R code to find optimal launch angles from Statcast data

optimal_angle.R

# read in tidyveres package

library(tidyverse)

# read in the Statcast data for the 2017 season

# the file statcast2017.csv can be downloaded from the folder
# http://www-math.bgsu.edu/~albert/statcast/

sc <- read_csv("statcast2017.csv")

# only consider balls put in play

sc_ip <- filter(sc, type == "X")

# exploring relationship between launch angle,
# exit velocity, and hit/out

# first remove home runs and focus on
# P(hit | in play and not home run)
# Also remove sac hits and sac flies
# define the hit variable

sc_ip  %>%  filter(events != "home_run",
              ! (events %in%
             c("sac_bunt", "sac_fly", 
               "sac_fly_double_play"))) %>% 
     mutate(hit = ifelse(events %in%
      c("single", "double", "triple", "home_run"), 
         1, 0)) -> sc_ip2

# only look at balls hit in the air and 
# define a factor variable Hit

sc_ip3 <- filter(sc_ip2, launch_angle > 0)
sc_ip3$Hit <- as.factor(sc_ip3$hit)

# scatterplot of launch angles and exit velocities

ggplot(sc_ip3, 
       aes(launch_angle, launch_speed, color=Hit)) +
  geom_point(size = 0.7, alpha = 0.1) +
  ylim(25, 125) +
  ggtitle("Scatterplot of Launch Angle
  and Launch Speed for Balls in Air") +
  theme(
    plot.title = element_text(
      colour = "blue",
      size = 16,
      hjust = 0.5
    )
  ) +
  xlim(0, 40) + ylim(50, 115)

# focus on hits with a specific velocity 
# look at the pattern of binned data

# write a function to do this for any speed

efit_model2 <- function(ev, width=1, delta=1){
  # width is for the binning of launch angle
  # delta determines blur of launch speed
  # only considering launch angles in (0, 30)
  d <- filter(sc_ip3, 
              launch_speed > ev - delta / 2,
              launch_speed < ev + delta / 2,
              launch_angle < 30)
  d$LA <- cut(d$launch_angle, 
              breaks = seq(0, 30, by=width))
  d %>% 
    group_by(LA) %>% 
    summarize(N = n(),
              H = sum(hit),
              Probability = H / N) -> S
  S %>% mutate(Mid = seq(0, 30 - width, by=width) + 
    width / 2,
    launch_speed = ev)
}

# run this function for a collection of exit velocities

map_df(c(75, 80, 85, 90),
       efit_model2, width=2, delta = 2) -> out
out$Launch_Speed = paste("Launch Speed =",
                         out$launch_speed, "mph")
write_csv(out, "plotdata.csv")

map_df(c(95, 100, 105),
       efit_model2, width=2, delta = 2) -> out
out$Launch_Speed = paste("Launch Speed =",
                         out$launch_speed, "mph")
write_csv(out, "plotdata2.csv")

##### graphing part

d <- read_csv("plotdata.csv")
d2 <- read_csv("plotdata2.csv")

ggplot(d, aes(Mid, Probability)) +
  geom_point(color="red") +
  geom_smooth(method="loess") +
  facet_wrap(~ Launch_Speed, ncol=2) +
  xlab("Launch Angle (degrees)") +
  ylab("Probability(Hit)") +
  theme(strip.text.x = element_text(size=16, color="red"),
        text = element_text(size=16))
        
ggplot(d2, aes(Mid, Probability)) +
  geom_point(color="red") +
  geom_smooth(method="loess") +
  facet_wrap(~ Launch_Speed, ncol=2) +
  xlab("Launch Angle (degrees)") +
  ylab("Probability(Hit)") +
  theme(strip.text.x = element_text(size=14, color="red"),
        text = element_text(size=16))

# graph the optimal angles

df <- data.frame(Launch_Speed = seq(75, 105, by = 5),
                 Optimal_Launch_Angle = c(20, 17,
                                  15, 13, 12, 11, 10 ))
ggplot(df, aes(Launch_Speed, 
               Optimal_Launch_Angle)) +
  geom_point(size = 3, color="red") +
  ggtitle("Optimal Launch Angle") +
  theme(
    plot.title = element_text(
      colour = "blue",
      size = 18,
      hjust = 0.5
    )
  ) +
  theme(strip.text.x = element_text(size=16, color="red"),
        text = element_text(size=16))