Anatomy of a pitch: Sliderby Josh Kalk
May 22, 2008
Last week, we looked at the attributes that go into making a good curveball, so it makes sense to follow up that piece with a look at sliders. Sliders are kind of like the "middle children" of pitches, with more cut than a cut fastball but less cut than a curveball. For this reason, sliders can often be hard to identify, as many pitchers throw their sliders with very similar movement to their curves or cutters. This resemblance also provides a lot of leeway to the pitcher for throwing a slider, as some are thrown hard with little vertical movement compared to the fastball and others are thrown softer with a big speed difference and large vertical movement. Some pitchers will actually throw two different types of sliders.
Like last week, we will investigate which attributes calculated by PITCHf/x correlate well to a successful slider. Again, we will be defining success using runs100, a metric invented by John Walsh that basically measures how many runs a pitcher saves per 100 pitches of a given type. For a pitcher to qualify for the study, I required at least 200 sliders that were tracked by PITCHf/x.
When comparing fastballs to curveballs last week, we used a side view of the pitch to examine the hump, so let's start there and see what happens to the hump when the pitch is a slider. And we might as well look at one of the best sliders in the game, that of C.C. Sabathia.
Again, this image was created by averaging all of C.C.'s fastballs and sliders. The black line between the two pitches represents the largest hump of the pitch, and the tick marks on the fastball and slider are the location of the pitch after 0.075 seconds, which represents the end of the information-gathering stage of the batter according to Bob Adair in his book The Physics of Baseball. Although there is still a small hump with the slider, the excellent downward movement shows you why Sabathia is the reigning Cy Young winner. His slider is almost completely hidden within his fastball while dropping off the table as it approaches home plate. I believe that the term "late break" might come from this—the break is late because it is hidden so well in his fastball.
Although C.C. is an excellent example, other pitchers can effectively hide their sliders in a similar manner because the difference in vertical movement for a slider is much less than that of a curveball. But what about horizontal movement? A fastball will tail in to a batter of the same handedness as the pitcher, but a slider is just the opposite: It tails away from the same batter because of the different spin applied by the pitcher.
So maybe we should change our angle and instead look at the pitch from above:
Here, we are looking down at Sabathia's fastball and slider. The plate is at horizontal position of 55 feet, and I have again begun tracking the pitch 55 feet away to make sure the ball has been released by the pitcher. I am calling the other dimension "depth" because I used "vertical" in the previous plot to describe up and down.
You can probably picture exactly what is going on here: Sabathia is left-handed, so he is releasing the ball near the mound a little less than three feet away, and when it crosses home plate it is crossing right at the heart of the plate (since I am taking the average path of all his fastballs). You can see that his fastball is slightly bending, trying to go more positive or closer to a left-handed batter. His slider is behaving exactly the opposite way, bending away from a left-handed batter and ending up near the edge of home plate (thus, the average Sabathia slider is on the outer part of the plate to left-handed batters and on the inner part of the plate to right handers). Again, I have placed tick marks representing the ball at 0.075 seconds, and the line between represents the largest difference in position before the pitches cross paths. So here, too, the difference between Sabathia's slider and his fastball is very small until the ball gets very close to home plate, and then the slider darts away to left handed batters.
Like I did last week with the curveball hump, I tried to correlate this maximum difference and its location to the value of runs100, but I got very poor results. So instead, I looked that the difference between the two pitches at intervals of 0.01 seconds and ran correlations at every step. Much to my surprise, the interval that had the largest correlation was the 0.08-second slice, which is almost exactly what Adair had suggested was the end of the information-gathering phase for the batter. Before that point, the correlation is poor, but it gets better and then stabilizes around 0.08 seconds before dropping again as the ball approaches home plate.
You might be wondering why I was surprised by this result. The reason is it is completely different from the result that I got for curveballs. As with the correlation for sliders, the correlation for curveballs starts low, but the correlation for the curveball rises much more slowly until a time of nearly 0.2 seconds where the ball has traveled almost half way to home plate. The correlation briefly levels off and then falls right before reaching home.
So what is going on here? Why are sliders matching up so well to Adair's idea of the end of the information-gathering stage but curveballs are not? I don't have a full understanding, but it appears that because curves tend to produce a larger hump, a fast-reacting hitter has slightly more time in which to put on the brakes (or alter his swing) when he realizes that the pitch is not a fastball. Because sliders tend to stay hidden much further down the line, a batter who is fooled in the information-gathering stage has much less time to recover. This "secondary reaction" might be something that could be measured in batters and it would be interesting to see if that correlated to their ability to hit (or not swing at) curves and sliders.
Looking at some other variables, slider success, as measured with runs100, showed greater correlation to the difference in horizontal movement compared to the fastball (0.27) but less correlation to the difference in vertical movement (0.15). Again, this is opposite to what we saw last week with curveballs. It is possible that extra horizontal movement in curves and extra vertical movement in sliders are additional giveaways for a batter. So if you are a pitcher and you want to hide your off-speed pitches as best as you can, try for large vertical movement with your curve and large horizontal movement with your slider.
Speed difference also played a smaller role with sliders than it did with curveballs. Thus, although a big speed difference is important for a curveball, the key for a slider lies in its movement. Difference in release point also is somewhat correlated with success, though it should be noted that all of these variables would be considered weakly correlated. It is possible that a further breakdown of different slider types would produce better results, but as of yet no pitch classification has broken down sliders into different types.