from Amazon
Friday, August 25, 2006
Pitcher’s Won-Loss Record
Patriot has the beginnings of an interesting look at a pitcher’s Won Loss record.
Here’s how I do it:
The problem, as Patriot points out, when you compare a pitcher’s W/L record to the rest of his teammates, you are essentially setting the bar at “.500”. Wins Above Team, where wins=0, treats the teammates as a .500 pitcher.
Instead, I do this. I’ll take the 1986 Mets, who went 108-54 (.667), and Dwight Gooden, who went 17-6 (.739). Take that record, and get rid of Gooden. That makes his teammates (pitchers AND hitters AND fielders) as 91-48 (.654). Assuming that off/def is split 50/50, then we’ll say that the .654 is broken up as .577 for the offense and .577 for the defense (sans Gooden).
As we can see, that it’s that .577 offense that bumped the .577 defense to an overall .654. Therefore, reverse Gooden’s .739 down by 77 point to .662. That’s Gooden (and fielders) record, with a .500 offense.
A couple of things. I’m treating the offense as if it was equal to the defense. We don’t know that it was.
I’m also ignoring fielding, or treating the team defense as zero. Not necessarily a good thing, but which can be similarly adjusted. That is, if the pitching + defense is .577, then maybe the defense is .523, and the pitching is .554. Knocking Gooden down another 23 points, his .662 becomes .639.
Of course, Gooden needed his fielders less, so the impact of his fielders wouldn’t have been as much.
As well, I would use the Odds Ratio method, not the differential method.
Finally, I know all about the problems, including the way wins and losses are assigned to begin with and the super tiny sample size. It’s 2006. We don’t need to discuss these last two issues, do we?
Odds Ratio would work out this way:
sqrt(91/48)=1.38 ratio, or .579 percentage (in place of .577)
17/6 / 1.38 = 2.05 ratio, or .672 percentage (in place of .662)