View Single Post
Old 04-06-2005, 09:48 AM   #752
dougaiton
Hall Of Famer
 
dougaiton's Avatar
 
Join Date: Mar 2003
Location: Location:
Posts: 3,414
Quote:
Originally Posted by Smoothie7745
How about in STF, which, according to your site, baseballprospectus.com, is "A rough indicator of the pitcher's overall dominance"? Ryan is a 34, whereas Finley is a 20. Also, Ryan's ERA (which is, in essence, a relative guage of an average game for a pitcher, right?) is almost a run better. Finley also allowed more pitching runs above average by a fair margin (200 to 186). But I think the biggest difference is their Delta-H, if I understand this stat correctly (it's a new one to me)...it appears, in essence, that the higher a Delta-H you get above zero, the worse, more or less. That said, Ryan's Delta-H was a -95, whereas Finley's was a +51.

These are all not translated apart, from stuff, which I'll get to in a minute.

If we translate them, than Finley has an ERA+ of 115 to Ryan's 112. Advantage Finley.

PRAA is better if it is higher (I know, it's back to front!). Advantage Finley.

Delta-H is a measure of luck (if that's what you believe BABIP to be). Finley had +51 bad luck (i.e., he allowed 51 hits in his career more than he should have), whereas Ryan was -93 (he allowed 93 hits less than he should have. Advantage Finley (although this isn't a measure of effectiveness, just a measure of how lucky the could/should have been).

STF is interesting. You'll need to talk to Josh, but there's a thread on Miguel Batista where we discuss it. I like it, but others don't. Advantage Ryan!

EDIT: More on STF:

STF is not really a measure of effectiveness, but rather pitching 'talent'.

I've defended it, so it's a bit hypocritical of me to disagree with using it. It isn't about results though, although maybe that's not what we're talking about.



Quote:
Originally Posted by Joshv02
That is true. I guess it depends on what we are using the metric to do. It seems a better approach is just to take HR/9 and regress towards league average at some higher rate than BB/HBP/SO. That would seem to be more intuitive to me. But, BPro doesn't seem to like regressing stats. I don't know why. (They also don't like PBP data. That I do know why, but it seems like an odd reason.)

So, if we think that pitchers have some control over h$, we would just regress that big time towards lg/tm avg. If we thought that pitchers have control over HRr we do the same thing, but less bigtime.

When they use PERA and EqERA, which is a pseudo ERC (iirc), that throws me, too: do they have any reason to think that PERA or EqERA is any more predictable than FIPS with translated stats? Why not just use a normalize H$, normalized R/G, or some other things. Anyway, sorry for the digression.

Last edited by dougaiton; 04-06-2005 at 09:52 AM.
dougaiton is offline   Reply With Quote