THE BOOK--Playing The Percentages In Baseball

Friday, January 28, 2011

How to calculate BABIP

Before we talk about how to calculate BABIP, first ask yourself WHY you want to calculate BABIP.  And don’t get hung up on the literal name ("batting average").

I’ll be presumptuous and tell you why you might want to know: you want to know, of the times that the fielders are involved, how often does the batter successfully reach first base.

There are two ways for the batter to confront the pitcher: bunt or swing away. 

The success rate of reaching base on bunts is much higher than swinging away.  It would have to be, because on bunts you can only gain one base, while swinging away let’s you get more bases.

Now, if a pitcher is going to be disproportionately bunted upon, or if a hitter will bunt a disproportionate number of times, then the BABIP can increase substantially even though the run impact may have been the same either way.  Do we necessarily want to count a bunt attempt the same as a swing away?

In addition, you can have a sac bunt (attempted bunt where you make an out) recorded under SH, but a bunt attempt where you get a hit recorded under H and AB.  From the batter’s persective, it should be counted together, but the stats don’t make the distinction that clear.

Let’s say you have a hitter who has the following stats:
500 non-bunt Balls in Play (BIP)
150 non-bunt hits
350 non-bunt outs

60 bunt Balls in Play (BIP)
10 sacrifice bunts
20 bunt hits
30 bunt outs

Now, you have three ways to calculate his BABIP:
1. Look only at his non-bunts.  So, 150/500 = .300

2. Look at ALL his BIP.  So 150+20 / 500+60 = .304

3. Lookat ALL his BIP, but remove his sac bunts.  So 150+20 / 500+50 = .309

From my perspective, I would prefer to look at only the swing-away attempts because I don’t want Juan Pierre’s BIP to be disproportionately influenced from Adam Dunn’s because of his bunt attempts.  To me therefore, I want that .300.  Now, if you instead decide to take all his bunt attempts but remove those counted as SH, but keep the “accidental” hits on sac attempts, you are going to get a skew.  You are not removing the BIP in a systematic manner, but a biased manner.  Therefore, I don’t like it if BABIP are counted with the SH removed, but you keep the hits on sac attempts.

So, look at all attempts (#1 above), or look at all non-bunt attempts (#2 above).

***

Finally: reaching base on error.  As you know, I think it’s ridiculous to count reaching base on error in the same way as you would count making an out, and not the same way as getting a single.  BABIP must include reaching on error.

***

Alternatively: you can have a wOBA-based BABIP.  Count a swing-away single not as “1”, but “0.9”.  Count a double or triple not as “1”, but “1.3”.  If you do this, then maybe you can come up with a better number for a bunt-single, so that there’s no bias.  Perhaps counting a bunt-single as 0.8 does the job.  But, now we are going away from a pure counting metric, to a more analytic metric.

Then again, what would you do with BABIP other than in an analytic sense?

***

Isn’t SH (sacrifice hit) the worse named stat around?  Hit?  It’s an intentional bunt out.  Call it sacrifice bunt or sacrifice out, but not hit.  A sacrifice hit would be an accidential bunt hit.