Writers of Pro Football Prospectus 2008

30 Dec 2008

Do Older Quarterbacks Wilt In December and January?

In Aaron's appearance on the B.S. Report (a free-flowing conversation that occasionally touches on mature subjects) this week, Bill Simmons asked us to analyze whether aged quarterbacks wilt in the winter months, their creaky bones too brittle for the 16-game schedule.

While Bill suggested that we put the answer in PFP 2009, this is exactly the sort of question the fantasy database that I use to project matchups each week has a readily-available answer to. While we don't have 2008 data in there as of yet, we can look at how quarterbacks have fared over the 13 years of game-by-game data we've compiled for the database already.

The data set here is every instance of a quarterback throwing 10 or more passes in a game from 1995 through 2007. The quarterback's age is defined as whatever his age was on September 1 of the season in question.

Of course, defining the age at which a quarterback is old is a difficult enough question. The 90th percentile of quarterback age amongst the 6500 games in the study is 35, so we'll define quarterbacks 35 and older to be the quarterbacks in question. I wanted to do this quickly, and the fantasy database doesn't have game-by-game DYAR, so we'll use quarterback rating as the measure.

Let's split the months to compare how a quarterback does in December and January (regular season games only) as opposed to how he does in the rest of the regular season. We'll split the remaining quarterbacks a couple of different ways.

QB Fantasy Points By Age, Month
Age Aug to Nov Dec & Jan Dropoff
21-26 77.0 74.3 3.5%
27-30 83.3 83.2 0.2%
31-34 81.8 79.2 3.1%
35-44 83.0 80.9 2.6%
 
21-34 80.5 78.5 2.5%
35-44 83.0 80.9 2.6%

Those elderly quarterbacks are losing more in the way of quarterback rating than their younger brethren, but that's because they gain more points during the warmer months than those quarterbacks do. That's due to selection bias: The quarterbacks who have made it as professionals to the age of 35 are a more talented, successful group of quarterbacks than the broader pool of quarterbacks who enter the league and play from the ages of 21 to 34.

If we measure the effect of the colder months as a percentage of their average quarterback rating in the previous months, old quarterbacks suffer from the cold weather and strain of a season almost identically as much as the broader pool. The difference is less than one-tenth of one percent.

In other words, while old quarterbacks play slightly worse at the end of the season, it's a trend that has nothing to do with their age.

Posted by: Bill Barnwell on 30 Dec 2008

24 comments, Last at 05 Jan 2009, 5:50pm by jjbtnw

Comments

1
by Joe T. (not verified) :: Tue, 12/30/2008 - 12:11pm

Well argued Bill.

I'd argue also that many quarterbacks who play beyond 35 have proved themselves more physically durable than others. 35+ year old quarterbacks are mostly physical anomalies themselves, i.e. Vinnie, Brett, Flutie. Plus, I'd say that many 35+ QBs, while not necessarily physical freaks like Vinnie or Brett, have not been subjected to first-string QB abuse their entire career. Flutie, Kerry Collins, Kurt Warner, etc all spent time either away from NFL football (or football altogether) or were bench jockeys for a substantial portion of their career. Riding the pine certainly helps to preserve you and prolong your career.

Interesting subject matter.

2
by NHRevsFan (not verified) :: Tue, 12/30/2008 - 12:25pm

...still probably worth a more thoughtful look, though:

1) Is 35 the right age? For positions dependent on power and speed as their primary athletic attributes it may be earlier. QB is about throwing ability (what's the rule of thumb for pitchers in baseball?) and judgment / game-management. For modern athletes, that age cutoff might be a couple of years higher at 37 or 38 before there's a real effect.

2) As you rightly point out, there's a winnowing process, and anyone playing anyRothing in the NFL well into their 30's is pretty good at what they do. Therefore, maybe you need a longitudinal study against the same population of quarterbacks. Take the population of 35+ quarterbacks performance at 35+ and compare it to when they were younger. What does that tell you?

13
by sundown (not verified) :: Tue, 12/30/2008 - 3:49pm

1) Tough to measure accurately because you have precious few starting QBs who are that old. (Which in and of itself makes a strong case that the cutoff age should be lower than 38.)
2) What do you suppose that study would tell you? You'll be looking at a tiny sample of aging superstars. Does Favre look like the same guy he was in the 90s?

3
by Bright Blue Shorts :: Tue, 12/30/2008 - 12:30pm

Not so much looking to criticise the study, as I think the results are about right. Purely as a matter of interest ... what are the sample sizes?

I'm struggling to think of too many regular starting QBs over 35 in recent years ... Rich Gannon, Brett Favre er ... um ... even Steve McNair doesn't qualify.

But I guess back in the late 1990s - Jim Kelly, Dan Marino, John Elway, Steve Young, Doug Flutie must make up a bulk of the reps ... with the latter two getting significant backup reps more recently ...

Was there a youth-movement effect earlier this decade?

