Coaches on Stats

Tom Dienhart has an interesting article up in which he asks various college coaches to describe which stats they view as important. (HT: Blutarsky.) Generally speaking, I was taken aback by the lack of sophistication in the stats cited by the coaches, including coordinators with reputations as advanced thinkers. For instance, not one coach listed yards per play as a relevant measure, despite the fact that most sophisticated gamblers use it as the basis of the systems that they set up to analyze teams. Tim DeRuyter comes close to explaining why a pace-neutral stat like yards per play makes sense...

Total defense also is important. A lot of it is a function of how many plays you give up. In our league, you have Oklahoma, Oklahoma State, Texas Tech, Baylor -- all these guys are no-huddle, up-tempo teams, so you are going to play more snaps and total defense gets thrown out. Instead, I am more concerned about pass-efficiency defense and third-down defense.

but then he cites third-down defense (as do a number of other coaches), a stat that Football Outsiders has shown (at least on an NFL level) to be a heavy luck element. Teams that do disproportionately well on third down tend to regress the next year and vice versa. Conceptually speaking, it makes no sense to focus on third down efficiency because there's no reason why third down is any more important than second or first. Can anyone conceive of a good reason why a defense would be bad on first and second downs and then magically get better on third? Maybe a defensive coordinator saves up his best stuff for third down after playing vanilla on downs one and two? Put another way, why do we pay special attention to third down conversion defense, but not to fourth quarter defense?

The general sense that I got from the article is that college football is a barren wasteland for statistical analysis among coaches. It's possible that some of these coaches have advanced stats for measuring their teams' performances and they just aren't sharing those stats because they want to retain an edge, but there is no hint of that sort of collusion. I doubt that dozens of baseball managers would say that they use batting average and RBIs to measure players and if they do, most GMs almost certainly don't. Since college football programs don't have general managers, it falls on the coaches to seek out and use better stats, but it appears that they don't, although at least some of them are curious enough to have read studies on the importance of turnover margin.

And here's the funny thing about innumeracy on the part of coaches: they have a ready-made solution at their universities. Do you really think that there aren't a dozen students (or professors) in the stats department at Alabama who would be thrilled to apply their regression analysis tools to help Nick Saban understand what stats have the strongest correlation with winning? Or what traits are most important for recruits at given positions? I'm not saying that coaches should use these sorts of analyses to make all of their decisions. As in baseball, stats are best used in conjunction with traditional scouting and coaching, and that's in a sport that lends itself to statistical analysis more than football. Likewise, I'm not saying that coordinators need to start spouting off about adjusted yards per play or S&P. Their jobs are still primarily to develop schemes and then to coach players to perform their roles in those schemes. However, there can be no doubt that they would be more effective at their jobs if they had better guidance as to what numbers matter.