Pitcher Ohtani's edgy slider that I saw at the WBC The name is "Sweeper"
The success of the Japanese national team in the WBC has become a big topic recently. In recent years, I haven't had much time to watch baseball broadcasts on terrestrial TV, so I'm watching this WBC closely because it's a big topic.
Among the Japanese players, Ohtani is the most popular player and has achieved results in pitching and hitting, but in this article Consider Ohtani as a pitcher.
I was watching Otani's pitching live on the WBC for the first time in a long time, but I felt that the content of the pitching was clearly different from the previous pitcher. As it was said in the commentary of the relay, instead of throwing a lot of fastballs and splits like before, he is throwing a lot of sliders, and moreover, sliders that don't sink much.
Pitcher Ohtani pitched against China and Italy, but a site where you can refer to MLB stats information Baseball Savant According to the 2019 edition, the lineup of pitches for the two games will be as follows: The pitch I used the most was the Sweeper. It's not very familiar in Japan, is it? To put it simply, would you call it an ugly slider?
I also didn't know until now, but the type of pitch measured in MLB is distinguished by "slider" and "sweeper". The image of a slider is that it bends in the horizontal direction and sinks slightly downward, but the sweeper seems to have a sharper curve than the slider and does not fall down much. Certainly, when watching the WBC, I encountered a scene where pitcher Ohtani took a whiff from a batter with this sweeper.
Difference between sweeper and slider
Let's consider the difference between a sweeper and a slider based on the pitches of Otani, who was also active as a pitcher last year (2022) and the year before last (2021).
Similarly, using data obtained from Baseball Savant, let's graph the amount of change in Otani's pitching in 2021 and 2022.
Horizontal variation (pfx_x) and vertical variation (pfx_z) are plotted and color-coded by type of ball. The horizontal direction is considered from the catcher's perspective, with negative values indicating the shooting direction and positive values indicating the sliding direction. 0 (zero) in the vertical direction represents the value when it falls as it is according to gravity.
It seems that the coordinate points are well divided according to the type of ball, but from now on, we will focus on sweepers (red points) and sliders (light blue points). As you can see from the graph, The sweeper has a larger lateral curve than the slider and does not fall too much I understand.
Another graph shows the coordinates (catcher's line of sight) and contour lines as they pass through home base for sweepers and sliders.
There is also a difference in the position when passing the home base between the sweeper and the slider.
Catcher Kai, who received Ohtani's pitching, said in an interview after the war with China that Ohtani was throwing three types of sliders. So, I thought that maybe sweepers could be classified into two types, and sliders might be three types together.
This is where statistical methods come in. Using 6 variables: amount of change (horizontal, vertical), coordinates on the base (horizontal, vertical), ball speed, amount of rotation, Let's classify sweepers into two groups by clustering method.
Classification and comparison of sweepers with Gaussian mixture
here, As a clustering method, we use the "Gaussian mixture distribution", which is often used in machine learning. It is a method of probabilistically expressing which class each individual (here, each pitch) belongs to. When classifying data that is difficult to classify clearly even if you draw a scatter diagram, like the sweeper this time may be used.
In JMP, under Cluster Analysis [normal mixture] It can be executed by selecting the menu.
This report classifies 1,423 sweepers into two groups using six variables, excluding data with abnormally low rotation speeds or missing data. In the JMP report, two clusters can be plotted (biplot) on the principal component space (principal component 1, principal component 2). It doesn't seem to be done.
Looking at the "Cluster Summary" and "Cluster Mean" reports, It can be seen that the 950 balls belonging to cluster 2 have an average faster speed and a higher number of rotations than the 473 balls belonging to cluster 1.
Somehow Cluster 2 feels like a better sweeper. So, here I named the sweeper of cluster 2 "Sweeper High" and the sweeper of cluster 1 "Sweeper Low". I don't know if it matches the 3 types of sliders that Kai is talking about, but for the time being, I classified them into 3 types of sliders "Sweeper High", "Sweeper Low", and "Slider" statistically.
These sliders are compared by ball speed (release_speed), spin rate (release_spin_rate), base coordinates (plate_x, plate_z), and amount of change (pfx_x, pfx_z). For these items, connect the average values of each slider with a line.
The big difference between "Sweeper High" and "Sweeper Low" clustered with mixed Gaussian distribution is ball speed and number of revolutions.
"Sweeper High" and "Slider" have almost the same ball speed and number of rotations, but it is interesting that there is a big difference in the amount of change and the coordinates at the base. From the batter's point of view, throwing balls with the same ball speed and spin rate but different amount of change would be quite troublesome.
Visualize the trajectory of 3 types of sliders
The data we are using this time also has the coordinates when the ball was released, so let's visualize the trajectory of the three types of sliders shown above.
The animation below was created using JMP's "Bubble Plot", using the most frequent point (the point with the largest density when estimating the density) for the coordinates at the time of release and the coordinates passing through the base for each slider.
(Caution: The animation below shows the coordinates at the time of release and when passing through the base in a straight line.)
Was the evolution of the sweeper a factor in his success as a pitcher?
Finally, we will compare the usage of the three types of sliders in 2021 and 2022.
The figure below is a mosaic plot that can be drawn in JMP's Bivariate Relationships. Even a quick glance reveals a clear difference between 2021 and 2022.
By 2022, nearly 90% of sliders will be “Sweeper High”. In other words, it is a slider that has a high ball speed and a high number of rotations, bends well and does not fall off very much. In the major leagues, it is said that the sweeper is an effective pitch to keep batters down, but Otani has become able to throw even higher quality sweepers, so it can be said that the pitcher's performance in 2022 was good. It may be.
It seems that Ohtani won't have another opportunity to pitch in the WBC, but I'm looking forward to seeing how he will handle the slider in the 2023 MLB season.
Of course, I have high expectations for him as a hitter as well!
