. So I'm going to look at fantasy scoring per game for two years at ages 27-28 and then see how QBs did in the subsequent two years at varying ages. I'll use the same fantasy scoring as above, but only go back to 1988, so I avoid the strike years cluttering up things. And to narrow things down to QBs who were at least part-time starters, I'm going to require a minimum of 25 attempts per game (Cousins averaged 35.9 the last two years). Using two-years smooths things out by limiting the impact of injuries, coaching changes, easy schedules, etc.

Method One

QBs from 1988-2016, Two Year FP/G

Age

Baseline Years

Future Years

Change

22

15.8

18.1

14%

23

16.1

16.7

4%

24

14.3

14.8

3%

25

15.3

15.6

2%

26

15.7

15.1

-4%

27

15.8

15.1

-4%

28

15.9

15.2

-4%

29

15.7

14.7

-6%

30

15.8

14.6

-8%

31

16.0

15.4

-4%

32

15.9

15.9

0%

33

16.5

15.9

-3%

34

16.7

14.7

-12%

35

16.7

15.8

-5%

36

15.8

14.7

-7%

37

16.4

15.1

-8%

38

14.0

11.7

-17%

39

15.8

15.1

-4%

40

15.3

14.9

-3%

41

14.2

8.1

-43%

42

13.5

6.0

-56%

What the table says is, for example, QBs who were 22 years old in this period, averaged 15.8 FP/G when they were 21 and 22 (Baseline Years) and then 18.1 FP/G at age 23 and 24 (Future Years), an improvement of 14% (Change). I'm showing all ages, with Cousins' age 28 highlighted. There are 707 QBs in this study overall and 70 of them were 28 years old. The sample sizes for ages 22 and 38+ are single digit and less than 30 for age 23 and 35+.

So looked at this way, QBs improve over two-year periods from age 22 to 25, then begin declining about 5% every period up to age 31. The decline flattens a bit in the periods after ages 32-33, then picks up again and is generally more severe.

So this story isn't very good for Cousins. He's already entered his decline phase. Of course, if your peak is high enough, losing 4% of your FP/G over the next two years wouldn't be too bad. In Cousins' case, this would drop him from 20.2 to 19.4 FP/G. Or from around QB5 the last two years to QB8. This contradicts Boswell's belief that Cousins will stay the same or improve, but it leaves him as a valuable fantasy QB.

Of course, the table may overstate decline by including players who weren't very good. The 28-year-olds in the study averaged just 16 FP/G in the Baseline Years compared to Cousins' 20. If I only use above average QBs (make the cutoff for Baseline Years FP/G = 16), this is the resulting table:

Method Two

QBs from 1988-2016, Two Year FP/G >16

Age

Baseline Years

Future Years

Change

22

16.8

21.5

28%

23

20.1

18.8

-7%

24

19.3

17.9

-8%

25

18.4

18.4

0%

26

18.4

17.2

-6%

27

18.5

17.2

-7%

28

18.5

17.1

-8%

29

18.6

17.5

-6%

30

18.4

16.6

-10%

31

18.5

17.2

-7%

32

18.0

18.0

0%

33

18.7

17.8

-5%

34

19.0

17.0

-11%

35

18.8

18.0

-4%

36

18.5

16.9

-8%

37

19.3

17.3

-10%

38

16.3

16.1

-2%

39

17.0

15.3

-10%

40

16.8

14.7

-12%

Using 16 FP/G as a cutoff lowers the sample size to 30 at age 28. All other ages are below 30, with single-digit samples at 22, 23, and 35+.

This makes it look like at most ages, QBs will have a future two seasons below their baseline years. Where this is not true, it's probably because of a few outlier performances affecting the small samples (ages 25 and 32 are noticeably unusual). But now it looks like Cousins will drop 8% of his fantasy value over the next two years compared to the last two. Again, he's at a high enough FP/G that a loss of 8% wouldn't cripple his fantasy value, and again, he's being compared to a group that didn't perform as well as he did anyhow.

Here I want to point out that the gradual increase in fantasy QB scoring since 1988 is a factor.

