Tuesday, September 02, 2008

A new kind of football stat

OK, not NEW, per se. But new to me...

I'm a stats geek, and I love them as they relate to baseball, to which they are perfectly suited. I think basketball is wide open for this kind of stuff, but needs new measurements--it can't be done just from what's in the box score. As for football, its so tough, because each play is a situation, and results differ in value based on the situation.

This exact problem was tackled in The Hidden Game of Football. They devised a measurement tool for offenses that bases the evaluation on the system.

For example, on first down, you need to get at least 40% of the yards for a first down, because, logically, if you got that on every play you'd make a first down. That play is one point and a winning play. Losing more than three yards deducts a point, and a turnover is -4. Making more than 80% of what you need is +2, and you get more points for big plays.

On second down, you only have two more plays left, so to get your winning play and +1 you need to get 60% of the yards you need for a first down. Other bonuses are the same as for first down.

On third and fourth down its simple. You need to get the first down to get a winning play and a point, though there are bonuses for bigger plays. An eight yard gain on third and 20 is worth no points, whereas in traditional stats, it helps a player quite a bit.

The best thing is that all you have to do is evaluate each play on these standards, and you come up with some pretty good measurements. (Note, you also leave out plays where one team takes a knee, and you an evaluate a sack as a passing loss, which it is).

So, I did this for the Pitt game and I got the numbers. I don't know what they mean until I get some more context, but let's take a look.


They ran 86 plays, 55 passing and 31 running. They had 39 winning plays and 47 losing plays, for 45.3%, which I suspect will end up being pretty good as we look back. They ended up with a net of 63 points, with is .73 points per play, which is probably going to end up as being not quite as good.

On running plays, they averaged .77 points per play and on passing .71 points per play, so they were pretty consistent.

Out of curiosity, I was wondering about their trend line. When the game was 14-0, they had run 33 plays and had 35 points, which is averaging a successful play on each play. Which means from there on in, they ran 53 plays for 28 points, an average something more on the order of .5 points per play.

I'm not sure that reveals a lot that I didn't see in person, but we're still learning here.

Bowling Green

As with the conventional stats, BG did not perform as well. I counted 57 plays we ran, 22 rushes and 35 passes. 22 of our plays were a win and 35 a loss, with is about 39%. We notched 33 points on those 57 plays, which is .57 per play...less than Pitt for the game, but better than for the period where it went from 14-0 to 27-17.

Our running plays netted only 10 points, which is .45 per play. Passing was much better....23 points and .657 points per play.

Well, like I said, this is a project for the year. I'm interested to see how it develops once the season starts.

One last point--obviously, the defensive measure is just the reverse of the opponent's offense.

No comments :