The Metric System: A Look At Pitchers' Control Of BABIP

NOTE: Today's A's-Angels game has been canceled due to rain. Boo. So instead of a game thread, today I bring you...The chance to argue about something!!!

NOTE #2: Tonight at midnight is the last chance to enter the AN Fantasy Baseball League. Enter now for your chance to play against other ANers!

First of all, I'd like to say that I think I'm often incorrectly dubbed an "anti stats" jerk terrorist fan, partly because I do value the art of "eyeball scouting," partly because I do believe in the importance of the mental side of baseball, and partly because I don't take all data at face value.

In fact, though, I firmly believe that stats are useful, that many metrics are far more reliable than most eyeballs, intuitions, and memories -- and so I prefer to think of myself as someone who balances an appreciation for good metrics with an appreciation for good observation, instinct, and logic.

So why do I find myself often at odds with "proven statistical claims" like "Pitchers do not control BABIP (batting average on balls put in play)"? I think it's because one of the factors that most reliably makes stats, equations, and theories more difficult to analyze or solve is the presence of many variables. For example, x+3=7 is a much different beast than, say, x+y=9. Whether Saturday or Sunday is a more popular day for AN readership is harder to measure if one Sunday is the Super Bowl while the next Sunday coincides with the trade of Dana Eveland and Bobby Crosby for Albert Pujols.

Today I want to talk about the notion that pitchers do not control BABIP, which often leads to the notion that once the ball hits the bat, from the pitcher's point of view the rest is in the hands of lady luck. I say yes and no.

The main thrust of this post is not going to be that pitchers do control their BABIP. What I'm putting forth today is just the idea that it's more complicated than just "they do" or "they don't." We know that over time, a pitcher's BABIP, whether he is Greg Maddux, Greg Cadaret, or Kevin Gregg, will generally be between .280-.320. Why?

The biggest factor is, quite simply, the number of fielders you are allowed to put on the infield and in the outfield, and the size of a baseball field. If you were allowed only 3 infielders and 2 outfielders, BABIPs would soar, and if you could employ a 5th infielder and a 4th outfielder, BABIPs would plummet. BABIPs would be much higher at Diamondbacks games if Arizona played its home games in the Grand Canyon. The main reason a batted ball can be expected to fall safely about 3/10 of the time, and not 2/10 or 4/10 of the time, is a direct function of how many fielders are allowed in how big a space, and those relative constants drive BABIP more than any other factors.

However, here are some other observations that I want to introduce into the equation:

* The spread of .280-.320 for "expected BABIP" is relatively small. It is .300 +/- only .020 points of batting average. However, it is not miniscule in that we do not generally regard a career .260 hitter and a .300 hitter to be the same, nor do we regard a career .220 hitter to be interchangeable with a career .260 hitter. So even when we start with the premise that a pitcher's "expected BABIP" ranges only from .280-.320, remember that the ability to hold opposing hitters to a BABIP of .280 and one of .320 are different enough that even the range of expected outcomes is actually significant -- or at least not entirely insignificant.

* I wonder if some of the data on pitchers' BABIP is skewed by the fact that pitchers who have true "high BABIPs" don't last long. What are the keys to a high BABIP? A straight fastball, the tendency to hang sliders, bad location -- also the qualities associated with pitchers who are quickly demoted and swiftly forgotten.

* Greg Maddux has been cited as evidence of how pitchers don't control BABIP, because here's a clear Hall of Famer whose BABIP was .289, just .011 points off the average and well within the expected range. I don't think Maddux is a great example, though. Maddux's game was to throw a lot of first pitch strikes, and thus to exchange some "hittable strikes" for the ability to maintain low BB totals and pitch consistently ahead and the count. I think that if Maddux's one goal in life was to have an exceptionally low BABIP, he could have had one -- he just knew he was a better pitcher overall if he approached hitters exactly the way he did.

So what's luck and what's skill? Intuitively -- and fangraphs agrees -- the biggest way a pitcher does exert influence over his BABIP is by allowing a higher or lower percentage of balls to be hit really hard. BABIP highly correlates, not surprisingly, with "line drive rate," and so while the bloops, bouncers, and looping fly balls reliably even out over time, what doesn't even out is when one pitcher keeps serving up a buffet of searing line drives while another pitcher allows the normal assortment of liners, bleeders, pop ups, and sharp grounders. GO/AO ratio is also a factor, since ground balls and fly balls don't have identical "expected BABIPs."

So why, then, is the range of "expected BABIP" still so small? I think a fair amount of it is "regression to the mean," which does not have anything to do with luck. A pitcher whose BABIP, at the start of the season, stands at .444 -- because he is pitching really badly and balls are being scorched off of him left and right -- will see that number fall quickly towards .300 when he makes adjustments, because a stretch of even 33/100 will send the number not falling but rather plummeting.

A pitcher whose BABIP stands at .182, meanwhile -- maybe because he is pitching exceptionally well, jamming hitters, getting late action on his sinker, and inducing a lot of funky swings and weak contact -- will see that number rise like Lazarus on crack even should just 27 of the next 100 balls fall safely into play. The mechanism that is pushing the number towards the mean is not just a force of "luck returning to normal" (as it would be with the flipped coin), but also just a force of statistics -- the force which sends stats always spinning towards the vortex of their statistical mean.

In summary, I believe that pitchers probably control their BABIP more over short stretches than over longer sustained stretches, that regression and the quick banishment of outliers (on the bad side) help to compress the range of expected BABIPs, and that the pitcher's ability to influence BABIP is far more complex than just "basically they don't" -- and that otherwise, yes: When you put exactly 4 infielders and 3 outfielders on a standard baseball field, and then consider the customary range of batted balls, a pitcher is going to watch balls in play drop safely about 30% of the time no matter who they are.