Writers of Pro Football Prospectus 2008

10 Oct 2006

FO on BSMW: Rundown

But, what about Randall Cunningham and Steve Young?
Burned to toast?
Vaporized to milk shake?
Or... escaped to ESPN?!?

Until we know, we'll have to see whether running quarterbacks do, as Bill Simmons hypothesized three weeks ago, flame out quicker than slower ones. Like Shredder.

Posted by: Bill Barnwell on 10 Oct 2006

35 comments, Last at 13 Oct 2006, 1:57am by B

Comments

1
by Vince (not verified) :: Tue, 10/10/2006 - 7:08pm

"Pretty obvious that even if rushing quarterbacks do throw fewer interceptions than slower ones, it's because they throw fewer times, not because they're better at eluding the rush. There goes my hypothesis."

What are you talking about? In the table listed, the rushing QBs average MORE attempts per INT than the slow QBs in their first six seasons, and then again in year 8.

Other than that, this was good stuff.

2
by Bill Barnwell :: Tue, 10/10/2006 - 7:35pm

My bad - forgot to edit that. Change should be going up shortly. Goober.

3
by NewsToTom (not verified) :: Tue, 10/10/2006 - 7:39pm

So, running QBs peak as fantasy players in their first six seasons in their careers, but don't fall off as QBs until year 11. And yet some people wonder why I don't play fantasy football because I think it's stupid. Takeaway point: BS is right when it comes to fantasy value, but wrong when it comes to NFL value? Takeaway query: do most people (a) recognize there's a difference and/or (b) care?

4
by Moridin (not verified) :: Tue, 10/10/2006 - 7:43pm

Yeah, but they throw less INTs, so attempt per INT * # of int is lower

4
by Pat (not verified) :: Tue, 10/10/2006 - 7:43pm

Yeah, I agree - I don't understand the interception bit. It looks pretty clear that mobile quarterbacks throw fewer interceptions.

Makes sense, and it helps explain why the quarterback with the best interception percentage in the NFL currently is a mobile quarterback. Also helps explain why it's one of the few QB categories where Vick is above average.

6
by Tally (not verified) :: Tue, 10/10/2006 - 7:52pm

That's because Vick has such terrible touch on his passes and throws such a bad ball that even the other team has trouble catching it.

7
by Vern (not verified) :: Tue, 10/10/2006 - 8:02pm

One theory I didn't see mentioned is that "fast" QBs might just be better at surviving their first few years of learning the pros just by running around -- whereas less mobile QBs have no such option to keep them in the game.

In short, there could be a kind of "natural selection" going on here that leads to the perception. Just looking at guys who made it to 1500 attempts won't capture that effect. You'd have to look at all fast vs. slow QBs overall to see how many of them make it.

8
by peachy (not verified) :: Tue, 10/10/2006 - 8:22pm

Two factors I'd be interesting in seeing analysed - sacks (or sack rate), and size (when people talk about 'rushing quarterbacks' in the current NFL, I think they usually mean Vick, and he's definitely on the small side compared to players like Culpepper or McNabb.)

9
by Theo (not verified) :: Tue, 10/10/2006 - 8:25pm

Spank my ass and call me Charly, but where's the comment about the rediculous peek in the stats of the fast QBs in season #13?
Whose season was it?

10
by Pat (not verified) :: Tue, 10/10/2006 - 8:43pm

Brunell's and Cunningham's. It's mentioned in the article.

11
by karl (not verified) :: Tue, 10/10/2006 - 8:51pm

that was probably cunningham's sick season in minnesota.

12
by DavidH (not verified) :: Tue, 10/10/2006 - 9:01pm

#3:
YPA and QB Rating are rate stats, while Fantasy Points is a counting stat. My guess is that this fact pretty much completely explains the difference between the "real" stats and the "fantasy" stat. Running QB's perform at a fine level when they're in the game, but get beat up and miss more time than the Redwoods.

