Prediction of 2nd Half Team Records

all_work.R

# illustration of producing graph for 2015 season
# I am assuming that the Retrosheet game log files are in the folder
# ~/Google Drive/gamelogs/gamelogs/
# output <- all_work(2015)
# output$p 

all_work <- function(season){
  require(readr)
  require(lubridate)
  require(dplyr)
  require(ggplot2)
  require(ggrepel)
  TH <- theme(
    plot.title = element_text(
      colour = "blue",
      size = 18,
      hjust = 0.5,
      vjust = 0.8,
      angle = 0
    )
  )
  file_name <- paste("~/Google Drive/gamelogs/gamelogs/gl",
                     season, ".txt", sep="")
  gldata <- read_csv(file_name, 
                   col_names = FALSE)
  headers <- read_csv("~/Google Drive/gamelogs/gamelogs/game_log_header.csv")
  names(gldata) <- names(headers)
  gldata$DATE <- ymd(gldata$Date)

  get_standings <- function(glyear){
    glyear <- mutate(glyear,
      Winner = ifelse(VisitorRunsScored < HomeRunsScore,
                      1, 0))
    S1 <- summarize(group_by(glyear, HomeTeam),
                W=sum(Winner), L=sum(1 - Winner))
    S2 <- summarize(group_by(glyear, VisitingTeam),
                W=sum(1 - Winner), L=sum(Winner))
    S <- inner_join(S1, S2,
                by=c("HomeTeam" = "VisitingTeam"))
    names(S)[1] <- "Team"
    F <- select(mutate(S, W=W.x + W.y,
               L=L.x + L.y), Team, W, L)
    Total <- sum(F$W + F$L)
    Total_season <- 162 * 30
    list(Standings = F, 
       Percentage = 100 * (Total / Total_season))
  }
  date1 <- paste(season, "-07-03", sep="")
  date2 <- paste(season, "-07-02", sep="")
  ff1 <- get_standings(filter(gldata,
                            DATE < date2))
  ff2 <- get_standings(filter(gldata,
                            DATE > date1))

  ff <- inner_join(ff1$Standings, ff2$Standings, by="Team")
  ff <- mutate(ff,
             WL_diff_1 = W.x - L.x,
             WL_diff_2 = W.y - L.y)

  p <- ggplot(ff, 
        aes(WL_diff_1, WL_diff_2 - WL_diff_1, label=Team)) +
   geom_text_repel() +
   geom_smooth(method="lm", se=FALSE) +
   geom_hline(yintercept = 0, color="red", linetype="dashed") +
   geom_vline(xintercept = 0, color="red", linetype="dashed") +
    xlab("W - L (First Half)") +
    ylab("W - L (Second) Minus W - L (First)") +
    ggtitle(paste(season, "Season, Slope = ", 
                  round(output$slope, 2))) + 
    TH
  fit <- lm(I(WL_diff_2 - WL_diff_1) ~ 0 + WL_diff_1, 
            data=ff)
  list(p=p, slope=fit$coef)
}