Some Thoughts Regarding the Run and Shoot That Have Nothing to Do with Mouse Davis

I lived in Arlington, VA for three simultaneously fantastic and miserable years. I grew to love a few of the area’s provincial aspects, including the John Thompson Show on 980 AM. Much like Steve “The Schmoozer” Somers, Thompson has a great radio voice, although the two sound completely different. Thompson makes a low rumbling noise, slowly and deliberately drawing out the vowel sounds.

Most, if not all, of the DC-area folks who read and write here seem to dislike the John Thompson Show, probably because they have different expectations than I when it comes to sports talk radio. Thompson knows a shitload about basketball, so when he gives an opinion it carries a lot of weight with me. Thompson admits that he doesn’t know much about any other sport, which may be a reason why others around here don’t like him. But I find it refreshing.

I remember an episode of the John Thompson Show where Thompson talked about learning from other coaches. He said that he learned a lot by talking to football coaches, because football and basketball have a number of similarities, including that one position has the ball in his hands on almost every play – the quarterback and the point guard.

Channeling his inner John Thompson, Mark recently opined during an editorial meeting that run-first quarterbacks are a lot like shoot-first point guards, and that he wouldn’t want either on his team. This inspired me to take a look at some numbers and see if this holds up.

First we have to figure out who the run-first quarterbacks (“RFQB”) are. I ran a search on pro-fooball-reference.com for all quarterbacks with 2000+ attempts since 1990. Due to the peculiarities of their search engine, I had to go back and fill in the gaps for players who started their careers before 1990. This also means that the search didn’t include players who had more than 2000 career attempts, but fewer than 2000 attempts after 1990, guys like Joe Montana and Phil Simms.

The search gave me 61 players. I then divided their number of pass attempts by their number of rushing attempts. My logic was that PFQB will have a small ratio of passes to runs because they run first. Here’s what I got:
 


(If you want to get a better look at all of the pictures in this post, right click on the picture and open it in a new tab. I'm too stupid to figure out how to properly size them.)

This looks about right to me. The guys who seem like they should be RFQB have the smallest ratio of passes to rushes, and the guys who can’t run have a higher ratio. As an aside, I started with a search going back to 1920 at the start of this nonsensical project and got too many players to cleanly fit in a graph. Including those extra players didn’t change much, but there was one massive outlier. Joe Namath has a staggeringly statuesque pass/run ratio of 52.99, which is 20 more passes per rush than the second highest (Ken Stabler with 32.14). Vick is also a major outlier going back to 1920. The only players with a ratio within 2.5 P/R of Vick are Kordell Stewart, Randall Cunningham, Steve Young, and two guys named Tobin Rote and Greg Landry.

Here’s how the numbers look graphically:





What I found interesting is that there appears to be two types of players with a low ratio: guys who run the football well and guys who aren’t good pro quarterbacks. I suspected that guys like Carr, Mirer, and Peete have so many rush attempts because they couldn’t find an open receiver, got shook, and ran for cover. Surprisingly (at least to me) there’s a pretty linear relationship between yards per carry and the pass/rush ratio (R2 = 0.72!) so guys like Carr, Mirer, and Peete really can be considered RFQB. This makes sense – if you’re good at running, you run more often:



Based on the foregoing, I will (somewhat arbitrarily) define RFQB as QB’s with a pass/run ratio of less than 9.0, giving us: Vick, Stewart, Cunningham, Young, Garrard, Culpepper, McNair, Harbaugh, Hostetler, Carr, Blake, Garcia, Gannon, Brooks, Mirer, McNabb, and Peete. Brunell, Elway, Banks, and Plummer miss the cutoff even though I think of them as “mobile” QB’s, if not RFQB, so the 9.0 cutoff clearly isn’t perfect. Luckily this isn’t a Ph.D. thesis and I’m not getting graded on it.

Now that we identified the RFQB, we need to figure out how good they are. Passing TD and INT alone fail to take into consideration the benefits and risks of RFQB, namely rushing TD and fumbles. So I made graphs for total TD (passing and rushing) and total turnovers (INT and fumbles):





I don’t think there are any majorly serious trends here. I also graphed total TD vs. total turnovers:



This graph shows that Jeff Hostetler is a tremendous outlier and the most boring QB of the past 20 years. Daunte Culpepper is the statistical opposite of Jeff Hostetler, even though they are both RFQB. Rick Mirer, Tony Banks, David Carr, Rodney Peete, and Trent Dilfer suck. Steve Young is superhuman (especially when you consider that he played before the current pass-friendly rules), while Peyton Manning, Tom Brady, Philip Rivers, and surprisingly Jeff Garcia are really good.

And I draw these conclusions with because I undertook the monumental effort of calculating the win percentage for all 61 quarterbacks evaluated (pro-football-reference.com gives you a win-lose-tie record but not a winning percentage) and found that (not surprisingly) the ratio of total TD to total turnovers is somewhat tied to winning percentage:



So then I plotted the total TD:turnover ratio against the pass:run ratio:



This is a pretty random distribution. Passer rating vs. pass/run was pretty random too, although the highest rated QB’s are not RFQB, except for Steve Young:



At this point in the analysis I wised up and realized that the simplest way to answer the question at hand was to make this graph:



This looks like about as random a distribution as you’ll get. You can win lots of games with a RFQB, and you can lose lots of games with a RFQB. The same is true for immobile QB’s. So my conclusion is that there’s nothing wrong with RFQB.

There are a lot of flaws with this analysis. For starters it compares complete careers with incomplete careers. Michael Vick didn't get old and slow yet and I imagine that his pass:run ratio will increase over the course of his career. If you only look at the first eight years of Randall Cunninham’s career, his pass:run ratio is 4.47. As he got older and slower he ran less and threw more. From 1997 to the end of his career his pass:run ratio more than doubled to 9.46 (although still very low). This makes sense given that he could throw to Cris Carter, Jake Reed, and Randy Moss or hand off to Robert Smith.

Which leads into another flaw inherent in almost all football analyses: it doesn’t take into account the quality of the QB’s teammates. Check out Tom Brady’s numbers in 2006 compared to 2007:



2006 was a good year for Brady but 2007 … laser show. The main difference between 2006 and 2007 (in my inexpert opinion)? He threw to Randy Moss and Wes Welker instead of Reche Caldwell and whatever was left of Troy Brown.

The same can be said of Kurt Warner. Look for yourself. He was a Hall of Fame caliber QB with Isaac Bruce, Torry Holt, and Marshall Faulk or with Larry Fitzgerald, Anquan Boldin, and Edgerrin James’ ossified corpse. He was merely average with Amani Toomer, Ike Hilliard, and Jeremy Shockey.

This analysis doesn’t include playoff numbers and it doesn’t include any sort of clutch factor, so Tony Romo looks better than he is and Ben Roethlisberger doesn’t look as good as he is.

I’m sure you can find other flaws, if so feel free to point them out in the comments.

And as a coda to this diatribe, passer rating doesn’t have much to do with winning, except at the extremes:



We have three losers with bad ratings, six winners with high ratings, and a blob of everyone else in the middle.

Stay tuned for Part II, Shoot First Point Guards.