THE BOOK--Playing The Percentages In Baseball

Tom,

My questions relate to a similar game - fastpitch softball. A very good friend of mine is a highly successful college softball coach who I talk with a lot about the value of stolen bases and bunts. Regarding bunting, I believe the bunt is much less valuable in fastpitch softball than it is in baseball.

Two reasons:
1) double plays are almost nonexistent (especially on her team where almost every one of her players have average to plus speed)
2) runners don’t seem to score quite as often from third base with less than two outs (strikeouts in college softball are a more common occurrence than in baseball - I would estimate about a 3:2 relationship between the two sports).

1st) Since there isn’t a large database available for college softball (nor a homogeneous sample), what would eliminating the groundball double play do to the ERT in baseball?

2nd) I understand the ERT and the conclusions that bunting is usually only a good play close & late. For example, the only time to bunt a lone runner from 2nd to 3rd with no one out is in the last inning, down by one. However, does this take into account the likelihood that the opposing pitcher will just try to strike out one of the next 3 hitters (walking the other 2 if necessary)? Or does the initial hypothesis take into account data for all innings - because in softball, when a runner gets on third with less than two outs, it seems as though the best pitchers go purely into strikeout or walk mode. Is pitching strategy involved in the thought that bunting with a runner on second and no one out might be a good play? (Because I’m not convinced it is a good play AT ALL)

Thanks!

I’m currently working on releasing my simple-Markov program, which does not allow for any outs on base (like a GIDP).  So, I can tell you that the effect is that your 1xx situation will be increased by about .03 runs or so more than it should.  That is, the value of being on 2B is not as much as you are used to, when compared to 1B.  This makes perfect sense, since part of the value of being on 2B is that you won’t be doubled up.

As for softball, or any league, I would really need to know the runs per inning figure (runs per game divided by innings per game).

As for changing pitching/hitting approach with men on 3b and less than 2 outs, of course this happens.  But, it’s not like it changes *that* much.  After all, the pitcher wants to K the batter, but the batter knows that, so he’s going to counter attack.

First, I get more and more pissed off when a mainstream researcher like Leavitt on a mainstream site like TBA is not familiar with and does not specifically site the seminal work (IMO of course) on sac bunting, which is the 40-page chapter in The Book.  I’ll gladly send him a free copy if he does not want to shell out the 15 or 20 bucks to buy one.

That being said, as is the case with most articles and research on the sac bunt over the years (Tippett excpeted), Leavitt completely missed the boat.  He does mention the boat, as is also the case with many bunt research papers I have read, but without the boat, the rest of the analysis is virtually meaningless.

And by the way, any use of the NL 9th place batter RE tables to analyze the difference between hitting away and bunting (by pitchers, presumably) is meaningless.  The numbers in the NL 9th hitter RE table (e.g., runner on 1st with no outs or even 1 out) reflect what happens mainly when the pitcher bunts.  You cannot use it as a proxy for hitting away.  In order to compare the #1 hitter with one out (or two, since pitchers usually bunt with 1 out as well) and a runner on second with what would happen if the pitcher did not bunt, you have to generate your own RE table (using a Markov model) for a pitcher hitting away.  You certainly cannot use the “runner on 1st no out” number, since, as I said, that already reflects a bunt 80% of the time (and the other 20% is by the best-hitting pitchers).  It is OK to include the occasional bunt for the other spots in the batting order.  It is not OK to do that for the pitcher’s slot.

I’m not really bothered by Levitt’s lack of citations, be it mgl’s, Tippett’s, Click’s, or anyone else.  After all, who has time to look around, and what if all you are interested in is displaying your work.  That’s what peer review is supposed to catch, right?

The bothersome thing is when a researcher makes it seem like he’s looked into other work, with statements like “Baseball analysts have been near universal...”, “Recently several baseball researchers have further dug into the advisability ....”, etc, etc.  The article itself reads as if the background research has already been performed, when it hasn’t.  And Levitt is hardly alone in this kind of presentation.  Who knows, I may have been guilty of it too.

That said, it’s a good article on its own merits, if you can strip away the comments I noted.

It’s not so much a lack of citation, but read the damn chapter in The Book and you’ll see that unless you know and analyze exactly what happens when the batter attempts a bunt (singles, ROE’s, DP’s, force outs, etc.), any other analysis is fruitless and will only yield erroneous conclusions.

If you read the chapter, you would also realize that one of the keys to the successful use of sacrifice attempt is the proper use of game theory (because the result - RE or WE - of the sac bunt attempt and of hitting away is directly and significantly related to where the defense is playing - i.e., whether they are expecting the bunt or not).

The article itself was well-done, but you are right that you need to write a 50-page treatise if you are going to get anywhere.  So, it’s a great effort, but still needs a long way to go.

As well, seeing that just a tiny fraction of people who visit this site actually bought the book, I’m not sure that Dan is supposed to know that your work contains the best piece on the subject.

Like I said, he’d be better off saying what he found, rather than all the other stuff.