Exploring game durations

duration_study.R

# load some packages

library(dplyr)
library(Lahman)
library(ggplot2)
library(readr)

# add a theme for the ggplot title 

TH <- theme(plot.title = element_text(colour = "blue", 
                                      size = 14, 
              hjust = 0.5, vjust = 0.8, angle = 0))

# collect gamelog data for a particular season

get_gldata <- function(season){
  file_location <- paste("~/Google Drive/gamelogs/gamelogs/gl",
                         season, ".txt", sep="")
  gldata <- read_csv(file_location, 
                     col_names = FALSE)
  headers <- read_csv("~/Google Drive/gamelogs/gamelogs/game_log_header.csv")
  names(gldata) <- names(headers)

  data.frame(Season = season, gldata)
}

# collect data for Halladay seasons

seasons <- 1999:2013
ALL_GAMES <- NULL
for(j in 1:15){
  ALL_GAMES <- rbind(ALL_GAMES, get_gldata(seasons[j]))
}

# create new dataframe with starter ids and duration

ALL_GAMES %>% select(Season, Date, 
          HomeStartingPitcherID, Duration) %>%
     mutate(Game = "Home") -> start1
names(start1)[3] <- "StarterID"
ALL_GAMES %>% select(Season, Date, 
             VisitorStartingPitcherID, Duration) %>%
  mutate(Game = "Away") -> start2
names(start2)[3] <- "StarterID"
start <- rbind(start1, start2)

# compare durations for different seasons using boxplots

ggplot(start, aes(as.character(Season), Duration)) +
  geom_boxplot() + coord_flip() + xlab("Season") + TH +
  ggtitle("Boxplots of Game Durations for 15 Seasons") +
  ylab("Duration (Minutes)")

# create an indicator variable for Halladay

start %>% mutate(Halladay = ifelse(StarterID == "hallr001",
"yes", "no")) -> start

# compare game durations with and without Halladay

ggplot(start, aes(as.character(Season), 
                  Duration, color=Halladay)) +
  geom_boxplot() + coord_flip() + xlab("Season") + TH +
  ylab("Duration (Minutes)") +
  ggtitle("Game Durations With and Without Halladay")

# find the 10 best (shortest) duration starters for
# each of the 15 seasons

start %>% group_by(Season, StarterID) %>%
  summarize(N = n(), M= median(Duration)) %>%
  filter(N >= 19) %>% arrange(Season, M) %>% slice(1:10) ->
  toplist2

# Find the starters who have appeared at least three
# times on the top-ten list

toplist2 %>% group_by(StarterID) %>% summarize(N=n()) ->
  Names
Names2 <- filter(Names, N >= 3)
Names2 %>% inner_join(dplyr::select(Master, nameFirst, nameLast,
            retroID), by=c("StarterID" = "retroID")) %>%
        mutate(Name = paste(nameFirst, nameLast)) %>%
        dplyr::select(Name, N) -> Names3

ggplot(Names3, aes(reorder(Name, N), N)) + geom_point() + 
  coord_flip()  + xlab("Starter") +
  ggtitle("Number of Top-10 Appearances by Game Duration") +
  TH + scale_y_continuous(limits = c(2, 10)) +
  ylab("Number of Top-10 Appearances")

# Compare durations of Greg Maddux and Roy Halladay

start %>% inner_join(dplyr::select(Master, 
                          nameFirst, nameLast,
           retroID), by=c("StarterID" = "retroID")) %>%
  mutate(Name = paste(nameFirst, nameLast)) -> start
ggplot(filter(start, StarterID == "hallr001" |
                     StarterID == "maddg002",
                     Season <= 2008),
      aes(as.character(Season), Duration, color=Name)) +
  geom_boxplot() + coord_flip() + TH + 
  ggtitle("Maddux vs. Halladay, 1999-2008") +
  xlab("Season") + ylab("Duration (Minutes)")

# How has the median game duration changed over the
# last 50 seasons 1967-2016?

median_duration <- function(season){
  file_location <- paste("~/Google Drive/gamelogs/gamelogs/gl",
                         season, ".txt", sep="")
  gldata <- read_csv(file_location, 
                     col_names = FALSE)
  headers <- read_csv("~/Google Drive/gamelogs/gamelogs/game_log_header.csv")
  names(gldata) <- names(headers)
  
  median(gldata$Duration, na.rm=TRUE)
}

seasons <- 1967:2016
medians <- sapply(seasons, median_duration)
ggplot(data.frame(Season = seasons,
                  Median = medians),
       aes(Season, Median)) + TH +
  geom_point() + geom_smooth(span=0.3, color="red") +
  ggtitle("Median Game Duration:  1967:2016") +
  geom_hline(yintercept = 180, color="blue", 
             linetype="dashed") +
  geom_hline(yintercept = 150, color="blue",
             linetype="dashed") +
  annotate(geom="text", x=2010, y=152, 
           label="2 1/2 hours", color="blue") +
  annotate(geom="text", x=1975, y=182, 
           label="3 hours", color="blue")