Percentage of Samples by Time

Period

25+ Att/G

16+ FP/G

1988-1996

27%

22%

1997-2005

35%

32%

2006-2014

38%

46%

I divided the 27 years from 1988 to 2014 into three equal nine-year periods. The first sample above (QBs with 25+ attempts/game) skews a little more recent: in 1988, teams averaged 31 attempts per game vs. 36 in 2016. So, more recent QBs are slightly more likely to throw more. But when I had a 16+ FP/G cutoff, the sample shifts more dramatically towards recent years. Passing has gotten much more efficient, which means more fantasy scoring, over this period. Note 2014 is the last year in the sample seasons because we don't have two years of post-2015 or 2016 performance yet.

What this means for Cousins though is that the smaller, higher-scoring sample is more relevant since it is generally more recent.

To get a group of QBs who matched Cousins' 20.2 FP/G in his Baseline Years, I needed to make the minimum FP/G over 18.2 FP/G.

Method Three

QBs from 1988-2016, Two Year FP/G >18.2

Age

Baseline Years

Future Years

Change

23

21.7

18.9

-13%

24

20.5

20.3

-1%

25

20.0

20.2

1%

26

19.5

18.0

-8%

27

20.1

18.8

-7%

28

20.2

18.2

-10%

29

20.0

18.6

-7%

30

19.8

18.2

-8%

31

20.3

19.6

-4%

32

19.9

19.7

-1%

33

19.9

19.6

-2%

34

20.4

17.5

-15%

35

20.8

19.9

-4%

36

19.4

19.1

-2%

37

21.8

16.1

-26%

Most of these sample sizes are single digit. The largest was at age 26, with 16 QBs, followed by age 28 with 15 examples. All age groups declined from the Baseline Years to the Future Years except at age 25. Overall, there were only 138 QBs in the total sample: less than a fifth of all QBs with 25+ attempts in their Baseline Years. Having looked at fantasy performance for a long time, this is not surprising. Most football players who post high scores tend to decline in the following season or seasons.

This sample indicates that Cousins will decline by 10% in FP/G in the coming two years. If you want to see the 28-year-old QBs that went into this calculation, here they are:

  • Drew Brees
  • Marc Bulger
  • Daunte Culpepper
  • Randall Cunningham
  • Brett Favre
  • Peyton Manning
  • Scott Mitchell
  • Philip Rivers
  • Aaron Rodgers
  • Ben Roethlisberger
  • Tony Romo
  • Matt Ryan
  • Mark Sanchez
  • Matt Schaub
  • Kurt Warner

That's a nice mix of Hall of Fame quality players (Peyton, Favre), very good QBs like Romo or Culpepper, and guys at a career peak like Matt Schaub, Scott Mitchell, and Mark Sanchez. You can make an argument for a big contract for Cousins or a cautionary tale out of those names.

To sum up, based on historical averages, Cousins will probably lose about 4-10% of his past two year fantasy scoring in 2017-2018 based purely on his age. Other factors will of course come into play – his health, his supporting cast, the offensive identity of the team he's on, etc. That makes him a low-end fantasy starter.

Now to generalize to all the QBs who have averaged 25+ attempts over the past two years, I have labelled the above tables Method One, Two, and Three. Method One uses all QBs with 25+ attempts in the Baseline Years. Method Two uses the 30 QBs with the Baseline Years FP/G closest to the 2016 QB's baseline. Method Three uses only QBs who had the same Baseline Years FP/G – a subset of Method Two, trading off sample size for more comparable QBs.

