Previous post:

Next post:

Another Quarterback Aging Curve Post (Adjusted Net Yards Per Attempt Edition)

by Neil Paine on August 6, 2013

in Passing, Quarterbacks, Statgeekery

No, Peyton, you are #1

No, Peyton, you are #1.

Back in March, Chase wrote a post investigating how quarterbacks age, finding that they peak at age 29 (with a generalized peak from 26-30) in terms of value over average. Today, I thought I’d quickly look at how quarterbacks age in terms of their performance rate — specifically, their Adjusted Net Yards per Attempt (ANY/A). For newer readers, ANY/A is based on the following formula: (Passing Yards + 20 * Passing TDs – 45 * INTs – Sack Yards Lost) / (Pass Attempts + Sacks).

First, I need to introduce a way of adjusting ANY/A for era: Relative ANY/A. Relative ANY/A is simply equal to:

QB_ANY/A – LgAvg_ANY/A

The table below lists the 30 single-season leaders in Relative ANY/A since the merger. You won’t be too surprised to see the 2004 version of Peyton Manning at the top. That year, Manning averaged 9.8 ANY/A, while the league average was just 5.6 ANY/A. That means Manning gets a Relative ANY/A grade of +4.1 (with the difference due to rounding).

Rk
Player
Year
Age
Team
G
GS
Cmp
Att
Yds
TD
Int
Sk
SkYd
ANY/A
LgAvgANY/A
Rel_ANY/A
1.Peyton Manning200428clt161633649745574910131019.85.6+4.1
2.Dan Marino198423mia161636256450844817131208.95.0+3.9
3.Roger Staubach197129dal13101262111882154231757.83.9+3.9
4.Bert Jones197625clt14142073433104249292847.84.1+3.7
5.Aaron Rodgers201128gnb15153435024643456362199.45.9+3.5
6.Ken Stabler197631rai121219429127372717192037.44.1+3.4
7.John Brodie197035sfo1414223378294124108677.54.2+3.4
8.Tom Brady200730nwe16163985784806508211288.95.5+3.4
9.Randall Cunningham199835min151425942537043410201328.55.3+3.2
10.Steve Young199231sfo16162684023465257291528.14.9+3.2
11.Mark Rypien199129was1616249421356428117598.35.2+3.2
12.Kurt Warner199928ram161632549943534113292018.35.2+3.1
13.Ken Stabler197429rai141317831024692612181417.03.9+3.1
14.Steve Young199130sfo1110180279251717813798.35.2+3.1
15.Joe Montana198933sfo13132713863521268331988.35.2+3.1
16.Craig Morton197027dal12111022071819157201667.24.2+3.1
17.Dan Fouts198231sdg992043302883171112947.74.8+2.9
18.Joe Montana198428sfo161527943236302810221387.95.0+2.9
19.Ken Anderson197526cin131322837731692111322477.04.0+2.9
20.Steve Young199433sfo161632446139693510311638.25.4+2.9
21.Boomer Esiason198827cin161622338835722814302457.85.0+2.8
22.Kurt Warner200029ram111123534734292118201158.05.2+2.8
23.John Hadl197333ram141413525820082211171266.63.9+2.8
24.Peyton Manning200529clt16163054533747281017818.05.3+2.7
25.Drew Brees200930nor151536351443883411201358.35.7+2.7
26.Philip Rivers200928sdg16163174864254289251678.35.7+2.7
27.Steve McNair200330oti14142504003215247191087.85.2+2.6
28.Brian Griese200025den10102163362688194171397.85.2+2.6
29.Peyton Manning200630clt1616362557439731914867.95.4+2.5
30.Tom Brady201033nwe16163244923900364251758.25.7+2.5

Using this, we can evaluate every quarterback’s season independently of era, and compute the year-to-year differences in Relative ANY/A at every age.

Taking every quarterback who had at least 15.1 dropbacks per game (which tends to correspond to the standard 14 attempts per team game) in back to back seasons, I fed the year-to-year Relative ANY/A deltas into a cubic regression and smoothed out an aging curve. (This is the same process I used to calculate an aging curve for basketball players for ESPN Insider last year.)

According to this methodology, here’s how a QB can expect his Relative ANY/A to change from year to year at each age:

From Age
To Age
Delta
2021+0.88
2122+0.66
2223+0.47
2324+0.32
2425+0.20
2526+0.10
2627+0.02
2728-0.04
2829-0.08
2930-0.11
3031-0.14
3132-0.15
3233-0.17
3334-0.19
3435-0.21
3536-0.24
3637-0.28
3738-0.33
3839-0.40
3940-0.50
4041-0.61
4142-0.76

Or in graphic form:

ANY-A by year

This would indicate that on average, quarterbacks peaks at age 27. To put the data in another light, if we created a passer who peaked at 8.0 ANY/A, and he perfectly followed this age curve, here is how his ANY/A would look each season:

ANY-A progression

One big caveat with this study: there’s probably a good deal of selection bias here, in the sense that only passers deemed to be good enough to keep playing will get a chance to put up 15.1 dropbacks/game the following year. Still, I think this provides a pretty good estimate of how much improvement/decline you can expect from a guy at a given age.

Armed with this aging curve, we can do a lot of cool things in subsequent posts, so stay tuned….

