First a bit of a disclaimer: my play-by-play database isn't nearly as clean as I'd like. Although it's very close in almost all cases, it doesn't quite match up with the official stats for various reasons. It also isn't set up to easily do queries like those that I'll do in this post. So while I'd bet that everything I'm going to say here is very close to being true --- close enough that the substance will remain unchanged even if the numbers aren't precisely right --- I wouldn't mind having some corroboration from those readers out there who have their own play-by-play databases.
Over the years I have gotten the impression that, whenever a team throws an incomplete pass on 1st-and-10, they typically follow it with a running play on 2nd-and-10. So I decided to check to see if my impression was right. Here's the data, 2003--2006:
==== 1st-and-10 play ===
2nd-and-10 play Run Pass
Run 739 3374
Pass 1236 3841
So what this says is: when a team throws an incomplete pass on first down (or completes a pass for zero yards), they run 46.8% of the time on second down. But when they rush for no yards on first down --- leaving them in the same second-and-10 situation --- they run only 37.4% of the time on second down.
I find this pretty interesting, and my first reaction is to think that it's evidence of irrational --- borderline superstitious --- behavior. Second-and-10 is second-and-10. What does it matter how you got there?
Then I realized that there are all sorts lurking variables.
For instance, what kinds of teams are most likely to allow zero yards on a first down rush? Teams with good rushing defenses, of course. The same kinds of teams you don't want to rush against on second down. And likewise with offenses. If the Cardinals and the Chiefs rushed the same number of times on first-and-10, the Cardinals would have more zero yard gains on first down because they're a worse rushing team. And they'd have fewer rushing attempts on second-and-10 for the very same reason. In other words, good rushing defenses and bad rushing offenses are possibly (probably) overrepresented in the first-and-10-run column. They shouldn't run on second and ten. Likewise, good passing defenses and bad passing offenses are probably overrepresented in the first-and-10-pass column, and those are the kinds of teams for whom it's worthwhile to try a run on second-and-10.
On the other hand, I'd bet that teams that are ahead, since they run more, are overrepresented in the first-and-10-run column. And those teams should be running on second down more often than teams that are trailing or tied. Check this out:
Run percentage on 2nd-and-10 following a 1st-and-10 pass: 56.6%
Run percentage on 2nd-and-10 following a 1st-and-10 run: 46.2%
When tied or trailing:
Run percentage on 2nd-and-10 following a 1st-and-10 pass: 43.1%
Run percentage on 2nd-and-10 following a 1st-and-10 run: 31.0%
In both sub-categories, the split is bigger than it appears in the aggregate data, which indicates that this lurking variable might be making the split look smaller than it actually is.
This is yet another post where I find myself closing with: I'm not sure what the point of this post is, but I think it's interesting. Let me wrap it up with a couple of questions.
If we looked at all second-and-10 situations and could somehow (perhaps with a regression, but it would be tricky) properly account for the offensive and defensive strengths of the given teams against the run and the pass, and all the relevant situational factors, do you think that whether the first down play was a run or a pass would be a significant factor in predicting whether the second down play would be a run or a pass? If so, is that evidence of systematically irrational coaching? Or would there be a good reason for it?
This entry was posted on Monday, January 29th, 2007 at 5:03 am and is filed under General. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.