THE BOOK--Playing The Percentages In Baseball

Would be interested to know more of the methodology.  Studies like these are particularly problematic because of the data.  In case anyone does not know, retrosheet provides just two things about the batted balls:  One, the type, according to them, or from whomever they got the data, G, F, L, or P.  I am assuming that Rally only looked at “P”.  Two, the section on the field.  The field is somewhat arbitrarily broken down into 60 some odd sections.  There isn’t really a “SS” section and “3B” section, etc., as you might be led to believe from the article (Rally, you really SHOULD explain more of what you mean - although I realize that this is just a blog entry and not a full-on article).

As to what sections Rally is including, I have no idea.

I just don’t think that given the problems associated with the data, you can get any meaninglful results on pop flies.  For example, parks are very important with this kind of stuff.  With large OF’s, the OF’ers are going to play deeper and catch fewer pop flies.  THAT is why (or at least principally why) Tulo catches more pop flies!  In Fenway, or Houston, the left fielder is practically on top of the infield and will catch many more pop flies than in other parks.  Etc.

Rally is looking at all popups caught by the 2B, 3B, SS, LF, and CF.

I look at every popup, as coded by retrosheet, that is fielded by the shortstop or an adjacent fielder (3B, 2B, LF, CF).

It is NOT zone-based (which is arbitrary).  So, he reduced two arbitrary recordings (zone, pop/liner/fly) into one (pop/liner/fly).

Whether that is better than the arbitrary zone, I don’t know.

He’s attempting to keep things as “factual” as possible.

Seems like a With or Without...Your Shortstop kind of article. The data is very noisy, since some guys are just pop-up hogs, which would help them while hurting their ‘mates at the same time.

Retrosheet doesn’t have zone for a lot of years.  I think its primarily there for the 93-98 data.

I just wanted to see if there’s any evidence for thinking that Jeter adds a lot of value for his ability to chase popups.  Seems like as many of them bloop in against the Yankees as against anyone else.

What do you include in your bloop hits (popups not caught)?  Any popup (coded as “P") that is not caught?

For 2005 to 2007 data only uses the P batted ball designation on outs when the ball is caught by an infielder.  When a popup is hit less than 110’ less than .4% of the balls are hits (43 of 11994) and most of those are flukes; bunts and wind blown balls.  What we really want to know are how many popups are hit to the left side of the infield farther than 109’.  The answer is 10342 of which 341 were hits.  I used 5 degrees to the second base side as the cutoff instead of directly over 2B to get balls that the SS might wander over and catch.  The SS catches 6952 of the 10001 outs, 3B - 2457 and 2B - 582.  So about 69.5%, 24.5% and 6% each. 

For the Yankees 2005-2007 there were 320 popups in this area.  Using the percentages above they should have converted 310 to outs with the shortstop catching 224 of them.  They actually converted 306 to outs with the SS catching 220.  So about -4 over 3 years for the left side of the infield as a whole and Jeter in particular.  This is using the “not very precise but probably good enough for this analysis” hit locations from mlb Gameday.  What we don’t know that would affect the analysis is how consistant Retrosheet is in recording a popup hit as opposed to a short fly ball hit for balls hit the same distance and hang time, and short fly ball outs caught by an outfielder that are within range of an infielder.

Yes, my “bloop hits” were coded by retrosheet as bbtype = P and hit value > 0.

Peter’s dataset is better than what I was looking at since he can determine ball location, at least to the extent that the data is reliable.