A note from a reader on BABIP

Reader Abbot Katz checks in with this note on batting average on balls in play.

It has been some time now since the BABIP established its place in the sabermetric canon, and understandably so. BABIP means to aim a quantified scrutiny at a rather interesting problem: the extent to which a batter’s skill at directing 90 mile-per-hour pitches away from the best preventative efforts of sagely-positioned, gloved men can be assayed.

Yet, I have long experienced a measure of unease with the BABIP, and while my hermetic disquiet won’t suffice to give pause to a sabermetrician—and it shouldn’t—perhaps this simple example will:

Imagine two hitters, A and B, both of whom assemble 600 at-bats, 180 hits, and 100 strikeouts. A hits 15 home runs, however, while B musters 40. Both players hit .300, of course, but their respective BABIPs look like this:

A — .340
B — .304

And therein lays the conundrum. In what manner shape or form are we entitled to conclude, on the basis of the evidence placed before us, that it is A who more adeptly interposes batted balls between the defenders assigned to thwart him? How can we possibly formulate such a judgment when the data affords us no warrant to do so?

All we know is that B outhomers A, and all we have is a metric

(Hits-HR)/(AB-HR-SO+SF)

which falsely skews its conclusion. And that is because by informing both tiers of the fraction, the subtracted home runs pare the numerator artificially for power hitters, culminating in lower BABIPs.

Again – why should we be entitled to thus declare that high-home run achievers commit inferior skills to the challenge of turning balls in play into hits? That inference is simply not available to us.

In fact, BABIPs typically exceed players’ averages ; but for big power hitters, the relationship undergoes a curious inversion. Babe Ruth’s .342 lifetime average turns into a BABIP of .340. For Hank Aaron, the split stands at .305/.295; yet Rod Carew checks in at .328/.361. Can we thus assert, with the appropriate, straight-faced measure of confidence, that Carew’s ball-in-play facility truly overwhelms Aaron’s by 66 points?

The inarguable mathematical point is this: that, all other things being equal, the player with more home runs suffers a relative decrement in BABIP. All else is speculation. If we proceed from the eminently clear-eyed premise that home runs tend to be hit harder than the average ball in play, we can go on to propose that, were these fence-clearers to fall short by a few feet, they would nevertheless fall for hits in greater profusion than typical batted balls – thus resulting in a higher projected BABIP for power hitters.

That too qualifies as conjecture, albeit a sensible one. Still, the fact is that the BABIP puts the caliper to a most intriguing property of the batter’s skill set in a manner that doesn’t quite measure up.

And thanks for sharing my unease.

1 These figures omit sacrifice flies, which contribute a very small effect.


11 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
DSMok1
14 years ago

When I calculate a predicted BABIP, I always adjust the GB/LD/FB/POP percentages after subtracting the predicted number of HRs from the FB total…

The problem with the writer’s conclusion is this: thinking that relative BABIP numbers between players means anything!  He is asserting that Carew’s BABIP being above Aaron’s is an inaccurate measure somehow.  That is wrong.  BABIP is a pure measure, all it says is what a player did.  You cannot compare one player with another… BABIP DOES NOT regress to the mean, it regresses towards the player’s true talent level.

Suppose two players hit 25%LD, 40%GB, and 35�.  One player hits 4% of LD and 25% of FB for HR, the other just 10% of FB.  That would yield a BIP split of 26.5%LD, 44.3%GB, and 29� for the first batter, and 25.9%LD, 41.4%GB, and 32.6� for the second.  I would expect the BABIP of the first to be higher than the BABIP of the second, unless the second is faster or the first is a dominant pull hitter.  So what?

So yes, “Carew’s ball-in-play facility truly overwhelms Aaron’s by 66 points”.  There is no problem with that; that is where part of Carew’s value lay.  The problem is thinking that BABIP is something it is not.

If you are trying to predict BABIP, you must consider the impact of HR… but there is so much more to consider than that.  If you are trying to compare players by looking at their BABIP… why?

Total
14 years ago

I echo Eric/OR; has there been some movement to make BABIP the be-all and end-all of measures?

And what, exactly, is “hermetic disquiet”?

Colin Wyers
14 years ago

Why are we looking at BABIP for hitters like this in the first place?

Sky Kalkman
14 years ago

I guess the point is that if you want to adjust for a player aging and losing power (although don’t hitters tend to peak late for power?) that lost power will help improve or at least maintain BABIP.  A loss of HR/FB skill is not independent of BABIP “skill”.

Detroit Michael
14 years ago

Without running the data, I would guess that the average batter with 40 HR per 600 AB strikes out considerably more than the average batter with 15 HR per 600 AB does.  Thus a hypothetical situation where he holds the number of AB, H, SO, and SF constant and changes only the HR creates an atypical comparison.

A player with more home runs has harder hit batted balls on average.  A player with harder hit batter balls will have a higher BABIP, but this effect might be offset by a GB/LD/FB distribution that leads to lower expected BABIP.  To conclude any more, I think one would have to look at data, not reason from hypotheticals.

Abbott Katz
14 years ago

To my public, as it were

Stepping out of my hermetic isolation, I’ll offer just a few replies to some of the remarks issued to my little piece:

-I never claimed the BABIP stands as the epitome of offensive metrics. It is, nevertheless, a widely invoked measure.

-In response to Detroit Michael’s observation that a player with harder hit balls will have a higher BABIP: true, more data need to be inspected, but consider Barry Bonds’ 2001, his annus mirabilis in which he hit, through whatever means, 73 HRs. Bonds’ BABIP that year came to .269—a year in which he actually hit .328. .269 is what happens when you average a HR every 6.5 ABs, and readers who write those numbers off as an outlier miss the point—and it doesn’t matter what Bonds was having for breakfast those days, either. It isn’t the steroids – the problem is the stat.

Thus to assert, as DSMok1 does, that BABIPs regress toward a player’s talent level, is in my humble view, arguable. The bottom line: more HRs, lower BABIP – and it is, in my judgment, proper to advance the all-things-being-equal proposition here.

Eric/OR
14 years ago

Sir: I think your observation is a valid one, but I think it and $1.50 will buy you a cup of coffee – I just don’t hear anyone that seriously follows sabermetrics making the contention that BABIP is an important skill indicator on a par with, say, wOBA or EqAv.  Are there metric-mongers out there disparaging Babe Ruth in favor of Rod Carew?

Dennis
14 years ago

This seems like a valid point. In addition to subtracting home runs, could we not just add back some percentage of home runs that we would expect to become doubles or any other sort of base hit?

Eric/OR
14 years ago

A needlessly painful read that could easily be boiled down to a few sentences.  Also, I’m not sure where the concern is being directed – is there some movement I’m unaware of attempting to badmouth the Babe Ruths of the world on the basis of BABIP?

Dennis
14 years ago

It’s almost as if it could be an xBABIP, much like xFIP—adjust a player’s BABIP by replacing HR with (FB * league average rate of fly balls becoming home runs). This seems like it achieves the desired effect of only removing the fly balls that aren’t fielded from BABIP calculations.

Eric/OR
14 years ago

In any event, regardless of its utility, the insight is valid.