THE BOOK--Playing The Percentages In Baseball

His great performance with men on base has a lot to do with his groundball/DP rates.

Sal/1 - but Jimenez’s double play rate is about twice the league average rate, while his groundball rate is “only” 120%” higher than the league average rate, so that’s one source of luck I think.  I really doubt that pitchers have any degree of control over how many double plays they get outside of their groundball rate. 

Tango, thanks for the reply.  I was wondering though, if you could somehow adjust the FIP coefficients based on how good you think the pitcher is at limiting damage on BIP?

Nick/#2: Kincaid did a write-up on pitching to contact and how that may affect FIP.

http://www.3-dbaseball.net/2009/11/pitching-to-contact-and-fip.html

I’d imagine you can do something similar if you just estimate how much you think Jimenez’ BIP are worth.

"I was wondering though, if you could somehow adjust the FIP coefficients based on how good you think the pitcher is at limiting damage on BIP?”

The whole point of FIP is to make it simple so you DON’T do that.  If you want to be more accurate you simply take his BABIP and regress it X% toward the league average and then use THAT number in the FIP formula rather than the league average (.300) or whatever it is.

I don’t know off the top of my head what the X/(BIP+X) regression formula for BABIP, but I am sure Tango can tell us.

Remember that FIP is simply a 100% regression for BABIP when it shouldn’t be.  For small samples 100% regression is close enough.  For larger samples, it isn’t that great.  Either way (large or small sample), you are always better off taking the next step which is to do what I just said - use the same formula, but instead of league average BABIP, you use the pitcher’s own BABIP regressed toward league average the appropriate amount.  Of course, a decent part of that pitcher’s BABIP is his defense, so you would also want to factor out the defense if you can.  But that is the NEXT step.  And of course, not all BABIP are created equal, etc.  Eventually, you might just as well use a component ERA with each component regressed the appropriate amount toward the league averages of that type of pitcher (hand, starter/reliever, fastball speed or “stuff” if you know it, etc.).

OK, so the league average BABIP is not actually in the FIP formula, so in order to use a pitcher’s own BABIP regressed toward that mean, you have to tweak the formula a little - maybe add another term or something like that.  I am sure Tango can also provide the correct formula if you wanted to use the pitcher’s own estimated true BABIP rather than league average.

Nick: I’m not disputing that his on-base numbers are “lucky” (or anomalous or whatever).  Some is luck, but we would expect him to “strand” a decent number of runners not because he’s stranding them but because he’s eliminating them.  Just a matter of how much you want to regress his strand rate.

Well, Jiminez is getting pretty close to the 1,500 BIP mark, so we could estimate that r=.4 or so, right?  So regress 60% and his adjusted BABIP for the season is .275.  That would be about 9 hits worth of “luck” or 7 runs worth.  Plug that into his ERA and it goes up to about 1.80.  Still not accounting for the Rockies’ defense or LOB%, of course.

I don’t know off the top of my head what the X/(BIP+X) regression formula for BABIP, but I am sure Tango can tell us.

3700

http://www.insidethebook.com/ee/index.php/site/comments/career_dips_numbers/

3700 is one number, and I think I’ve used 1800 or something another time.  It depends on the sample data used.

In any event, FIP is FIP, and I wouldn’t change it any more than I’d change OBP’s weighting of walks and HR (1 each).

***

Nick is asking that if a pitcher really does have fewer baserunners on base, shouldn’t that affect the FIP coefficients.  The HR coefficient wouldn’t change, and the others would go down a bit, but I wouldn’t worry about it.  It just doesn’t happen to an extent that the coefficients have to change.

If you really want to know, plug everything into BaseRuns and use that.

"Nick is asking that if a pitcher really does have fewer baserunners on base, shouldn’t that affect the FIP coefficients.”

Oh, OK.

Again, FIP is an approximation. If you want to be more exact, don’t use it.  As you say, if you want to be more exact than OPS, use something else like wOBA or lwts.

Thanks guys.  As I said to Tango in the email, it’s my understanding that FIP is composed of 4 values:

-the run value of the pitchers strikeouts
-the run value of the pitchers walks + hbp
-the run value of the pitchers home runs
-the run value of the pitchers balls in play

The first three are directly included in the FIP formula, while the latter is indirectly, and for each the league average value is assumed.  I’m just curious if there is a way to plug in different values for BIP and see how FIP would change. 

Kincaid’s article linked to above looks like a pretty good start, so I’ll play around with that.