This is our old blog. It hasn't been active since 2011. Please see the link above for our current blog or click the logo above to see all of the great data and content on this site.

More on second-and-ten

Posted by Doug on January 30, 2007

Yesterday I posted some data about team tendencies on second-and-ten. I also pointed out that that data might be tainted by some situational variables that are tough to account for. Here is another set of data that, at least in my mind, removes any doubt about whether the effect is real.

First three quarters of the game, point differential within +/-7:

Run percentage on 2nd-and-10 following a 1st-and-10 pass: 55.7%
Run percentage on 2nd-and-10 following a 1st-and-10 run: 29.2%

[NOTE: this table was corrected shortly after posting.]

I chose this slice of data because I wanted to remove clock management considerations from the equation as much as possible. If it’s a one-score game with more than 15 minutes to play, clock management should be a relatively minor factor in the run/pass decision.

Contrary to some of the commenters in the previous post, I am now convinced that this is systematically irrational coaching. Whatever the profile of your team, there must be some second-and-10 run/pass mixture (in game theory parlance, a particular “mixed strategy”) that optimizes your chances of getting a first down within the next two plays. I find it very difficult to believe that the one single play just before the second-and-10 could provide enough information to alter that optimal mixture so drastically.

No, more likely the reason is something suggested by Vince:

I think you’re overlooking something: most coaches (most good ones, anyway) don’t want to be one-dimensional and are seeking a balance between rushing and passing. Barring a turnover, you’re guaranteed three plays on any possession: Most coaches want to get at least one run in there, and 3rd and long is not the time to do it.

I actually wasn’t overlooking that; I was just considering it an example of irrational decision-making.

It reminds me of my younger days when I was something of a baseball player. I had a coach — and I don’t think he was unique — who absolutely positively would not tolerate a called third strike. If you swung and missed at three straight pitches that were over your head, he’d growl at you a little, but taking a third strike — even on a full count — was unpardonable. It was almost sure to get you benched.

My coach thought that a strikeout looking was worse than a swinging strikeout because a strikeout looking made you look apathetic or unagressive or something. And for some reason, an apathetic or unagressive out is worse than another out.

I think that for many football coaches a three-and-out with three passes is the equivalent of taking a third strike. It’s not just a failure; it’s the ultimate failure. Not because it makes you look apathetic, but because it makes you look desperate. And a desperate failed drive is apparently worse than an equally-failed drive with a run in there. The second-and-10 run is the coach’s way of saying, “Look at me. I am not in panic mode.”

Even though the end result is identical, they’d much rather have a pass-run-pass three-and-out than a pass-pass-pass three-and-out. That’s fine, but (unless clock management is a factor) three-and-out is three-and-out. If you can increase your chances of getting a first down by not “mixing it up,” then it’s hard for me to imagine that the benefits of mixing it up outweight that.

But all my rambling above assumes that 32/68 is closer to the optimal second-and-10 run/pass mixture than 57/43 is. The proof of that would be in the data. As commenter Jim A suggests, the next step is to look at the results of the different kinds of second-and-10 choices. I will attempt to do that with my next post.