Batting Average on Balls In Play. "BABIP" at some point, most of you have probably seen people refer to a pitcher's BABIP when assessing whether his success is due to a repeatable skill, a bit of luck or a combination of both. But how well do you understand BABIP? How is it calculated? What is the league average? How is BABIP affected by a pitcher's ability to induce ground balls, a defense's efficiency in converting batted balls into outs, and a ballpark's dimensions?
In this article, I'll provide an overview, attempt to answer the above questions, and take a look at the Twins pitchers in 2010, see who's been lucky and who may be due for a rebound. As usual, I'll probably beg more question than I'll answer, but more details after the jump.
What is BABIP?In a nutshell, BABIP measures a batter or pitcher's batting average by only considering batted balls that are put in play. This attempts to remove components such as home runs or strikeouts that are outside the defense's control, and it tends to normalize for pitchers right around the major league average.The exact formula is (H - HR) / (AB - K - HR + SF).
The basic concept is that pitchers generally tend toward the league average (around .300), so BABIP can give us an indication of whether a pitcher's good or bad luck will continue.
The generally accepted league average BABIP is around .300. In fact, based on my analysis of all batted balls during the 2009 season, I calculated a league average BABIP of .294. This is different from other calculations of league average BABIP because I excluded bunts from consideration (so many bunts are sacrifices), I did not consider sacrifice flies as non-at bats (outfield defense still has to catch the ball, and the batter usually isn't giving himself up), and I broke out pop flies from other fly balls.
From a first order approximation, if a pitcher's BABIP is significantly above or below .294, we may expect a regression to the mean. But when have you known me to stick to the first order approximations? Assuming a pitcher's BABIP will regress to the mean is not entirely accurate. If a pitcher gives up a high (Aaron Harang had a 23.7 LD%, highest in 2009, .339 BABIP) or low (John Danks 14.9 LD%, lowest in 2009, .274 BABIP) percentage of line drives, we would expect his BABIP to be above or below average. But let's not get ahead of ourselves quite yet.
BABIP and Type of Batted Ball
Logic would dictate that certain types of batted balls (e.g., line drives) are going to be much more likely than others (e.g., pop flies) to become base hits. How much more likely? Again, looking at all batted balls during the 2009 season, here are the numbers:
| Type | BABIP |
| Line Drive | .720 |
| Ground Ball | .231 |
| Fly Ball | .171 |
| Bunt | .179 |
| Pop Fly | .019 |
As the table above shows, line drives are far more likely to fall for a base hit than any other type of batted ball. Duh. What's interesting here is that there is a 60 point gap in BABIP between fly balls and ground balls. All things being equal, a sinker ball pitcher like Nick Blackburn would expect to have a higher BABIP than a fly ball pitcher like Scott Baker. The evidence suggests this to be the case, as Blackburn's career BABIP (.309) is higher than Baker's (.306) despite a sizable 17 point gap in LD% (19.2% for Blackburn vs 20.7% for Baker). Obviously, a sinker ball pitcher's ability to avoid home runs or long doubles relative to a fly ball pitcher has a great deal of value in itself, but solely looking at BABIP, ground ball pitchers tend to have higher BABIPs than fly ball pitchers.
Even though I did not include bunts in my calculation of BABIP, I included in the table above to give an idea of how often defenses are able to convert bunted balls into outs. You can see that defenses are somewhat more efficient than for ground balls, but since I didn't break out sacrifice attempts versus bunting for a base hit, take it with a grain of salt for now.
"Expected" BABIP
You may see a simplified version of expected BABIP, typically denoted "xBABIP", which only takes line drives into consideration. xBABIP = LD% + 0.120. This can be a useful metric, but I don't find it terribly useful since line drive rate can fluctuate quite a bit based on sample size. If a pitcher's xBABIP is above or below BABIP, it can be difficult to determine whether the line drive rate will itself regress, taking xBABIP down with it, or LD% will remain the same and BABIP will regress to xBABIP.
Using the above BABIP by batted ball type, we can calculate a pitcher's "expected" BABIP (eBABIP), against which we can compare actual BABIP to determine where we might expect a pitcher's BABIP to regress over a larger sample size. Very simply, the formula is eBABIP = LD% * 0.720 + GB% * 0.231 + FB% * .171 + PF% * 0.019. Here's how eBABIP stacks up versus BABIP for the Twins starting rotation in 2009:
| Player | eBABIP | BABIP |
| Nick Blackburn | .280 | .308 |
| Carl Pavano | .288 | .344 |
| Scott Baker | .275 | .287 |
| Francisco Liriano | .285 | .324 |
| Kevin Slowey | .279 | .352 |
In 2009, all five starting pitchers had a higher BABIP than eBABIP. In the case of Pavano, Liriano and Slowey, BABIP was much higher than expected. Based on our previous discussion, we might say the entire rotation experienced some bad luck last year. Let's compare to 2010:
| Player | eBABIP | BABIP |
| Nick Blackburn | .283 | .285 |
| Carl Pavano | .306 | .302 |
| Scott Baker | .296 | .317 |
| Francisco Liriano | .285 | .312 |
| Kevin Slowey | .325 | .337 |
Strange. Over a smaller sample size, less than a quarter of the season, we see smaller gaps between BABIP and eBABIP. What is happening here? In order to answer this, we have to look at...
