The defensive shift: pitch by pitchby Jonathan Hale
August 05, 2009
With the recent trend of using a massive infield shift against some power hitters, modern baseball strategy has clearly embraced the fact that who is at the plate can have a huge difference on where the ball is going to go. But what about the pitch the man on the mound is throwing? Sometimes you will see an middle infielder join a mound conference to find out what is coming, or relay information to his double-play partner once he sees the signs. But should he be more interested in location, or the type of pitch that is coming? Would outfielders be more effective defensively if they repositioned themselves based on the incoming pitch as well?
Using HITf/x, we can look at the horizontal angle of the ball leaves the bat, broken down by both what type of pitch is coming and where it is located. As with everything written about HITf/x so far, this is an extremely advance peek behind the scenes, with all the problems associated with only having one month of data to play with. The sample sizes get ever trickier because we have to split up grounders and fly balls, as they can be completely different. (i.e. for the famous "Ortiz shift", the outfield usually plays straight up or even slightly to the opposite field because despite being a dead-pull power hitter when the ball stays on the ground, that type of hitter is able to lift the ball when going the other way).
As a result, for the following analysis I have had to leave out several groups of hitters, the first being left-handed hitters. What a tease, I know—but everything I've seen so far from PITCHf/x so far suggests that their spray patterns are a mirror image of right handed batters' (and that the baseball lore such as that lefties as a group are all "good low ball hitters" is a myth). If there is a significant difference, it is not detectable with what we've got so far, and we're looking for incredibly broad trends, anyway, as obviously all hitters on both sides have unique swings. So let's get started looking at what the majority of the league does with four-seam fastballs:
What you’re looking at here is the average angle that a (four seam) fastball is put in play, based where it was thrown (as seen from behind the plate). I have adjusted the data so that 0 degrees (in yellow) is a ball hit right back up the middle, blue (-45 degrees) is down the first base line, and red (+45 degrees) is right down the third base line. You'll notice that the average angle never gets all the way to blue, because of the natural tendency for hitters to pull the ball (see Dave Allen's article on this tendency). Each contour line represents five degrees of difference in the average horizontal angle that the ball is hit, and going by Harry Pavlidis' introduction to HITf/x, I have used 7 degrees of elevation as the cutoff between a grounder and a fly ball.
Hitters clearly don't pull inside fastballs on the ground when facing left-handed pitchers as often as they do against right-handed pitchers. I would suggest this is because the pitch starts on the other side of the mound, and travels across the plate at an angle their hands to get there, so they are getting jammed, or having to "fight it off" the other way (an effect that will come up with other pitches as well). Only when it is down and in do they get around on the ball and drive it towards the third base line. I wrote an article ages ago that suggested that throwing a rising fastball down and in was never a good idea, which this speaks to.
However, they are more likely to pull a ball in the air against opposite-handed pitchers, especially when pitches are left up in the zone. As Mike Fast showed, pulled fly balls are more likely to go for home runs, and so we are almost certainly looking at one cause of the platoon advantage. If a lefty is going to come inside with a fastball, he has to keep it low and go for one of those ground balls.
But back to the task at hand, if a fastball (that has any chance of being middle-in) is on its way, the defense should pay attention to who is on the mound. If the pitcher is right handed, the infield should play the hitter to pull, and the outfield the other way; vice versa if the pitcher is a lefty. If a pitcher is painting the outside corner, the difference is nowhere near as great, although lefties have the advantage on balls pulled in the air if they can pick out the outer edge of the zone.
Time for another compromise, and a somewhat uncomfortable one: lumping breaking balls in together. Trust me, I know there's a big difference in trying to hit a slider and a curveball, but for the sake of this investigation, it isn't enough to split them (both for the sake of sample size and blinding you with identical graphs). I look forward to the day when we can examine more subtle differences like this (or between other even more similar pitches like cutters and sliders), but they have very similar horizontal spray patterns as they both move in the same direction and come in slower than a fastball.
As expected, we see a lot more pulled pitches, but the difference when facing right- and left-anded pitchers is even more pronounced, as almost all right handed curves hit for grounders are pulled (likely "rolled over"), with the exact opposite when hitting off a lefty. In defensive terms, this leads to the same general rule as before: With a right handed pitcher on the mound, the infield should be set more to the third base line when a breaking pitch is coming, and the outfield straight up or slightly away; but the exact reverse with a lefty on the mound.
Along with the knuckleball, I've glossed over different kinds of fastballs because they're too hard for PITCHf/x to distinguish from each other, or even sliders at this point. But extreme sinkerballers are of particular interest due to their apparent ability to limit a hitter's BABIP. Defining a true sinking fastball as one that has more than three inches of drop (a PITCHf/x value of 6) compared to a normal fastball, we can see that they are very rarely pulled when thrown by right-handed pitchers (the sample size is too small to evaluate left-handed pitchers). The first graph shows right-handed batters' results when hitting sinkers in the air, the second shows their results when sinkers on the ground.
On the ground, the average ground ball off a sinker goes straight back up the middle, even when the pitch is on the very inside edge of the plate (a pitch that is breaking into the hands of batters like a slider from a lefty) . And almost no matter where they are thrown, fly balls hit off of sinkers tend to go to the opposite field, the only pitch for which this is the case.
In a previous article on what makes a home run pitch, I showed that the sinker is the only pitch whose home-run rate is almost the same from one side of the plate to the other. Now we know that part of this effect sinker is due to hitters having difficulty pulling the sinker, even when they do manage to get it in the air. It seems that unlike other pitchers, sinkerballers do not have to worry as much about keeping the ball down and away.
The results for changeups are too ugly and inconclusive for me to show here, but take a look if you like. All I can tell from that mess is that the only time anyone ever hits a changeup to the opposite field is when it is thrown by a left-handed pitcher to the outer edge of the strike zone (likely due to changeups typically having tailing action away from the batter than due to them being behind it).
- In general, play the infield more to pull and the outfield away with a like-handed hitter pitcher matchup.
- Do the exact opposite with opposite-handed pitchers and hitters.
- Right-handed batters pull almost every grounder when facing breaking balls from the right side, and vice-versa from the left.
- Don't play hitters to pull, either in the the infield or outfield, with a sinkerballer on the mound.
- Part of what keeps sinkers in the park is likely their tailing action as well as their drop.
- If a pitcher is paints the outside corner with a changeup, batters are likely to take it the other way.
References and Resources
Although any mistakes are purely his own, the authur would like to thank Daniel Brooks of brooksbaseball.net for his invaluable statistical advice, who would in turn like to give Dave Allen due credit for his examples from the PITCHf/x summit. Collaboration runs rampant in the PITCHf/x community, and that is amazing.
Jonathan Hale be found mixing cold hard statistics with reactionary conjecture at The Mockingbird . He welcomes questions and criticism via e-mail.