Sacks Ain't Rushes

I like statistics. And I like football. But football statistics don't do much for me.

Sure, individual football stats are fun. We all know what 100 rushing yards for a running back or 300 passing yards for a quarterback mean, but there's no elegance to the overall statistical framework of the game the way there is for basketball.

This is really just a function of football being a very complicated game. The specialization among the players and inherent team-reliant nature of individual performances make statistical comparisons between players less valid. And the large degree of variance between the different situations in which plays are undertaken creates a lot of skewing. A team that plays with a lot of big leads will select its plays a lot differently than a team that has to play from behind all season.

The good people at Football Outsiders have done a magnificent job of accounting for all this in their advanced metrics, but those metrics are fairly clunky in terms of providing an intuitive sense of on-field performance. (Example: MSU was 8.0% good on defense last season.) That makes them hard to use in writing intended for general audiences (or at least audiences unwilling to spend more than 15 minutes studying the statistical glossary).

I can offer no comprehensive solution to this quandary, so instead I'll suggest a more modest adjustment: stop counting negative sack yardage in rushing totals. College football box scores classify all yardage gained/lost as either rushing or passing. If you don't attempt to pass the ball, then the result of the play goes into the "rushing" category--even though a quarterback who gets sacked clearly wasn't attempting to rush the ball.

And it's not as if scorekeepers aren't making that judgment--sack yardage is specifically tracked in the box scores and does not include all instances where a quarterback gets tackled behind the line. A loss of yardage on a clear designed quarterback run isn't scored as a sack. So just go ahead and count the sack yardage against the passing numbers, like they do in the NFL. The reality of any given football game is complicated enough without calling some of the plays conducted in the game something they're clearly not.

(I'm not the first person to recognize this idiosyncrasy of college football scorekeeping. Brian normally makes these adjustments in his treatise-length game previews. But I am, to my knowledge, the first person to write up a full blogger rant on the topic.)

If we take this philosophy, apply it to the offensive numbers Big Ten teams put up last season (conference games only, Gasaway-style), and (of course) scatterplot it, we get this:

Note that the scales the horizontal and vertical axes run along are closer to each other than you might have guessed. Football fans are often pass crazed--why not throw for an average of 7 or 8 yards per attempt instead of running for 4 or 5 yards per attempt? Adjusting for sack yardage leaves the differential between average yards per attempt at just a single yard (6.1 vs. 4.9). That remaining difference can be explained away by (1) the risk of throwing an interception, which isn't accounted for above, (you can fumble the ball either way) and (2) the fact that running the ball generally gets you positive yardage more consistently than passing it does, which is useful for accumulating first downs and keeping drives alive.

Your major outlier here is Ohio State, which somehow allowed opposing defenses to sack the quarterback on 21.1% of intended passing attempts (pass attempts plus sacks). The traditional stats credit the Buckeyes with 6.6 yards per pass attempt and 4.5 yards per rush attempt. As illustrated above, though, they clearly experienced more success when they attempted to advance the ball with their legs than with the arm of Braxton Miller or Joe Bauserman.

OK, now to some 2012 Big Ten Preview-type substance:

• Candidates for team improvement on offense include teams with quarterbacks entering their second years as full-time starters: Iowa, Minnesota, and Purdue (I'm applying college basketball theory to the football side here, and being generous with the hated-by-a-deity Purdue quarterback situation). Those teams all generally return a good number of offensive starters, as well.
• Ohio State could be very dangerous if they reduce their sack rate to anything within normal operating parameters. And I think that's pretty likely with Urban Meyer now at the helm. The second highest rate in the league last season was less than half Ohio State's (Northwestern, 9.6%).
• Expect MSU to move to maybe the midpoint between Iowa and Michigan. The run blocking should be better and LeVeon Bell will get the bulk of the carries from the get go; the passing game has to regress somewhat, MAXWELL HYPE to the contrary. (Did you know that Andrew Maxwell got bored with memorizing the offensive playbook this summer and memorized the collected works of Shakespeare, too? Get thee to the wrong part of the field, opposing safeties!)
• Wisconsin has got to come back to the pack. They return only 4 starters on offense and saw 4 assistant coaches on the offensive side of the ball depart for greener pastures during the offseason. I'm really hoping Wisconsin becomes the poster child for a new top-notch-assistants-are-the-key-to-everything meme a few months from now. Also: Bielema's ACC-BOT 2012 cannot, by the sheer laws of mathematics, be as efficient at QB as ACC-BOT 2011 was.
• Five of the top six offenses (those closest to the upper, right-handed corner) were in the Legends North Division last season. That could be true again this season, depending on how much Northestern regresses sans Persa and whether Purdue and/or Minnesota can make substantial efficiency leaps.

I'll post the comparable defensive scatterplot in the near future. For now, here's a bonus offensive scatterplot, just cuz: