Is There Any Advantage in Keeping Fielders on Their Toes?
Ever heard someone say that a pitcher wants to "keep his fielders on their toes" to succeed? Let me present you a quote from an online paper to explain this idea:
Greg Maddux is the type of pitcher that works his spots around the plate and throws a healthy diet of off speed pitches. He will keep fielders on their toes by getting the batter to hit ground balls. This way, the fielders aren't just standing around falling asleep while the game is being played.
This attitude has always seemed a little strange to me, to be honest. Personally, I've never seen a fielder just standing around because the pitcher allows so few balls in play. In fact, and this is based on personal observation only, it seems to me that pitchers who allow many walks and strikeouts (in other words, pitchers who do not allow many balls in play) tend to be fly-ball pitchers, and fly-ball pitchers tend to allow lower Batting Average on Balls in Play (BABIP) than ground-ball pitchers.
I decided to take a look at this question by conducting five studies. What I found was very interesting.
First of all, I took every pitcher season with at least 350 Batters Faced (BFP) between 2002-04 using the Lahman Database. This gave me 529 pitcher seasons, more than enough for a large sample size. The first Study I ran was pretty simple: I calculated each pitcher's Balls in Play per Batter Faced (BIP/BFP), and split the data into three groups, pitchers who were one Standard Deviation (SD) above average in BIP/BFP, pitchers who were one SD below, and the rest. What interests me are the first two. Take a look at the results:
N BFP BIP BIP/BFP H-HR BABIP R IP RA High BIP/BFP 67 44600 35465 0.795 10502 0.296 5686 10360 4.94 Low BIP/BFP 90 62826 39982 0.636 11353 0.284 6090 13502 4.06
The first (High BIP/BFP) is the group that is supposed to keep batters on their toes. As you can see, its BABIP against is higher than that of the second group, and by quite a bit. It seems that the more Balls in Play (BIP) you allow, the higher your BABIP becomes. There is one problem, however, with this initial Study, and it has to do with the last column in the above table. The Run Average (RA) of the first group is much higher than the RA of the second group. This creates a potential bias for which we need to control. That forms the basis of my next two studies.
What I did to attempt to control for this bias is split the 529 pitchers into three groups: pitchers who were a SD below average in RA, pitchers who were a SD above average in RA, and the rest. For studies two and three, I used the first two groups.
Study #2 focused on the bad pitchers, those that were one SD below average in RA, of whom there were 102. I then repeated the process I used in Study #1 on this group of pitchers, and came up with the following result:
N BFP BIP BIP/BFP H-HR BABIP R IP RA High BIP/BFP 19 8548 6811 0.797 2083 0.306 1309 1914 6.16 Low BIP/BFP 16 8224 5468 0.665 1646 0.301 1266 1810 6.30
Again, the pitchers who supposedly do not keep their fielders on their toes had a lower BABIP. This time there was no bias in terms of RA; in fact, the group that allowed more BIP/BFP had a lower RA.
Let's move on to Study #3 which replicates Study #2, but using the best players, those with the lowest RAs in the data set. Here are the results:
N BFP BIP BIP/BFP H-HR BABIP R IP RA High BIP/BFP 11 8655 6576 0.760 1822 0.277 763 2112.3 3.25 Low BIP/BFP 10 6439 3655 0.568 1006 0.275 510 1598.3 2.87
The same problem that arose in Study #1 is apparent here as well. The low BIP/BFP group has a much better RA than the high BIP/BFP group. This is because the low BIP/BFP group strikes out so many more batters, and it's a problem that cannot really be addressed within the parameters of this study. However, the BABIPs for the groups are so close that when we factor in that the BABIP for the low BIP/BFP group has to be regressed a little more due to a smaller sample, we can conclude that there is no difference in BABIP between good high BIP/BFP pitchers and low BIP/BFP pitchers. I think. (I say that because in a minute I will show this conclusion to be incorrect). Still, we see no evidence that "keeping fielders on their toes" will result in better fielding.
Studies #4 and #5 repeat what I did in studies #2 and #3, respectively, except I adjusted pitcher BABIPs for their team BABIP, thus filtering out other potential biases (fielders, park, etc.). The groups are the same; the only thing that will change are their BABIPs. Let's take a look at the high-RA pitchers:
N BFP BIP BIP/BFP H-HR adjBABIP R IP RA High BIP/BFP 19 8548 6811 0.797 2083 0.312 1309 1914 6.16 Low BIP/BFP 16 8224 5468 0.665 1646 0.309 1266 1810 6.3
Adjusted BABIP was calculated by taking each player's hits on balls in play, finding his expected hits on balls in play based on his team's BA, and then dividing hits on balls in play by expected hits on balls in play and multiplying that by the average BABIP of the group.
Here, we again see that those who allow few BIP/BFP have a better BABIP. Let's move on to my final study:
N BFP BIP BIP/BFP H-HR adjBABIP R IP RA High BIP/BFP 11 8655 6576 0.76 1822 0.269 763 2112.3 3.25 Low BIP/BFP 10 6439 3655 0.568 1006 0.256 510 1598.3 2.87
Now you can see why I said that among pitchers with low RAs, the ones who have low BIP/BFP rates also have better BABIPs. Adjusting for team makes a big difference.
Anyways, as you can see, in all five studies, the pitchers who allowed few BIP/BFP, pitchers, who could also be called "three true outcomes" pitchers because they allow many home runs, walks, and strikeouts, had much better BABIPs than those who allow the ball to be put into play most often. Why? It seems to me that the positive correlation between strikeouts and BABIP is part of it, probably a large part. But no matter what the exact reason, it seems fairly clear that this piece of conventional wisdom is wrong. "Keeping fielders on their toes" does not ensure better defense; in fact, it does the opposite.
Note: I also thought of doing a study using pitch count data since throwing more pitches results in longer at-bats, and fielders that supposedly get bored. However, as walks and strikeouts are really what dictate how many pitches a pitcher throws, and since they largely dictate a pitcher's BIP/BFP, the correlation between BIP/BFP and Pitches/BFP is almost perfect (.95 r-squared). In other words, doing such a study would add nothing new.
[Additional reader comments and retorts at Baseball Primer.]