{ 16 comments… read them below or add one }

Dan August 6, 2013 at 1:28 am

The cubic polynomial is a pretty strong assumption, and can distort things (especially near the edges of the graph). Have you looked to see what the data look like if you don’t assume a particular functional form, and just use the raw data from each age (or smoothed with nearby ages)?

Reply

Neil Paine August 6, 2013 at 12:26 pm

I definitely graphed it out beforehand, and the cubic model is the best fit (as it tends to be for aging curves in other sports as well). The problem with using the actual computed values, as opposed to the theoretical ones from the model, is that you won’t see a clean progression. And if you’re going to smooth between nearby ages, you might as well just use the regression to smooth the whole thing out.

Reply

Shattenjager August 6, 2013 at 1:34 pm

As a Broncos fan, the right end of that graph looks very bad.

Reply

Neil Paine August 7, 2013 at 10:03 am

Even if Peyton was a true +1.5 RANYPA QB last season (the average of his last 3 full seasons), instead of the +2.0 he put up in 2012, you’d still expect him to be a +1.2 passer this season (losing 0.3 RANYPA as he moves from 36 to 37). If you know nothing about a team except that it produced a +1.2 RANYPA, you’d predict that team to put up a WPct of .640 (10.2 W/16g).

And that doesn’t even take Denver’s strong defense (granted, one that will be missing Von Miller for 1/4 of the year) into account. I think the Broncos are going to be fine in 2012.

Beyond 2012, they knew the kind of bargain they were getting into when they grabbed a 36y/o Manning coming off what was thought to possibly be a career-ending injury. When a QB misses his entire age 35 season with injury, anything you get out of him at 36 and beyond is gravy IMO.

Reply

Shattenjager August 7, 2013 at 12:30 pm

The problem is actually after this season, because the declines get even more dramatic. If he were to follow the aging curve exactly as a 1.5 RANYPA player as you said, he’s under 1 next season, which suddenly means that the passing game isn’t enough to carry the team, which worries me.

Plus, it’s really just not fun watching a great player decline like that. I lived in Minnesota for three years and I have hated him since 1996, but even watching Brett Favre fall apart in 2010 was just sad.

And of course the real best thing about signing Manning was being able to get rid of his predecessor.

Reply

Richie August 13, 2013 at 6:27 pm

Yeah. In a slightly injury-shortened 1996 season, Dan Marino led the league in ANY/A at age 35. The decline was so fast that in 1999 he threw more INTs than TDs in his final season.

Reply

Chase Stuart August 6, 2013 at 6:02 pm

HADL!

Reply

Neil Paine August 7, 2013 at 10:14 am

Hadl’s 1973 is constantly disrespected whenever people delve into the annals of great QB seasons. Maybe it was a fluke (it was sandwiched between 2 below-average seasons), but it still stands up as one of the best performances ever, relative to its era. Translated to the modern era, Hadl’s year was the equivalent of something like 3,300 yd/60 cmp%/30 TDs/10 INTs — shades of a year on par with Randall Cunningham’s 1998.

Reply

Chase Stuart August 7, 2013 at 10:42 am

We really need a full bio of Hadl one say.

Reply

Chase Stuart August 7, 2013 at 10:46 am

Interesting note: According to this, Hadl’s best season was ’67: http://www.footballperspective.com/the-greatest-qb-of-all-time-iv-part-i-methodology/

Part of the reason was attempts, but also, he had a great sack rate that year (assuming we can trust these numbers): http://www.pro-football-reference.com/years/1967_AFL/

Reply

Neil Paine August 7, 2013 at 10:54 am

Were pre-merger numbers relative to the overall pro football average or the specific NFL/AFL average? I think it’d have to be the latter, but then find some way to determine the relative quality of competition in each league by looking at guys who played in both and how it impacted their numbers. This kind of messiness is why I usually avoid pre-merger stats entirely.

Reply

Chase Stuart August 7, 2013 at 10:57 am

Pre-merger stats are the only things that matter!

Reply

Chase Stuart August 7, 2013 at 10:58 am

And yeah, I compared NFLers to NFLers and AFLers to AFLers when I did that (at least, I think). In some studies, I’ve included an AFL modifier (definitely with WRs, can’t remember if I did with QBs) but in any event, I think that was eliminated by ’67. Seeing as how the AFL won the SB in ’68 and ’69, and there have been lots of post-merger years with a bad conference, I think I’ve always considered the AFL equal to the NFL starting in ’67.

Reply

Richie August 13, 2013 at 6:41 pm

Here is the study Jason Lisk did on AFL-NFL a few years ago: http://www.pro-football-reference.com/blog/?p=4409#more-4409

By 1968 the leagues were pretty close, but the NFL was better all along, with the AFL getting closer and closer each year.

Reply

Shattenjager August 7, 2013 at 12:32 pm

Brian Griese’s 2000 is in the same boat (though it was not a full season).

I actually think if he had never gotten injured that year, his entire career may have gone very, very differently.

Reply

James August 11, 2013 at 10:20 pm

Thoughts on Brian Burke’s QB aging study that showed a QB’s last year was almost always bad but within normal variation, and once removed QBs showed nearly no decline at all?

http://www.advancednflstats.com/2011/08/how-quarterbacks-age.html

Reply

Leave a Comment

You can use these HTML tags and attributes: <strong> <em> <pre> <code> <a href="" title="">

{ 12 trackbacks }