Today we'll look at TFLs, but from the offense's perspective. Unless you're running the Hail Mary as your base offense, the key to moving the ball and scoring is getting 1st downs. And unless you're facing FSU with ACC Refs on your side, you'll need to get those 1st downs legitimately with positive yardage plays. Sounds simple enough. I provide a breakdown of TFLs into Sacks and Stuffs, and then calculate the percentage of plays that an offense incurred negative yardage on.
You could stay here, but I'd jump.
Here's a rundown of the above table: TFL.a is tackles-for-loss allowed by a team's offense; Sacks.a is sacks allowed; Stuffs.a is stuffs allowed; Rush.adj is offensive rushes with sacks removed; Pass.adj is offensive pass attempts plus sacks; TFL.rate is the percentage of total offensive plays that ended in a TFL; Sack.rate is the percentage of adjusted pass attempts that ended in a sack; Stuff.rate is the percentage of adjusted rushing attempts that ended in a stuff; and OSOS (FEI/Brian Fremeau/FootballOutsiders.com) reflects the overall quality of defenses, where a lower number indicates a better set of defenses faced. Baseline stats are from CFBStats.com.
I'm mostly providing these stats for your perusal, but below are some of my thoughts. Note that you can click on the column heading to sort by ascending/descending order within the table. We're cool like that here at TN.
The percentages range from 3.86% (Stanford) to 13.97% (Mississippi). This seems already to be a pretty good indicator of great and abysmal offenses, respectively. Think about those percentages for a second. If we assume a pace of 60 offensive snaps per game, Stanford averages about 2.3 TFLs against per game. Mississippi? 8.4 TFLs against. However, note the difference in qualities of defense (OSOS) faced: Stanford - 69th, Mississippi - 25th.
I ran a simple linear regression of the TFL.rate % on OSOS, trying to see if (and how much) the quality of defense faced determined a given offense's tendency to go backwards. The result was a significant negative relationship, but explained little variance of offense's TFL.rate (R-squared of 6%). This shows us that the strength of defenses faced was important but not at all dominant in describing an offense's anti-"splash play" rate.
Look at the bottom 5 teams - all SEC teams (boy, Florida's offense was pathetic this year). I don't do it here, but if we grouped the individual teams into groups by conference membership, I'm fairly certain we'd see that the R-squared would rise explain more variance seen - for instance - in the SEC.
Oklahoma's offense is the benchmark, incurring sacks on only 1.59% of all passing plays. At a 60-snap per game pace, that comes out to just under 1 sack allowed per game; and that coming against some better defenses (23rd). Ohio State's offense incurred a sack on an astonishing 15.27% of all passing plays. At a 30-pass per game offensive pace, that comes out to 4.5 sacks per game. That is an ugly number, allowing 40 sacks on 262 passing plays against the 28th OSOS. At least Pitt's 56 on 450 (12.44%) came against the #1 set of defenses faced nationally. Note that the Sack.rate could be an indicator of offensive line fail, but sacks also happen when you're receivers are covered well or can't get open. A dumb QB can also work himself into a sack by holding too long or escaping the pocket too quickly.
The R-squared value from a regression model of Sack.rate upon OSOS is even smaller than the TFL.rate/OSOS relationship (4.3%) described above. Though statistically significant, it again shows that OSOS doesn't explain the TFL.rate of a given offense. Thusly, I contend we're seeing the fail rate of offenses moreso than the success rate of defenses captured in these TFL derivatives.
I've probably been conducting the "Stuff" train a bit too long but it's fun to see who is overwhelming the point of attack, whether through scheming via run blitzes, winning 1-on-1's, or some combination of the two.
The #1 and #3 offenses incurring the lowest Stuff.rate are your traditional triple-option teams in Army and Georgia Tech, allowing 2.7 and 3.7 Stuffs on average for a 60-rush pace. This makes sense; they use play-action to pass and often don't even dropback passblock unless they're down by multiple scores (note how poor Army is at incurring sacks when passing - Georgia Tech is at least respectable). Sandwiched between them is Stanford. Damn if Jim Harbaugh didn't build up one of the best offensive lines in football. Where the offensive line goes...
Florida and Mississippi bring up the rear with a Stuff.rate of 17.16% and 17.91%, respectively. At a 30-rush per game pace, that comes out to 5.1 and 5.4 Stuffs per game average. That is horrendous. No wonder Weis went 'Jeff Bowden swing pass to the RBs' as a base offense. While both teams Sack.rate percentages aren't horrible, their overwhelmingly atrocious ability to run the ball explains their overall TFL.rate rankings. Interestingly, the Stuff.rate/OSOS linear relationship is not statistically significant.
All this talk about a linear relationship between TFL/Sack/Stuff.rate and OSOS is really just meant to serve as a caveat. What I mean is this: On any given play, the result is some mixture of offensive/defensive execution and failure. What I'm showing with these statistics may be an indictment of offensive ability (rather than defensive). The linear models help to extract the seasonal trend related to quality of defenses faced. However, since the R-squared values are so low we are led to think that the story is more about offensive failure (high TFL.rates) than defensive success.
How does FSU look?
Bad. FSU is very close to the bottom of the conference, careening neck-and-neck with Wake Forest toward the basement with a TFL.rate of 10.99% (Wake Forest - 10.18%). At a 60-snap pace, FSU's offense would incur about 6.6 TFLs.
We can break apart the TFL.rate statistic into Sack.rate and Stuff.rate. In Sack.rate, FSU is in fact the bottom-dweller of the ACC at incurring a sack on 9.04% of all pass attempts. At a 30 pass attempt pace, FSU's offense incurs about 2.7 sacks. In Stuff.rate, FSU is at least closer to the conference median at 8th with 10.99% of all rushing plays ending in a Stuff. Assuming 30 rushing attempts per game, this comes out to about 3.3 stuffs incurred per game.