Third-Down Give-Up Draws, Gary Kubiak, And You

Last season (I hated writing those words), it almost became a running joke amongst Battle Red Blog staffers that in any third-and-long situation, the Texans would turn to the draw because ... uh, well, nobody really knows. At first it seemed like it was because the defense was dominant, then later it happened because of the illusion that the defense was still dominant. Or something like that.

Either way, I spent about an hour in the handy dandy Football Outsiders spreadsheet today. Here's what I can tell you about third-down give-up draws, Gary Kubiak, and you.

First, I sampled the entire NFL season. This is every third-and-long (for my purposes, I'm using third-and-7 or longer) down for the entire NFL:

• There were 3,289 total third-and-long attempts by an offense this season. 257 of them were rushes. That's 7.8%.
• Of those 257 rushing plays on third-and-long, 27 of them resulted in a first down or touchdown. That's a success rate of ... 10.8%
• The other 3,032 plays were either passes or scrambles. 28.3% of them resulted in first downs or touchdowns.
• Finally, we account for the fact that passing has the potential to lead to a more negative play. On third-and-long, there were 83 non-Hail Mary interceptions and 272 sacks (or fumbles on scrambles). That's a negative-play percentage of 12.2%

Now we have our baselines. Let's look at how Dr. Kubiak managed things:

• The Texans faced 101 third-and-long situations. 23 of them were runs. That rate of 22% is nearly triple the league-average rate. A better way to say that? The Texans called nearly 9% of all third-and-long runs in the entire league last year.
• On those 23 third-and-long runs, the Texans converted ... one first down. That's a success rate of: 4.3%.
• On the other 78 third-and-long pass plays or Matt Schaub scramb--- hahaha no, just pass plays -- the Texans converted a first down or touchdown on 21 plays. Success rate: 26.9%.
• And then we hit the negative plays. Schaub threw three interceptions and was sacked four times on those unsuccessful plays. Negative-play percentage: 8.9%

It was hard to understand the rationale behind these moves before I did the research. Looking at more evidence just made it even more puzzling. Schaub did not have an inordinately poor rate of success on third down (over the course of the entirety of the season, anyway), and he was actually less likely to take a negative play than the average quarterback. Moreover, the Texans were extraordinarily terrible on their third-and-long runs. Small sample size theater, sure, but that's the nature of situational NFL stats.

And if we apply actual game theory and logic to this -- shouldn't a better defense make you want to take more risks on offense? Despite the fact that the Wade Phillips defense got plastered by New England twice and Green Bay once, it was easily the better unit over the course of the season. Leaning on that unit to bail out the offense could have enabled Houston to take more risks -- and risks don't have to be fourth-and-2 on your own 40, they can be as simple as "actually throwing the ball." It's allowed in the rules and everything.