Tuesday, November 17, 2009

Belichick Update

Greg Mankiw is also on Belichick's side, and links to an interesting story on Belichick's reliance on data.

I read a little bit more of the paper but am still not convinced that Romer's analysis translates to the Pats situation. Using data from 3rd downs and the 1st quarter just doesn't translate to 4th and 2 with 2 minutes to go. I don't think I can write anymore about it without getting some thicker nerd-glasses, but two quick points: At the end of a tight game like that, the returns to getting exactly two yards are exactly equal to the returns to getting 20 yards -- you win the game in either case -- so the defense is going to sit right on the first down line and not let you past. If it's the first quarter and a third down, the returns to 20 yards are much greater than the returns to 2, so the defense has to respect the deep ball and play off the line, increasing your chance to make a 2 yard gain.

The second point is that points come in important, discrete bunches (3 points for a field goal and 7 for a TD). This may have been addressed in the paper and I just didn't read it closely, but you would have to do the analysis looking only at the probability of the opposing team scoring a touchdown to replicate the Pats situation. If you are just averaging the opposing points scored the field goals would raise the average but with a 6 point lead the Pats could care less if the Colts kicked a field goal at the end of their drive.

No comments:

Post a Comment