Bill-
Would have been nice to see FP/game and/or some kind of traditional counting stat, so that the "real" and "fantasy" stats were more comparable. Otherwise, interesting article.

13
by DavidH (not verified) :: Tue, 10/10/2006 - 9:08pm

Bill-
Also, you should probably define how you calculated your fantasy points. I'm sure I can figue it out if I look at the data dump, but I shouldn't have to.

14
by Travis (not verified) :: Tue, 10/10/2006 - 9:22pm

I wish the NFL followed common sense and counted QB kneels as team rushes, which would make studies involving QB rushing easier. Otherwise, Peyton Manning (who has averaged about a kneel/game since 2000, looking at the FO stats, and hasn't been over a normal rush/game since 2002) and QBs like him would probably be included in the immobile group.

Was any adjustment made for the oddness of Doug Flutie's career? Not many QB's have an 8-year gap between NFL seasons.

15
by andrew (not verified) :: Tue, 10/10/2006 - 10:22pm

The 13th or later spike for running quarterbacks... Tarkenton was arguably the first running quarterback (although I don't consider running the same as scrambling, but for the samke of this article they're both fast), his best years were probably 1973-1976, his 13th-16th, with his best being 1975 when he wsan the MVP.

None of these stats factor in the quality of the o'line, changing teams, etc. Tarkenton's late spike was due to his going back to a Vikings team that was much better than anything he's ever played for before. Then again, part of the reason he was a rushing quarterback was his original lines weren't very good...

Culpepper's decline is directly related to the Vikings offensive line faling apart last year (Birk & Dixon gone), and now going to a MIami team with an inferior line.

Steve Young was much better with San Fran than Tampa, and so on...

16
by the fumble (not verified) :: Tue, 10/10/2006 - 10:55pm

I did a very similar study before the season with a much smaller sample size and ran correlations between years experience and ypa, ypc, run %, and cmp%. Here's the results.

(YR EXP VS. RUN% -0.84)

(YR EXP VS. YPC -0.91)

(YR EXP VS. CMP% 0.58)

(YR EXP VS. YPA 0.49)

Here's the QB's I used (Douglass, Vick, Kordell, Rote, Landry, Cunningham, Young, Culpepper, McNair, McNabb, Staubach, Elway,D Williams, Moon.

If a QB wants to stick around (historically speaking), he learns how to pass. I think you're onto something #7. I think that fast qb's tend to run before they have to learn the offense. Eventually defenses can expose this though, so the QB better adapt.

17
by Bill Barnwell :: Tue, 10/10/2006 - 11:35pm

I wasn't aware there were different permutations for fantasy points out there! I used (pass yds / 20) + (rush yds / 10) + (pass td * 4) + (rush td * 6) - int.

18
by DavidH (not verified) :: Wed, 10/11/2006 - 1:17am

Thanks Bill. Yeah, there are tons of different scoring systems. I don't know about other sites, but Yahoo lets the commish set the scoring however he likes. For example, I'm in a league where passing and rushing attempts are negative, and the points for yardage is higher, which ends up having the effect of rewarding high YPA and YPC. But anyway, the default scoring in Yahoo, which is the one I bet a LOT of people are most familiar with, is:

Passing Yards (50 yards per point)
Passing Touchdowns (6)
Interceptions (-2)
Rushing Yards (20 yards per point)
Rushing Touchdowns (6)
2-Point Conversions (2)
Fumbles Lost (-2)

At any rate, most scoring systems are pretty similar, with just varying weights for yardage and TD's, so using a different scoring system probably would change the results very little.

19
by Andrew (not verified) :: Wed, 10/11/2006 - 2:17am

Bill:

I don't understand how you can come up with a definition of running QB's that excludes John Elway (#6 all time in QB rushing yards) and Rich Gannon.

20
by Andrew (not verified) :: Wed, 10/11/2006 - 4:27am

That data after year 10 screams "Outliers! Outliers! Outliers!" Specifically, it looks like a bunch of your rushing QBs haven't hit a certain year threshold yet and one or two good performances at year 13 and 14 allowed the performance of rushing QBs to spike dramatically. It'd be interesting to put n values for each of your data points, in order that your readiers might be able to distinguish data obtained from small sample size from data obtained with the larger sample size.