8
by Bill Barnwell :: Tue, 12/30/2008 - 1:24pm

35+: 657 games August through November. 256 games December and January.

10
by Bright Blue Shorts :: Tue, 12/30/2008 - 2:56pm

Thanks Bill ... would I appear ungratefully if I asked you to further refine the answer? :-)

How many quarterbacks do those games represent? Beyond the eight I mention, are there any significant omissions?

I must say that a total of 957 out of 6500 matchups represent a lot more than I would have estimated being played by QBs over 35. Did Vince Evans make it onto the list?!?

12
by Bill Barnwell :: Tue, 12/30/2008 - 3:26pm

No, not a problem. Quick head count has the total around 41, yes, including Vince Evans.

4
by MJK :: Tue, 12/30/2008 - 12:51pm

What if, instead of age, you look at seasons played? As Joe T. pointed out, a lot of QB's playing at age 35+ took some years off before then. If we're worrying if old bones and joints feel the cold more, maybe the question is not how old they are, but how much punishment they've taken.

5
by Bright Blue Shorts :: Tue, 12/30/2008 - 1:01pm

I think age is the right thing to use because we're looking to 'prove' a biological drop off here. That said perhaps it would be better to use some sort of hair-loss or wrinkles scale. Brett Favre is noticeably much greyer in his beard these days than during the Super Bowl years ...

6
by Evan D. (not verified) :: Tue, 12/30/2008 - 1:09pm

I'm wondering if there's some warm weather or dome effects going on here. Is there any way to break it down by location of the game as well? (Generalizations like New England is always cold/San Diego is always warm are probably fine.)

7
by Bright Blue Shorts :: Tue, 12/30/2008 - 1:17pm

Do warm-weather/domes matter given that you're sampling all the QBs in the league? Basically for every Brett Favre playing in Green Bay; there's a Kyle Orton/Tom Brady/Peyton Manning playing in Chicago/New England/Indianapolis.

I guess that's why it might be important to know the sampling. If weather does affect it, you might see more conservative gameplans in these months.

Another factor to consider is schedules. If I'm not mistaken the NFL tends to schedule more within-division games for the last quarter of the season (i.e. Dec/Jan). Doesn't the familiarity/rivalry thing play into this somewhere?

But again, does it matter if you're sampling all the QBs in the league?

9
by Bobman :: Tue, 12/30/2008 - 1:25pm

Not sure how you'd reliably measure mileage. There's precious little data on QB hits and hurries the farther back we go and even for sacks, pre 1980 or so... nothing reliable.

Which QBs have played every game this year? Do they all have the same mileage from those 16 games? I'd argue no. Manning, with only 12 sacks, has been pounded. Ben R,with 99 sacks against him, has been pounded more. With a super running game and solid OL, I assume Eli has not taken too much of a beating. Brady, who only played one quarter of one game suffered an injury that might, to some, mean his year was pretty low on the pounding scale... of course it could also negatively affect every throw he makes from now til the end of his career as well. Time will tell

Might as well go with age--at least it's objective and measurable. And as we see, doesn't matter all that much. An NFL QB is an NFL QB. There are probably more stiffs in the 22-28 age range (backups thrust into the limelight for their one chance to do well) than there are guys at 35 (if you're a 13-year veteran QB, you clearly do not suck. You might not be what you once were, but you are better than most first-time starters who were backups the week before. The old guys usually have *something* whether it's craftiness, a cannon arm, quick release, etc. even if other aspects of their game have faded.)

11
by Neoplatonist Bolthead (not verified) :: Tue, 12/30/2008 - 3:26pm

Don't really young QBs also conventionally struggle in the postseason? If you look at that chart, it would seem to bear out that the postseason is least hard on guys in their late 20s, which is what one would expect if there were a "rookie wall" and a "veteran wall," no? This would seem to say that Rivers and Roethlisberger would seem to be in good shape on this particular metric.

15
by sundown (not verified) :: Tue, 12/30/2008 - 3:53pm

Excellent point. There's a small window in there between being "young and unproven" and "old." And, the numbers do suggest that (gasp) guys play best in what most would call the prime years of their careers.

17
by Bright Blue Shorts :: Tue, 12/30/2008 - 4:26pm

If I'm not mistaken you didn't read carefully ... it's "(regular season only)" ...

14
by Uncle Willy (not verified) :: Tue, 12/30/2008 - 3:53pm

What, if any, impact would "meaningless" Week 17 games have? Or would the number of meaningless games for the old QB's teams vs meaningless for their opponents cancel out? Perhaps the last week of the season should be removed?

16
by sundown (not verified) :: Tue, 12/30/2008 - 3:57pm

How do you define "meaningless"? If you're a QB in your mid-30s I can assure you that you're treating every single game like it may be your last. The only exception to that would be if your team has sewn up the division and are resting for the playoffs.

18
by ceolaf :: Tue, 12/30/2008 - 5:21pm

I think that there is a conceptual problem with this study.

You are assuming that late season older QB's are the same people as early season old QBs, and that might not be the case.

