Explores impact of length of plate appearance using 2018 Retrosheet data

pitch_length.R

# Load tidyverse packages and read in the Retrosheet data

library(tidyverse)
load("~/Dropbox/Google Drive/Retrosheet/pbp.2018.Rdata")

# if you have trouble getting the 2018 Retrosheet data, you can
# use several complete Retrosheet datasets from previous seasons
# see http://www-math.bgsu.edu/~albert/retrosheet/

# Create new variables:  pseq containing only the pitch
# results, pseq_length gives the length of the PA, and
# Count records the ball strike count

d2018 %>% 
  mutate(pseq = str_remove_all(PITCH_SEQ_TX, 
                              "[.>123N+*]"),
     pseq_length = str_length(pseq),
     Count = paste(BALLS_CT, STRIKES_CT, sep="-")) -> d2018

# Only want to consider PA's where the PA is completed

end_of_pa_code <- c(2, 3, 5, 13:24)

d2018 %>% filter(pseq_length > 0,
                 EVENT_CD %in% end_of_pa_code) -> d2018a

# Find Brandon Belt's long PA in 2018

library(Lahman)
d2018a %>% filter(pseq_length == max(pseq_length)) %>% 
  inner_join(select(Master, nameFirst, nameLast, retroID),
             by = c("BAT_ID" = "retroID")) %>% 
  select(GAME_ID, nameFirst, nameLast, pseq, pseq_length, EVENT_CD)

# For each possible length of PA, find the number of PAs, the
# mean run value and the standard deviation of the run values

d2018a %>% group_by(pseq_length) %>% 
  summarize(N = n(),
            M = mean(RUNS.VALUE),
            S = sd(RUNS.VALUE)) -> S_count

# Plot of mean value of PA as function of number of pitches

ggplot(filter(S_count, pseq_length <= 9), 
       aes(pseq_length, M)) +
  geom_point(size = 2) +
  geom_hline(yintercept = 0, color = "red") +
  scale_x_continuous(breaks=1:9) +
  xlab("Number of Pitches") +
  ylab("Mean Run Value") +
  theme(plot.title = element_text(colour = "blue",
      size = 18, hjust = 0.5)) +
  ggtitle("Value of Plate Appearance\nas Function of Number of Pitches")

# Plot of standard deviation of value 

ggplot(filter(S_count, pseq_length <= 9), 
       aes(pseq_length, S)) +
  geom_point() +
  xlab("Number of Pitches") +
  ylab("SD Run Value") +
  ggtitle("Value of Plate Appearance as Function of Number of Pitches")

######### Summarize for each PA length and Count

d2018a %>% group_by(pseq_length, Count) %>% 
  summarize(N = n(),
            M = mean(RUNS.VALUE),
            S = sd(RUNS.VALUE)) %>% 
  filter(N >= 10) ->
  new_summ

## This adds the count info to the previous plot

ggplot(filter(new_summ, pseq_length <= 9),
       aes(pseq_length, M, label = Count)) +
  geom_label(color = "blue", size = 5) +
  geom_hline(yintercept = 0, color = "red") +
  scale_x_continuous(breaks=1:9) +
  xlab("Number of Pitches") +
  ylab("Mean Run Value") +
  theme(plot.title = element_text(colour = "blue",
                                  size = 18, hjust = 0.5)) +
  ggtitle("Value of Plate Appearance\nas Function of # of Pitches and Count")


# Find the HR rate for each length of PA and count

d2018a %>% 
  group_by(pseq_length, Count) %>% 
  summarize(N = n(),
            HR = sum(EVENT_CD == 23),
            HR_Rate = mean(EVENT_CD == 23)) -> HR

# graph the HR rates against PA length

ggplot(filter(HR, pseq_length <= 9, HR > 9),
       aes(pseq_length, HR_Rate, 
           label = Count)) +
  geom_label(color = "blue", size = 5) +
  scale_x_continuous(breaks=1:9) +
  xlab("Number of Pitches") +
  ylab("Home Run Rate") +
  theme(plot.title = element_text(colour = "blue",
                                  size = 18, hjust = 0.5)) +
  ggtitle("Home Run Rate\nas Function of Number of Pitches")