21
by Josh (not verified) :: Wed, 10/11/2006 - 6:03am

Do the rushing attempts by the "fast" QB's decline after a certain year? I'm thinking of McNair, Young, Brunell, Plummer, Elway, etc as QB's who were initially running QB's but who changed their style.

22
by Bill Barnwell :: Wed, 10/11/2006 - 8:29am

Andrew,

the data dump does include the n values at the far left of each column under the "#" field.

As for Gannon and Elway, well, they simply didn't run as much as people thought they did. They each had one year where they ran a lot - Elway 1987, Gannon 2000, but otherwise Elway's rushing numbers are a guy who runs a decent amount for 174 years, while Gannon didn't get the PT. I'll include their carries per game numbers below.

Gannon
0.00
1.33
3.71
2.87
3.75
2.63
4.00
3.00
3.67
3.58
2.88
5.56
3.94
3.13
0.86
1.67

Elway
2.55
3.73
3.19
3.25
5.50
3.60
3.20
3.13
3.44
2.83
2.75
4.14
2.56
3.33
3.13
2.85

Compare that to, say, Daunte Culpepper's numbers:
3.00
5.63
6.45
6.56
5.14
5.56
3.29

23
by Bill Barnwell :: Wed, 10/11/2006 - 8:52am

Josh,

If you look at the average attempts per season at the end of the column in the rushing quarterbacks section, the average number of attempts drops down by a third in Season 9 and never gets higher than that again.

24
by DB (not verified) :: Wed, 10/11/2006 - 11:28am

RE 18 : I'm pretty sure that Mr. Barnwell was being facetious when he said that he didn't know there were other schemes out there.

25
by Andrew (not verified) :: Wed, 10/11/2006 - 11:41am

Bill:

If your definition was 4 rushes per game in 2 years, Elway and Gannon would be in.

And I don't see how Flutie qualifies. He only had one year - 1999 - running alot.

26
by Andrew (not verified) :: Wed, 10/11/2006 - 11:50am

Bill:

Reading over your article, your methodology doesn't quite cover the same ground as what Bill spoke of, since you look at careers, and he looked at years starting.

By years starting, the following guys fell off a cliff around years 5-8. The number of good years is listed after their name.

Blake (7), Brooks (6), Brunnell (8), Culpepper (5), Cunningham (5, 7, 9), Garcia (5/6), McNabb (7), McNair (8), Stewart (7), Young (8 in SFO).

Only Brunnell and McNabb appear to have picked themselves up. If you also include Elway, he collapsed after year 9, then came back. Gannon collapsed after 8 years of starting.

27
by andrew (not verified) :: Wed, 10/11/2006 - 1:02pm

Culpepper's rushing numbers include a lot of the "one-yard sneak", which he was IMO one of the best in the league at. These kind of carries are different than the drop back, get flushed out and take off type of runs, and both are different than the designed draw or naked bootlegs...

28
by Bill Barnwell :: Wed, 10/11/2006 - 3:07pm

Andrew,

I don't think Simmons was factoring in years Jeff Garcia spent on the bench as years that were taking time off his career. It wasn't his point -- that rushing QBs age due to tear, not wear.

As for the other guys, I don't have the numbers in front of me, but I'll see how the group expanded with the four-carry threshold. I feel like that might be a little low.

As for Flutie, he averaged 5.9 attempts in '99 and 4.7 in 2003.

29
by Kevo (not verified) :: Wed, 10/11/2006 - 4:43pm

You say that McNabb is screwed around season 11, but one of Simmons's points was that he seems to have learned to survive without running. He now rarely runs, and I'd guess it's definitely less than 4.5 attempts per game (kneeldowns aside). Isn't there any way to chart the progression of rushing attempts over the whole career as opposed to just picking two seasons where a guy rushed a lot?

