THE BOOK--Playing The Percentages In Baseball

What do you mean by gun bias?  Employees in AL parks are more speed conscious that NL ones?  Is there actual data on this?

Tango:  Using the data from MGL’s AL-NL articles in THT, I think you can do a comparison of NL and AL parks.  Using 2000-2005 interleague data, I found the difference was tiny, with NL parks adding about 7 runs per season compared to AL parks.  See this thread:  http://www.insidethebook.com/ee/index.php/site/comments/why_did_the_al_cream_the_nl_this_year_in_interleague_play/.  So I don’t think that would change John’s analysis much.  His 60 run estimate is also very close to the estimate I came up with there of an AL pitcher advantage of about 50 runs.

Tools: not more speed conscious, but at what point they pick up the ball.  If one park picks up the ball when it leaves the pitcher’s hand, and another one picks up the ball when it crosses the plate, you’ve got a 7-8 MPH difference.  So, every 7-8 feet or so, you are talking about 1 MPH here or there.

If this was random, no one would care.  But, if you have a few AL parks that pick up the ball closer to the pitcher, and a few in NL parks that pick up the ball closer to the batter, then that’s a systematic bias.  And that’s what needs to be corrected.

And, it’s easy enough.  Figure out RJ’s fastball at Yankee Stadium and on the road.  Do the same to Mussina, et al.  Do it to the visiting pitchers.  Voila: a gun park factor, just like you’d have a HR park factor or a LH/RH park factor, or a GB/FB park factor.

***

Guy, is that 7 runs per 162x40 PA (.001 runs per PA)?

The difference was .6 runs per 500 PAs (which was how MGL did his analysis).  That is, a hitter with 500 PAs generated 0.6 more runs in an NL park vs. an AL park.  (Of course, both leagues’ hitters do better in their own league, because you’re capturing homefield advantage). 

MGL’s data was only presented to one decimal and per 500 PAs, so I could be off a little due to rounding, but at least for those 6 years it appears the parks were pretty equivalent.

I guess that was Guy posting #4.

Ok, so that’s .0012 runs per PA.  1 point of wOBA is .0009 runs per PA.  OPS is 2.2 times wOBA.  So, 1 point of OPS is .002 runs per PA (if I’m doing all this correctly).

So, agreed, less than half a point of OPS difference between the parks.  John was proved right here (though my objections stand with merit).

The other issue is fielders.  Again, are fielders equal or not in both leagues?  Dual-leagued fielders is one way to do this.  Another is to look at the Fans’ Scouting Report.

A technical note:  A lot of people use the wrong “uncertainty calculation” when estimating the uncertainty of a “difference between (or sum of) two sample values” such as NLOPS-ALOPS.

Remember that the the standard deviation of the sum of or difference between two values is the square root of the sum of the variances and NOT just the square root of one of the variances and definitely not the standard deviation of the two samples combined (in fact, it is exactly twice that I think).

I am assuming that John did it the right way, but many otherwise competent researchers forget the “rule.”

Tom—Thanks much for the kind words. 

Regarding your point about AL v. NL parks, you’re perfectly right of course.  I tried to make it clear the assumption I was making about the equivalence of NL and AL parks, an assumption I felt intuitively was reasonable.  Although, as you say, I did not present any data to support my view.  I had not seen the thread here on MGL’s articles from last summer, so I was not aware of Guy’s conclusion the the park differences in the two leagues were small. I’m glad my assumption appears to be valid, although this seems like a topic that could use a dedicated study.

As for the starter/reliever issue that is a good point. I agree with you that it’s probably minor, but should be checked.  Another issue that I thought about, but did not study, is that of older players moving preferentially to the AL where they can DH.  (I did not make any age corrections, nor did I exclude DH’s.)

MGL—yes, I calculated the uncertainty on the OPS difference correctly: var(a-b)= var(a)+var(b), where var is the square of the 1 SD uncertainty. Doesn’t everybody do it this way?

John, thanks for stopping by, and taking the critique in the spirit in which it was delivered.

Yes, the aging issue is also interesting.  Because you limited your study to a 3-year gap, that curtails alot of it.  And, maybe only 60% of players switch to the AL as they get older, as opposed to 50%?  So, the expectation should be a very low change in OPS as a result of the aging issue, in your sample.

In terms of fielding, I think you could do this by looking at DER, excluding PAs by pitchers and DHs.  Calculate these 8 DERs:
INTRA-LG
ALF vs. ALH on Rd (AL parks)
ALF vs. ALH at Hm (AL)
NLF vs. NLH on Rd (NL)
NLF vs. NLH at Hm (NL)
INTER-LG
ALF vs. NLH on Rd (NL Parks)
ALF vs. NLH at Hm (AL)
NLF vs. ALH on Rd (AL)
NLF vs. ALH at Hm (NL)

Weighting each one equally to control for Hm/Rd, hitter quality, and park effects, just calculate an average DER for ALF and for NLF.  Of couse, the one thing this doesn’t control for is pitchers.  However, it seems pretty unlikely that there could be anything more than a very small league difference in pitcher BABIP ability, once you control for park and defense (and I believe in BABIP ability as much as anyone!).  So any significant difference in league DER would probably be a real measure of different fielding ability. 

That said, my guess is you’d do a lot of work to find no difference (or a tiny difference).
an analysis similar to what