Not for attribution my ass

(Form wanted me to post something else on my blog, but it wasn't terribly interesting, so instead I'm going to post our hot three-way conversation about applying sabremetrics to Football. If it doesn't make sense [more so than, say, everything else on my blog], it's because you're dum. I've copied it straight from my email, so read it from the bottom up or something... I don't really care.)

Well..according to my transcript, I took Calculus for Social Scientists, so I should be able to understand all of this. Unfortunately I do not. However here are some thoughts....

1)Mike's point about running the ball 30 times does not consider the sabremetric method. The panthers win when Stephen Davis runs the ball 30 times would be one of those stats that people seem to pay attention to, but means nothing. Mike is right to point out that this is more a symptom of the Panthers winning, rather than a cause. (What I forgot in Calc for SS, I make up in things I remember from Aesthetics and Hermeneutics.)

2)I think sabremetrics could never be applied completely to football, but the thinking could be used to discover new stats which tend to create wins. For example, are all interceptions equal? We count them as equal. But Tom Brady's interception really resulted in nothing different than a punt, although potentially costing the Pats 3 points. However, if he threw that interception from his OWN endzone, it would be much more costly. I stand by Mike and My Barry Sanders theory. Maybe something like "runs over 3 yards" or "carries without a loss" should count for more than yards per carry. While these stats might be more reflective on team play than individual play, it could lead to better coaching or drafting. Maybe a GM should pick the guy who runs straight ahead rather than the big play guy. Maybe calling sweeps are a waste of time.

3)Finally, sabremetrics could change how players are evaluated. If the stats that people pay attention to are different, maybe things like raw speed and arm strength are over rated too, like they are in baseball. No one can explain why Tom Brady is so good. He is slow and does not have tremendous arm strength. Well part of it is luck. Part of it is having a great defense. But I think his style/the way he was coached helps increase the odds of success. Dumping the ball off before taking sacks. Not taking excessive risks. Do offensive coordinators who have QB's with big strong arms gameplan to that and take excessive risks? Is Bledsoe's arm strength advantage negated by his inability to step away from blitzers (and his false belief in his own arm strength). You know Bledsoe had some great years with an All-Pro Tight End named Ben Coates. The offense started to stink when they made Coates a decoy to force the ball to Terry Glen the game breaking speed receiver everyone said he needed. there you have my thoughts. Any other stats lurking there to make more sense out of football?


Mike is right about sabremetrics (the British spelling) not making much sense for football if you don't take the situation into account. Of course, the response to that is to take situation into account. Just track some meta information about each play (reamaining time, field position, score differential, etc.) and you could do analysis filtered on that. If you've got that information available, it's really simple to plug in your current situation and evalaute the expected outcomes of each possible play. So you could massage some sabremetrics out of that data.

The tricky part might be assigning a stanadard value to the outcome of each play so that you could evaluate plays whose aims are different (gaining yards vs. running down the clock vs. kicking a field goal when you need a touchdown). But you could probably come up with some sort of situational utility function or at least a comparitive ranking of different outcomes if you sat and thought about things hard enough. It definitely is more complicated than in baseball, since the game clock necessitates discounting yardage and points.

So football is like a 5th order differential equation while baseball is a quadratic equation.



I think that's probably true. I read part of that study when it came out and it said that football coaches were too conservative. I agree, and Belichick takes advantage of that. When he has a chance to go for points, he takes it. He didn't sit on the ball in Super Bowl XXXVI in the last minute. He didn't let a 1st-and-20 on his own side of the field deter him from going for the winning score.

In baseball, one basic assumption behind sabermetrics (and Jeff can correct me) is that while situations change there's really no difference between the object of an at bat in the first inning or the ninth inning. But in FB the situation of the game changes the decision, run versus pass, playing to run clock rather than playing for points. If Carolina wins when Stephen Davis touches the ball 30 times, why not hand the ball off on the first 30 offensive plays.

I just don't think the same principles apply here.



As I was telling Jeff when we watched the Superbowl, this might be the closest Football gets to sabremetrics. While there might never be a way to fully rate players' win total contributions, there might be a better way to look at play calling.


Lest anyone doubt that Belichick is a genius --

Go Cowboys!


Popular posts from this blog

Are you acquainted with our state's stringent usury laws?

Eddie Vedder is Still an Incoherent Drunk