30
by jason (not verified) :: Wed, 10/11/2006 - 5:43pm

My question is, isn't there a lot of bias in the way that these groups were selected?

For example, take the 1500 career attempts threshold: seeing as how, theoretically at least, running qb's would seem to have fewer pass attempts, it seems to me that this criteria would automatically weed out a lot of the "flame out" qb's bill was talking about.

What's more, this criteria would also result fewer "recent" qb's, such as vick from being included in the study - this would seem to eliminate a lot of the qb's bill had in mind when he originally made the statement

Another possible selection bias would appear to be the rushing attempts per game: wouldn't it make more sense to differentiate between running and pocket passers by looking at their season rushing yards?

31
by Peter (not verified) :: Wed, 10/11/2006 - 6:25pm

30: Simmons specifically mentions 8 QBs, all of whom are included in the study. I don't think he's talking about players who play 3 years, suck, and don't play anymore. Wear isn't a concern, and even tear is more catastrophic problems that can happen to slow QBs as well. Think about it... 1500 attempts, over five years of starting is only 300 attempts per year, which is not a large number. Even QBs who don't throw that much would reach 1500 by season 6 or 7, and Simmons is only interested in people playing that long.

On the same note, what use does Vick have for the study? He only has 4 years of experience, how can we talk about what happens to him after 6 or 8?

I sort of see your point on the rushing yards per season, but it doesn't really matter if the player rushes WELL, just if they do it a lot. The idea is that they take a beating by running repeatedly, not by running long distances.

32
by NewsToTom (not verified) :: Wed, 10/11/2006 - 6:57pm

Re #29
Take a look at the data dump at the end of the article. Rush attempts are at their highest in Y3 and fall as time goes on. There's a noticeable blip upward in Y8, but I suspect that's connected to three QBs dropping out (see N falling from 10 to 7, then remaining constant for 4 seasons). McNabb has developed into more of a pocket passer, but this isn't interesting because that's what pretty much every rushing QB does.

33
by Bill Barnwell :: Thu, 10/12/2006 - 2:54am

Setting the threshold to 4.0 would add the following guys as rushing QB's:

- Elway
- Gannon
- Jim Harbaugh
- Rodney Peete

Does Harbaugh really belong in this discussion? I wonder...

The other thing is the reason that I chose 4.5 is because there's no quarterback I could find who has a season over 4.5 carries per game and ISN'T a running quarterback - that, to me, seemed like a very good definition of a running quarterback. On the other hand, there are lots of quarterbacks who have seasons where they average over 4 carries per game, including:

- Tony Banks
- Dave Brown
- David Carr
- Boomer Esiason
- Jay Fiedler
- Don Majkowski
- Jim McMahon
- Rick Mirer
- Warren Moon

I feel like, even if we were to include people who only specifically had two seasons at 4.0, the grouping might not represent what it's supposed to.

The only thing that might have affected the data set is that the PFR database may list 12 games played for a player when all he did was hold a snap on one play or, alternately, down the ball. I actually prepared a spreadsheet for the QB games played projection system in this year's book that has the "true" games played data, but it's only for 2001-2005. It usually isn't TOO far off, but in 2005, for example, 22 of 32 backup quarterbacks had incorrect games played data.

34
by JDawg (not verified) :: Thu, 10/12/2006 - 12:14pm

How can you make a comparison like this without checking if any of these differences are statistically significant, especially with such a small sample size? Otherwise, you could conclude that being either a fast or slow qb makes you better than an average qb (at least in terms of fantasy points). My guess is that if you put in error bars all of these "differences" evaporate.

35
by B (not verified) :: Fri, 10/13/2006 - 1:57am

For the "rushing QBs throw fewer Ints" I have a theory on this. When nobody is open, a stationary QB has two choices, throw the ball away, or force a throw. A mobile QB has three choices, the first two or scramble for yards. With that extra choice, it's likely that the QB would force less throws, which would be less Ints.