Reproducing MLB Baseball Savant

The goal of this tutorial is to recreate several graphics from Major League Baseball’s Savant, which utilizes Statcast data. Specifically, we will recreate three visualizations for Gerrit Cole.

• Gerrit Cole | Baseball Savant

Complete materials for the workshop can be found on GitHub:

• Part One: daviddalpiaz/cmsac-2024
• Part Two: bayesball/Workshop

This tutorial makes heavy use of dplyr and ggplot2. Documentation, including cheat sheets, can be found at:

• dplyr Documentation
• ggplot2 Documentation

Setup

# list of required packages
packages = c("ggplot2", "dplyr", "readr", "ggdensity", "ggforce")

# install packages if not already installed
for (package in packages) {
    if (!requireNamespace(package, quietly = TRUE)) {
        install.packages(package)
    }
}

# load required packages
library("ggplot2")
library("dplyr")
library("readr")
library("ggdensity")
library("ggforce")

# load data
cole2024 = read_csv("https://daviddalpiaz.github.io/cmsac-2024/data/cole2024.csv")

• Statcast Search
• Statcast Search CSV Documentation
• Statcast Data Reference | Analyzing Baseball Data with R
• saberpowers/sabRmetrics

# preview data
cole2024

# preview columns of interest
cole2024 |>
    select(
        pitch_type,
        pitch_name,
        release_speed,
        release_spin_rate,
        plate_x,
        plate_z,
        pfx_x,
        pfx_z
    )

# calculate mix pitch and pitch summary statistics
cole2024_stats = {
    cole2024 |>
        group_by(pitch_type, pitch_name) |>
        summarise(prop = n() / nrow(cole2024),
                  ave_velo = mean(release_speed),
                  ave_spin = mean(release_spin_rate),
                  .groups = "drop"
        ) |>
        arrange(desc(prop))
}
cole2024_stats

Pitch Velocity Distribution

Starter Code and Graphic

ggplot(cole2024) +
    aes(x = release_speed) +
    geom_density()

Full Code and Final Graphic

ggplot(cole2024) +
    aes(x = release_speed,
        y = 100 * after_stat(count) / nrow(cole2024),
        fill = pitch_type
    ) +
    geom_density(alpha = 0.5) +
    theme_bw() +
    xlim(c(70, 100)) +
    scale_y_continuous(
        limits = c(0, 13),
        breaks = setNames(
            seq(0, 12, by = 2),
            paste0(seq(0, 12, by = 2), "%"))
    ) +
    xlab("Pitch Speed (mph)") +
    ylab("Frequency of Speed") +
    labs(fill = "Pitch Type")

Pitch Arsenal

Starter Code and Graphic

# get pitch types in order of frequency
cole2024_stats |>
    pull(pitch_type) |>
    dput()

c("FF", "KC", "FC", "SL", "CH", "SI")

# make pitch types a factor ordered by frequency
cole2024$pitch_type = factor(
    cole2024$pitch_type,
    levels = c("FF", "KC", "FC", "SL", "CH", "SI"))

• Where Is the Strike Zone? | Analyzing Baseball Data with R

# create geom to visualize the strikezone
geom_strikezone = function(sz_top = 3.8, sz_bot = 1.1) {
    plate_width = 17 + 2 * (9 / pi)
    sz_left = -(plate_width / 2) / 12
    sz_right = (plate_width / 2) / 12
    strikezone = data.frame(
        x = c(sz_left, sz_left, sz_right, sz_right, sz_left),
        y = c(sz_bot, sz_top, sz_top, sz_bot, sz_bot)
    )
    geom_path(
        mapping = aes(.data$x, .data$y),
        data = strikezone,
        linewidth = 0.5,
        linetype = 1,
        color = "black"
    )
}

cole2024 |>
    ggplot() +
        aes(x = plate_x, y = plate_z) +
        geom_point(
            aes(color = pitch_type),
            alpha = 0.6
        ) + 
        geom_strikezone()

Full Code and Final Graphic

location_plot = {
    ggplot(cole2024) +
        aes(x = plate_x, y = plate_z) +
        xlim(c(-3, 3)) +
        ylim(c(-2, 6)) +
        xlab("") +
        ylab("") +
        theme_bw() +
        coord_fixed() +
        facet_wrap(vars(pitch_type))
}

location_plot +
    geom_hdr(
        method = method_kde(h = 0.75),
        probs = c(0.90, 0.70, 0.50, 0.40, 0.30, 0.20),
        show.legend = FALSE,
        aes(fill = after_stat(probs)),
        alpha = 0.75
    )  +
    scale_fill_brewer(
        palette = "RdBu",
        direction = -1
    ) +
    geom_strikezone()

location_plot +
    geom_point(
        aes(fill = pitch_type),
        alpha = 1,
        show.legend = FALSE,
        pch = 21
    ) +
    geom_strikezone()

Movement Profile

Starter Code and Graphic

cole2024 |>
    mutate(hb = 12 * pfx_x) |>
    mutate(ivb = 12 * pfx_z) |>
    ggplot() +
        aes(x = hb,
            y = ivb,
            colour = pitch_type
        ) +
        geom_point(alpha = 0.75)

• What is Induced Vertical Break? | Reddit
• The Physics of Baseball | Alan Nathan

Full Code and Final Graphic

set.seed(2024)
cole2024 |>
    mutate(hb = -12 * pfx_x) |>
    mutate(ivb = 12 * pfx_z) |>
    slice_sample(n = 200) |>
    ggplot() +
        aes(x = hb,
            y = ivb,
            colour = pitch_type
        ) +
        geom_hline(yintercept = 0, linewidth = 0.1) +
        geom_vline(xintercept = 0, linewidth = 0.1) +
        geom_circle(
            data = data.frame(
                x0 = 0,
                y0 = 0,
                r = 6 * 1:4,
                lt = factor(c(2, 1, 2, 1))
            ),
            mapping = aes(
                x0 = x0,
                y0 = y0,
                r = r,
                linetype = lt
            ),
            inherit.aes = FALSE,
            linewidth = 0.25,
            show.legend = FALSE
        ) +
        geom_point(
            alpha = 1,
            size = 2
        ) +
        geom_text(
            data = data.frame(
                x = c(-22, -16, -10, -4),
                y = 1,
                text = c("24\"", "18\"", "12\"", "6\"")
            ),
            mapping = aes(x = x, y = y, label = text),
            inherit.aes = FALSE,
            family = "mono"
        ) +
        scale_x_continuous(limits = c(-24, 24), breaks = 12 * -2:2) +
        scale_y_continuous(limits = c(-24, 24), breaks = 12 * -2:2) +
        theme_void() +
        coord_fixed() +
        xlab("Horizontal Break (Inches)") +
        ylab("Induced Vertical Break (Inches)") +
        labs(colour = "Pitch Type")