The Liar Stat

Comrade KJ has previously talked on this issue at a game by game level, this is just going further

On October 26th, Connor Cook headed into the locker room of Memorial Stadium with what would generously be called a bipolar stat-line. On 11 passing attempts, he'd managed 10 completions for 111 yards and a ridiculous 91% completion percentage. He also, on a healthy 7 rushing attempts (more attempts in that half than any full game of his other than USF), had managed a laughable -2 yards (what's known in college football circles as 'a Toussaint'). It was in many ways a near perfect dichotomy of a great passing and awful rushing performance that would lead anyone glancing at the halftime score-sheet to make some pretty solid assumptions about what kind of QB Michigan State was rolling out there and also to wonder quite why he was almost averaging a carry for each passing attempt.

The thing is, of course, that that stat-line was a lie, as indeed all college quarterback stat-lines have progressively become more and more of an irritating lie, and it's all thanks to the sack, or more accurately, how the NCAA has decided to score-keep the sack.

See, before he threw a 29 yard jump-ball to the defensive back covering Bennie Fowler that was improbably reeled in for a TD (itself a bit of a statistical fib that is much harder to correct for, and is definitely a subject for a separate post), Cook was felled by a pair of ugly sacks on first and second down for a total loss of 15 yards. These were two downs where MSU was clearly trying to throw the ball, and the scorekeeper correctly credited each as a sack, and yet when it came time to file '1 attempt, -6 yards' and '1 attempt, -9 yards' into the box-score they went, as they always do, into Cook's (and the team's) rushing column instead of his passing column.

To ask 'Why do they do it that way?' is almost unnecessary, because the answer is likely 'for no good reason'. To ask 'Well, what's the big deal?' is to uncover a multitude of annoyingly minor and annoyingly major bugs with how we view both quarterbacks and modern offenses, bugs that could be fixed with relative ease.

In short, a sack should not be considered a rush attempt. Furthermore, a sack should be, for the purposes of a boxscore, considered a pass attempt that is a. an incompletion and b. for negative yardage.

What does this relatively minor change do?  My friends, it shows us a nearer vision to the truth.

Sack adjusting the Michigan State Quarterback

Sack adjusting the Michigan State offense.

All that data can be compressed into something more manageable like so:

So we're talking tangible shifts of like, a perceived half a yard to three-quarters of a yard here, as well as an almost two and a half percent shift in perceived run-pass balance, all due to score-keeping these passing attempts as running attempts. In a world where more and more QBs are picking up big yards on the ground, properly score-keeping sacks also more accurately assesses that position as a running threat at a glance. Many of the stat questions in college football (yards in the air vs yards after catch on pass plays, QB rush yards on scrambles vs designed runs, how much is a QB really under pressure, even when he isn't sacked, defensive yards per attempt at the individual DB level) require a bunch of time-consuming film study. This fix doesn't.

In the future, I'm curious to look at sack-adjusted data beyond just the Spartans (someone like Devin Gardner is a much better case study than Cook, for example), at the Big Ten, and maybe even the national levels, to see if I can get a better idea of how much of an perception effect this has on college football as a whole. Off-season project, maybe.