Strategy February 18, 2012

# Does BABIP Steal Our Common Sense? Part 2

By Michael Caron

If you read Part 1 of this article and you’re still with me now, I commend your dedication. If you skipped over Part 1, good luck trying to follow my seemingly endless rambling. It’s time to take yet another step down the rabbit hole. In the previous article, I displayed the extent a players speed (or lack thereof) can play in determining an xBABIP.

Starting with the formula Tristan Cockroft used in his article on understanding BABIP ((GB% * .235) + (FB% * .137) + (LD% * .716) + (BUNT% * .388)), we tweaked the number we multiply GB% by to more accurately reflect a player’s speed (.293 for players with at least 25 SBs and .226 for players with 0-1 SBs and at least 500 PAs instead of the initial .235 number). We’re getting closer to a more accurate xBABIP.

Now it’s time to take even more factors into consideration. The most glaring oversight I see in Cockroft’s formula is the double usage of bunts factored into both GB% and BU%. This is an easy enough fix; we simply subtract the bunts from the ground balls to come up with an updated GB%, separating the two categories. Otherwise, players have GB% + LD% + FB% + BU% that equal greater than 100%.

The second correction to Cockroft’s formula is to eliminate home runs from a player’s FB%. This is xBABIP after all, and HRs do not count as balls in play. This is another easy fix. We simply subtract HRs from a player’s fly balls to come up with new batted ball percentages. Obviously, lowering the number of fly balls will lower a player’s FB% and thus raise their other percentages. Why is this relevant? Fly balls have the lowest success rate (.137 in 2011) of all the batted ball types. By lowering FB% and raising GB% and LD%, which have much higher success rates, we inch closer to a more accurate xBABIP.

Here is a table reflecting how the adjustments we have made in Parts 1 and 2 have influenced a more accurate xBABIP (New xBABIP) compared to Cockroft’s xBABIP.

 Name BABIP xBABIP New xBABIP Adrian Gonzalez .380 .307 .311 Matt Kemp .380 .307 .346 Emilio Bonifacio .372 .354 .371 Michael Bourn .369 .356 .377 Michael Young .367 .335 .339 Alex Avila .366 .302 .311 Miguel Cabrera .365 .308 .320 Hunter Pence .361 .291 .298 Alex Gordon .358 .304 .312 Dexter Fowler .354 .312 .312 Jose Reyes .353 .307 .330 Ryan Braun .350 .301 .337 Joey Votto .349 .335 .347 Andre Ethier .348 .326 .328

I considered switching from stolen bases to a speed score (speed score is a statistic developed by Bill James that rates a player on speed and base running ability) as the speed indicator to find the average BAGBIP (batting average on ground balls in play) among the fastest and slowest runners in the MLB. However, after finding these averages, I discovered that they are very similar to the averages I came out with using SB as the speed indicator, and not worth the switch.

This is how speed scores (Spd) compared to SBs.

 Year >6.0 Spd <2.5 Spd >25 SB 0-1 SB and >500 PA 2011 .285 (29 players) .242 (29 players) .292 (22 players) .239 (21 players) 2010 .290 (23 players) .211 (24 players) .298 (25 players) .219 (29 players) 2009 .290 (23 players) .228 (26 players) .290 (28 players) .220 (26 players) 2008 .294 (24 players) .228 (21 players) .290 (23 players) .224 (28 players) AVG .290 (99 players) .228 (100 players) .293 (98 players) .226 (104 players)

As you can see, the averages came out similarly. I did, however, notice a few interesting trends when breaking down the speed scores further. Let’s take a look.

 Spd score Average BAGBIP 6.1-6.5 .282 (33 players) 6.6-7.0 .287 (23 players) 7.1-7.5 .291 (18 players) 7.6-8.0 .299 (16 players) >8.1 .309 (9 players)

 Spd score Average BAGBIP 0-1.5 .232 (14 players) 1.6-2.0 .233 (42 players) 2.1-2.4 .222 (44 players)

Notice the average BAGBIP increased corresponding to the increase in speed scores among the players with scores of 6.1 or higher. Interestingly, the reverse did not occur with the group with speed scores less than 2.5, with the fastest players among that group (those with speed scores of 2.1 to 2.4) having lower BAGBIPs than the slower groups.

What does this tell us? Simply put, small increases in speed among fast players helps more than small decreases in speed among slow players hurts. The explanation, I hypothesize, is that a small increase in speed can make a difference in the outcome of infield ground batted balls, whereas there is a point where small decreases in speed do not affect the outcome of close plays at first. The slower players BAGBIPs start to plateau once they’ve reached the point that they are slow enough to not beat out infield ground balls with any regularity. Neither a 2.4 speed score player nor a 0.7 speed score player has enough speed to beat out infield hits. Their ground ball hits tend to come from ground balls that escape the infield. This is influenced much more by luck.

While there are still further ways to adjust an xBABIP formula, I am satisfied enough with the results to stop here. Hopefully you have gained a better understanding of how to properly use xBABIP to determine the extent to which a player’s BABIP was determined by luck. This will assist you in identifying players due for regression to the mean, while allowing you to bid confidently in players others may be misinterpreting as lucky/unlucky.