Thursday, July 21, 2011

Rotisserie Values and Pitcher Metrics: Part II

Last month, I examined the theory that certain pitcher metrics might be worthwhile in determining pitcher value. I concluded that K/9 is an excellent metric to determine pitcher value in Rotisserie, BB/9 isn't quite as good, and G/F is hardly a predictive metric.

Rotoman wanted to know if I could make one little substitution.
Not that you guys aren't busy, but it seems worthwhile at this point to run the same numbers substituting HR rate for G/FB.
I did this and added it to the spreadsheet I already created for this project. Table 1 below contains the previous results using G/F and the new results using HR/9.

Table 1

$ G/F











Rotoman's educated guess was quite correct. Substituting HR/9 for G/F makes these combined metrics far more predictive.

Table 2 breaks down each metric separately.

Table 2

HR/9 does nearly as well as K/9. The pitchers in the 21-30 grouping are significant enough of an outlier that K/9 still stands as a superior metric to use to tie performance to Roto value. However, it's close, and the best HR/9 pitchers simply blow away the best K/9 pitchers in terms of matching performance to value.

This is all very interesting to look at. However, all this really does is tell me a lot about last year's performance; it is hardly instructive in terms of analyzing future performance. Thanks to xFIP, we know that last year's low HR/FB pitchers are likely to regress to the median the following year.... if not sooner.

I can't read Eno Saris' mind, but this is what I believe he was trying to get to in his original piece that sparked all of this in the first place.
We do know that there are three main components to a pitcher's performance: How often do they strike a guy out, how often do they walk a guy, and how often do they get a guy to ground out. These three aspects of pitching are the parts of the game that a pitcher has the most control over. Generally, you want a pitcher to be above average or better in two of the three categories before you place your trust in them.
While HR/9 are predictive of Rotisserie value, they aren't necessarily predictive of how good a pitcher is, since home runs aren't as much in the realm of a pitcher's control as we used to believe they were. A pitcher with a high ground ball will allow fewer fly balls and in theory should allow fewer HR.

In practice, though, G/F by itself is nearly worthless, and definitely isn't as important as K/9 or BB/9. You want to get a guy who controls his own game by striking hitters out and keeping them off base by walking as few as possible, in that order. A strikeout pitcher who also generates a lot of ground outs is better than a strikeout pitcher who doesn't, but if you ignore G/F rates entirely, you'll still do pretty well if you only use K/9 when looking for arms for your staff.

1 comment:

George said...

It's LIMA, basically. Find pitchers who strike out a lot of people and don't give up home runs. Might be worthwhile running your analysis with %K and %HR rather than HR/9.