Defensive Effect on BABIP
The pitcher is only one input into BABIP. He has control (some, the batter obviously has control as well) over whether a batted ball is a line drive, ground ball, fly ball, etc. But once the ball is put into play it's up to the defenders to convert into an out. As a team, the Twins were a poor defensive ballclub in 2009. The Twins -23.7 UZR last year was 24th in the majors, 10th in the American League. One would expect higher BABIPs as a result. And the Twins pitching staff's .308 team BABIP is in line with UZR, 22nd in the majors, 9th in the AL. Did the Twins team BABIP differ for the various types of batted balls?
| Type | MLB Avg | BABIP | MLB Rank |
| Line Drive | .720 | .723 | 19 |
| Ground Ball | .231 | .253 | 30 |
| Fly Ball | .171 | .191 | 24 |
| Pop Fly | .019 | .026 | 24 |
The table shows the Twins were below league average for all types of batted balls, and were worst in the league converting ground balls into outs. While the fast Metrodome artificial turf likely played a role (more on this later), it doesn't tell the whole story as Toronto ranked #2 in MLB with a .209 BABIP for ground ballsand Tampa Bay ranked #5 at .217.
Let's put this into perspective. What does the difference between a .253 and a .209 BABIP mean over an entire season of ground balls? By my calculations, the Twins fielded a total of 1,938 ground balls last year, giving up 491 base hits on those ground balls. At a .209 BABIP, the Twins would have given up a total of 405 base hits. That's 86 fewer ground balls that get through the infield, more than one every other game. Which explains why we moved to shore up our infield defense by acquiring J.J. Hardy and Orlando Hudson. Unfortunately, I haven't been able to run the 2010 numbers to determine BABIP on ground balls relative to the rest of the league, but considering our team UZR sits at +9.1, I suspect we'd see a marked improvement.
Ballpark Effect on BABIP
The last question I want to consider is the effect a ballpark can have on a team's BABIP. Larger, more cavernous outfields may result in more fly ball outs, but it could also result in more fly balls falling just outside the outfielder's reach. Artificial turf may result in more ground ball outs. A large foul territory may result in a higher percentage of pop fly outs. During my analysis, I looked at BABIP by batted ball type for every major league ballpark. Unfortunately, my ballpark totals include the home team's defense, which will be weighted heavily toward that team's defensive ability. In the case of the Metrodome, the overall BABIP on ground balls was .243, which means opposing defenses had a BABIP around .233, in line with league average. Considering that the Twins had a relatively fast base running team last year (another aspect of BABIP I have yet to analyze), this tells me that the Metrodome was roughly neutral against ground balls in 2009.
Fly balls are another story. Ballpark by ballpark, we see a huge range of BABIP, from Fenway Park's .257 (even higher when one considers Boston's .209 BABIP on fly balls defensively) to the Rangers Ballpark in Arlington (seriously, is the "In Arlington" even needed?)'s .101 BABIP allowed (.124 Rangers team defense). As one would expect, the cavernous Safeco Field (.117, 6th in MLB) and Petco Park (.141, 9th in MLB) are well above average. The Metrodome had a .233 BABIP on fly balls, which when normalized compared to the Twins .191 BABIP means the Dome allowed a far above average percentage of fly ball base hits. Considering the baggy in right field, this is also no surprise.
I expected to see the effect of Oakland's McAfee Coliseum's spacious foul territory on pop fly outs. While McAfee had the 3rd most pop flies (366 total), the Metrodome actually had the most pop flies (404 total) and pop fly base hits (12), again, not a surprise considering the ability for teams (especially the White Sox) to lose balls in the roof.
Conclusions
In summary, BABIP can be useful metric, but it's most useful when one breaks down all of the components, including batted ball type, defense and ballpark. In 2009, Twins pitchers were affected quite a bit by the porous defense and the Metrodome's tendency to generate base hits. In 2010, given what appears to be a more neutral Target Field and a much improved defensive ballclub, we can expect the team BABIP to fall quite a bit.