from Amazon
Thursday, June 03, 2010
No, no, no, no, NO! (Part 2)
I don’t remember what my last no,no,no,no,no thread was about. It must have bothered me then, and this one bothers me now: running regressions of salary to wins.
I just don’t know what to say at this point any more. I’ve got at least a dozen threads on this. Correlation increases as the number of games increases. It’s really that simple. There’s a huge difference between running a regression against 70 games and against 700 games. Every time I see one of these regressions, the implicit treatment of the OBSERVED winning percentage is that it’s a TRUE winning percentage.
Even if God were to tell you the exact talent level of every single player in MLB, you will never be able to get r=0.9999 between talent and winning %. Not unless you’ve got one million games played.
Please, guys, stop it with regression analysis. Apologies to Hawkonomics for using his/her post as my target practice. I otherwise enjoy that blog.