So, yes it is true that older QB who start early in the season are likely pretty good. But what about veteran backups for injured projected starters? If you Flacco or Ryan had gotten hurt, who would have started for these teams in December? Heck, look at Vince Young and Kerry Collins. Is Collins really that good? No, of course not. He's a competent veteran backup.

Competent veteran backups are old, almost by definition.

And so, you've answered the question about the groups (old QBs, young QBs), but not whether individual old or young QBs get better/worse in the cold.

*********************

So, how to fix this?

1) Well, you still want a good sample size, and so you still want individual game ratings. Fine.

2) You need to somehow correct for whether each game is better or worse that THAT QB did the rest of the season. That's pretty easy. Use the difference between their season's rating and their game rating as you individual data points.

3) Run the same averages.

This would tell you whether individual QBs get better or worse, on average, grouped by age. QBs who only played part of the season (e.g. Vince Young) would not affect the numbers. It would automatically remove the impact of late season "let's see what he's got" starts for young QBs, or late season replacements by veterans who didn't start at all early in the season.

You'll see any trends clearly, if they exist. If you run columns for each month, you'll be able to see if the youngest (i.e. <25) QBs actually improve, or where the big jump or The Wall is.

**************************

You could also easily run this by years experience (easy to get), which might get you some different stuff. Heck, you might be able to run it by games experience, but that might be harder. (Obviously, I am thinking more about young QB development here.)

Another relatively easy one might be to look at what start it is of the year for that QB. That is, address the "wisdom" of the idea "he can fill in for a couple of weeks, but that's about it." Do younger QBs hit a wall at a certain start in the season? Do older QBs wear out around week 13?

It would be harder to add game time temperature, but that IS possible. That's really the question, isn't it? Comparing younger QBs performance in cold weather games to their performance in warm weather games, and comparing that comparison to older (or just old) QB's comparison.

comparison comparison comparison

So, if you want to do the article, try different versions of the question.
IV variants #1 : Age, experience, games started (career), game started (season).
IV variants #2: Month, week, start # (in-season)
DV correction: rating(week) - rating(season) -------------- instead of rating(week)

19
by Sunil (not verified) :: Wed, 12/31/2008 - 8:19am

Am I reading this right? Drop-off of old QBs = 2.6% vs 2.5% for younger QBs. That's a differential of 0.1% .. is that even statistically relevant? IMHO - I just don't see any performance difference in the 2 groups.

20
by Papa Narb (not verified) :: Wed, 12/31/2008 - 1:04pm

Actually, if you look at the original breakdown, it looks like there is a "sweet spot" of age between 27 and 30 where there is no drop off in production. All the other age brackets show a 2.5% to 3.5% decline.

One could argue that in their prime years, QBs do not see a decline, but in any other age range, they do.

21
by ceolaf :: Wed, 12/31/2008 - 3:20pm

In fact, I don't think that this approach allows any such claims.

This analysis only says that the older and younger QBs who play in December and January are a little bit less effective than the once who play the rest of the year, and that the sweet spot QBs perform at the same level.

We don't know if individual QBs show drop offs or not. This approach does not support such inferences.

22
by Appaloosa (not verified) :: Thu, 01/01/2009 - 3:58pm

A couple of years ago I did a study on the careers of all Hall of Fame and probable future Hall of Fame QBs who entered the league after WWII. One of the things I was interested in was how age affects the quality of the QB. Since all of the guys in the study are among the elite, there is no bias for better QBs playing at a greater age just because they win that right be being great earlier.

My findings were that passer rating, completion percentage, and yards per attempt remain virtually unchanged for as long as an elite QB cares to play. However, after 16 seasons in the league (age 37 or so), total yardage and TDs per attempt begin to drop and really drop by season 18. I think the reasons for these observations are that older QBs are very savy, having seen virtually everything a defense can throw at them, but on the other hand the wear and tear on the arm and shoulder after 16 seasons in the NFL take a toll. Older QBs are generally smart enough to limit mistakes, but stamina has become an issue, which would indicate a tendency to fade a bit late in the season.

23
by Bright Blue Shorts :: Fri, 01/02/2009 - 9:03am

Sounds like a good study.

One query ... if Completion Percentage and YPA remains the same but total yardage drops; doesn't that just indicate that they threw less pass attempts?

Did you consider other factors as a possible causation for these changes? Older QBs tend to go for a last season or two in the sun (Favre - Jets, Montana - Chiefs, Unitas - Chargers); or be surrounded by ageing teammates.

24
by jjbtnw (not verified) :: Mon, 01/05/2009 - 5:50pm

Is the study supposed to show the effects of a long season ? In which case shouldn't the study only consider QBs with 10+ passes in all 16 games.
Also, the data presented indicates that quarterbacks do indeed tend to wilt in the late season as they get older. Once a QB gets to age 27 they show a definite decline as they get older.
The more interesting question seems to be "Why do young QBs also fall off over the long season ?"