2016 QBs with 25+ Att/G in Baseline Years

Player

Age

Baseline Years

Method

Projected Change

Future Years

Jameis Winston

22

18.9

One

14%

21.5

Marcus Mariota

23

19.4

One

4%

20.2

Teddy Bridgewater

24

13.9

Two

6%

14.7

Trevor Siemian

24

14.2

Two

2%

14.5

Blake Bortles

24

20.9

Three

2%

21.4

Derek Carr

25

18.6

Three

-3%

18.1

Matt Barkley

26

12.9

Two

-1%

12.7

Robert Griffin III

26

13.0

Two

-1%

12.9

Brock Osweiler

26

14.5

Two

-2%

14.2

Nick Foles

27

10.4

Three

-9%

9.5

Mike Glennon

27

13.6

Two

-4%

13.0

Kellen Moore

27

15.7

Two

-2%

15.4

Blaine Gabbert

27

16.5

Two

-6%

15.5

Tyrod Taylor

27

19.5

Three

-9%

17.7

Andrew Luck

27

22.5

Three

-6%

21.1

Cam Newton

27

22.6

Three

-6%

21.2

Case Keenum

28

11.5

Three

2%

11.8

Ryan Tannehill

28

17.4

Two

-4%

16.7

Matthew Stafford

28

19.4

Three

-9%

17.7

Kirk Cousins

28

20.2

Three

-10%

18.2

Russell Wilson

28

20.3

Three

-12%

17.9

Colin Kaepernick

29

15.8

Two

-5%

15.0

Sam Bradford

29

16.2

Two

-6%

15.2

Andy Dalton

29

18.8

Two

-5%

17.8

Brian Hoyer

31

15.7

Two

0%

15.7

Joe Flacco

31

17.6

Two

-4%

16.9

Matt Ryan

31

19.9

Three

-6%

18.7

Alex Smith

32

16.9

Two

1%

17.1

Jay Cutler

33

15.9

Two

-3%

15.4

Aaron Rodgers

33

22.4

Three

-6%

21.0

Ryan Fitzpatrick

34

16.3

Two

-12%

14.3

Ben Roethlisberger

34

20.0

Three

-14%

17.2

Eli Manning

35

18.4

Three

-5%

17.4

Philip Rivers

35

19.4

Three

-3%

18.8

Josh McCown

37

16.8

One

-8%

15.4

Carson Palmer

37

19.8

One

-8%

18.2

Drew Brees

37

22.6

One

-8%

20.8

Tom Brady

39

22.2

One

-4%

21.3

The table is sorted by age, then by Baseline Years FP/G. I've indicated the method I used to project future performance. Basically, it's Method Two if I could find a sample of 30 QBs at the appropriate age. If the QB in the player column was exceptionally good or bad, I used Method Three except for the outlier ages (22-23 and 37+). There just aren't that many players like Jameis Winston or Tom Brady, so I felt I shouldn't narrow it down too much.

If you think the chosen method is too pessimistic, you can look at the Method One Table and use those percentages. But the bottom line is that age comes for every QB, even Drew Brees and Brady. My perception is that the Russell Wilson and Ben Roethlisberger projections are the most controversial in this table.

Wilson's two-year baseline is probably depressed because of his injury this year. But here's a small sample of the comps I used for him: Aaron Rodgers – decreased 11% from Baseline to Future; Randall Cunningham – down 15%; Daunte Culpepper – off 19%; Peyton – down 17%. When you put up two big years, you're very likely to decline. Now other changes in Seattle could offset aging: the defense could weaken and the team might have to throw more; the RBs could struggle, causing a greater reliance on the passing game; a switch in philosophy could occur if Pete Carroll leaves after 2017. Remember, this is a two-year projection. Wilson's comp declined more than Cousins – maybe it doesn't make sense to penalize the former more than the latter.

As for Roethlisberger, here's the complete list of comps I used for Method Three:

  • John Elway
  • Trent Green
  • Dan Marino
  • Vinny Testaverde
  • Steve Beuerlein
  • Tony Romo
  • Peyton Manning
  • Joe Montana
  • Warren Moon
  • Tom Brady
  • Steve Young
  • Drew Brees

Of those 12 QBs, 8 are Hall of Famers (now or to be). If I just use those 8, they averaged 20.7 FP/G in their Baseline Years and 18.0 in their Future Years. That's a 13% decline even eliminating Green, Testaverde, Beuerlein, and Romo. Ben could last like Testaverde, but I think age and injury are more likely to take a toll on him than most would admit. If I owned him in a dynasty league, I'd try to move him now while he's on a deep playoff run with a deep WR corps and strong running game. His value is not going up after this.