BAbip and the Count

BAbip, or "batting average on balls in play," is an integral statistic to current analysis of baseball. It's calculated by dividing the number of hits other than home runs (probably excluding inside-the-park home runs in theory, although I'm not sure in practice) by the number of at-bats other than home runs or strikeouts. The idea is that these at-bats are the ones that result in a batted ball for the defense to handle, and the times when a hit results are more about some combination of luck and the defense than the hitter or the pitcher. Now, the logical extreme toward which that idea is tending is that hitters and pitchers should receive basically no credit or blame for abnormal BAbip performances, and that all that matters in evaluating them is home runs, strikeouts, and walks. I don't think very many people believe that, especially as regards hitters, who are known to have characteristically high or characteristically low BAbip tendencies. Pitchers, who have to face the entire league of hitters with above- and -below average BAbip tendencies, are assumed to have less ability to have genuinely non-average tendencies in this regard, although not no ability; in particular, knuckleballers are known to have systemically low BAbips. But the fact still remains that a lot of people think that pitchers have fairly little control over whether a ball that's put in play results in a hit or an out.

Enter this article by Fangraphs, which analyzes On-Base Percentage and home runs per batted ball for each count. (Note: OBP is for any plate appearance which at any point reaches that count, even if the at-bat doesn't end on that count, while HR/batted ball is just when the ball's put in play on that count.) There are strong and very-nearly-matching tendencies for some counts to favor hitters on both measures and for other counts to favor pitchers on both measures. Overall we get the following order of counts from most pitcher-friendly to most hitter-friendly (more or less): 0-2, 1-2, 0-1, 2-2, 1-1, 0-0, 2-1, 3-2, 1-0, 2-0, 3-1, 3-0. Now, if we find that home runs per batted ball varies tremendously with the count, that seems to suggest that balls are hit systemically harder on some counts than on others. Obviously the count is substantially subject to the pitcher's control. So, are there any interesting splits in batting average on balls in play by the count on which the ball got put in play?

Looking at data for the 2011 National League as a whole, the answer seems to be "yes." Here are the BAbip numbers for each of those 12 possible counts, in that order from most- to least-pitcher-friendly: .272, .275, .297, .293, .296, .300, .316, .303, .296, .325, .304, .305. There's a pretty clear trend here, with higher BAbips on more hitter-friendly counts, though it's not monotonally increasing or anything. What if we do something I've often been intrigued by, and consider that, while the defense isn't involved on home runs, it's not clear that the details of whether the ball flies just over the wall or just short of it and into a waiting glove is much more under the pitcher's control than whether the ball happens to land where no one can touch it? That is, suppose we create the statistic "batting average on batted balls," or hits divided by at-bats minus strikeouts, and list that for each different count? Then we get .290, .294, .318, .316, .321, .329, .346, .330, .328, .369, .346, .366. That's much closer to being strictly increasing.

Now, with graphs!

The orange dots are batting average on batted balls, the blue dots are batting average on balls in play.

The count seems to determine about 55.3% of the variance in batting average on balls in play, and about 82.4% of the variance in batting average on batted balls. Both metrics seem to suggest that 2-0 is a more friendly count for the hitter than 3-1, unlike home runs per batted ball or on-base percentage after reaching. Likewise, these metrics like 2-1 better than 3-2 or 1-0, in contrast to the ones discussed by Fangraphs. But overall we can see pretty clearly that getting into favorable counts should permit a pitcher to have a substantial degree of control over the batting average against them on balls in play, or on batted balls. And since everyone admits that pitchers have substantial control over the count, that should mean pitchers have substantial control over BAbip. A pitcher who's routinely getting hitters into two-strike counts should experience a lower BAbip than one who's running up lots of 3-ball counts, or letting hitters put the ball in play on 1-0, 2-1, etc.

And we're what, surprised by this?