Writers of Pro Football Prospectus 2008

16 Nov 2009

Curse of 370 question

I've just discovered this site in the wake of "the worst coaching decision in NFL history"gate, and I really like it, but the 370 carries articles were a little disappointing. From what I read, it basically says running backs who are very successful have a larger percentage decrease the following year than running backs who are less successful. It implies this is due to injuries from overuse.

To me it seems like a lot of the statistics could simply be from random variation. To take an extreme example, assume all the running backs in the league are exactly the same, but there outcomes in a given year are taken from some random distribution. Then the following year on another random draw, it is obvious that the top 10% from year 1 would have a bigger average decline than the top 20%-10%. Obviously running backs are not all the same, but I would argue much of the variation is independent of the prior year, in which case those with fortunate luck would on average decline more than those with not so fortunate luck.

Is there something wrong with my logic, has this been analyzed in a more satisfying manner? I would be interested in comparing total yards per carry of those with >370 compared to those with 300-369 in the following year rather than % changes from the previous year.

Posted by: mermel on 16 Nov 2009

2 replies , Last at 18 Nov 2009, 4:34pm by tally

1
Re: Curse of 370 question
by Raiderjoe :: Tue, 11/17/2009 - 12:29am

lots of rbs with 360-369 carrys in one year did not have big drop .

vvvery sneaky that FO use 370 as magic number.

2
Re: Curse of 370 question
by tally :: Wed, 11/18/2009 - 4:34pm

I presume a good portion of the drop results from regression to the mean. You don't give 370+ carries to a RB who isn't having a good year; it's likely that he isn't going to repeat such a performance.

I don't think 370 is really a magic number so much as a significance threshold. Certainly the degree of dropoffs increases with workload.

Login or register to post comments