In the hopes that these pieces will eventually lead us to understanding all three phases of the game in the context in which it’s actually played, I thought we would zoom out today to consider Tennessee’s offense on the whole. Using 2015’s play-by-play data - which, ya know, has some relevance to the 2016 team with the same coach, same OC, and same skill players - what are Tennessee’s offensive tendencies in different down and distance combinations?
Prepare to be shocked...
Under Mike Debord, Tennessee likes to run. A lot. After manually scrubbing the overall play-by-play data set across 13 games (I likely mislabeled a play here or there, but the trends should hold), UT ran the ball on 578 of 965 plays considered in 2015, good for 59.9% overall.
As has been pointed out in other weeks, though, that result doesn’t mean much without some additional context. Let’s start by looking simply at the run/pass split by down for the Tennessee offense in 2015.
This chart does nothing to disabuse us of the notion that Debord can be a bit predictable at times. 72% of the time on 1st down, 301 times in 418 tries, Tennessee ran the ball. This does not account for the exact play call (zone read vs. power vs. speed option to the short side of the field vs. Dobbs scramble, etc.), but there’s a pretty clear tendency here.
It’s interesting to note that the percentage of plays dedicated to passing increases on 2nd and 3rd down, but there’s a good reason for that (namely that getting behind schedule on 1st or 2nd down makes a pass more expedient on the subsequent play). That says more about footbaw strategery than any sudden surge of inventiveness on the part of the offensive staff.
Ok, you say, so we ran a lot. Big deal. On average, maybe the runs were more effective.
Strictly looking at the mean...that’s not true. Yards per play are higher for passes than runs across downs last year. That said, I maintain that we can’t really learn a lot about the variability and what we should really "expect" in any given situation without looking at the distribution of outcomes, median, and variance. So what does it look like if we focus only on those first down play calls, the vast majority of which occur on 1st and 10?
As General Neyland once (allegedly) said, "When you throw the ball, three things can happen — and two of them are bad." We see that reflected on the chart by the large spike at 0 yards reflecting, mostly, incomplete passes. If we dig into this chart a bit more using the percentiles we talked about in week 1, we can see what a 1st down run versus a 1st down pass really looks like.
- 25th Percentile: 2 yards
- 50th Percentile: 4 yards
- 75th Percentile: 7 yards
- 25th Percentile: 0 yards
- 50th Percentile: 0 yards
- 75th Percentile: 7 yards
Well now we at least somewhat understand why, given Butch’s and Debord’s oft-discussed preference for limiting risk, we like to run so much (or at least did last year)! The median run on first down in 2015 was more successful than the median pass on first down. This was also true at the 25th percentile, and the 75th percentile play by ground or by air was the same.
Could you use this knowledge in a game?
We toyed with the concept of game theory in the comments section a few weeks ago, and the high likelihood that Tennessee will run on 1st down provides an interesting case study. I am not an expert on the subject, but in a football context, it would go roughly like this.
Saban knows that Debord runs on 1st and 10 more than 7 out of 10 times. The "rational" call for Saban to make is a defensive scheme geared toward stopping the run. Debord knows that he runs it 72% of the time on 1st and 10, and he suspects that Saban has been watching film and is well aware of Debord’s tendency to run it on first down. Given that, Debord expects that Saban will come out in a defense designed to minimize a running play.
Should Debord call a pass to take advantage of the defense’s weakness? What if Saban knows that Debord knows that Saban knows?
Calling plays as an offensive or defensive coordinator could be considered a simultaneous game (both sides do get to see alignments before the snap, but the individual executions are unknown). Yes, you’re trying to maximize your team’s strength and minimize its weaknesses, but you’re also anticipating what the other side will do and adjusting your calls accordingly to increase the chances of offensive success.
At least, that’s the idea. In practice, Tennessee ran 19 times on 28 1st down plays against Alabama in 2015, with 68% being pretty much right on the season average...soooo Debord tried to confuse Saban by calling his normal game. That’s a bold move. Let’s see if it pays off!
We’re still only scratching the surface. What were the distributions of outcomes for run and pass plays on other downs? Did the balance of calls shift dramatically depending on who the opponent was? We’ll see!
Got something you’d like to see in a future week? Shoot me a note or use the comments section below and I’ll be more than happy to steal your idea and pass it off as my own.