Run-Expectancy.R

#' Sample MiLB Run Expectancy
#'
#' How to calculate Run Expectancy
#' ___________________________________________________
#' 1) Get PBP
#' 2) Runs at current moment
#' 3) Get Runs at end of half inning
#' 4) Runs yet to be scored = runs at current moment - runs at end of 1/2 inning
#' 5) Get State of game (runners on and outs)
#' 
#' Ideally dataframe would be:
#' RUN ON 1ST | RUN ON 2ND | RUN ON 3RD | OUTS | RUNS YET TO BE SCORES
#' 
#' 6) Average by state

# Libraries
library(tidyverse)
library(baseballr)
library(progressr)

# level_ids key
#' 1 = MLB
#' 11 = Triple-A
#' 12 = Double-A
#' 13 = Class A Advanced
#' 14 = Class A
#' These are also technically possible but not worth my efforts tbh
#' 15 = Class A Short Season
#' 5442 = Rookie Advanced
#' 16 = Rookie
#' 17 = Winter League
#' 
level_ids_key <- data.frame(level_id = c(1,11,12,13,14),
                            league = c("MLB", "AAA", "AA", "A+", "A"))

# Create RE Function
create_re <- function(year_input = 2015, 
                      league_input = "AA", 
                      level_id_key = level_ids_key) {
  
  file_name <- paste0("Data/", league_input,"-", year_input, "-RE.csv")
  
  level <- level_ids_key %>% 
    filter(league == league_input) %>% 
    pull(level_id)
  
  ### Step 1: Get Data
  # Get All ids of a certain level
  schedule <- mlb_schedule(season = year_input,
                    level_ids = level)
  
  # Get some Game IDs from 5 games
  # Eventually, this will be all IDs
  ids <- unique(schedule$game_pk)
  
  # Help prevent errors
  daily_stats <- function(game_id){
    tryCatch(
      mlb_pbp(game_pk = game_id) # |> 
      # Data cleaning/manipulation
      ,
      error = function(cond){tibble()}
    )
  }
  
  # Get All pbp
  pbp <- map(ids, ~{ daily_stats(.x)}) %>%
    bind_rows()
  
  ### Steps 2-5) Data Cleaning
  run_expectancies <- pbp %>% 
    # Only last pitch of each at bat, to reduce size of data
    # and thus improve speed.
    group_by(game_pk, about.inning, about.halfInning, atBatIndex) %>%
    # Take last row of each at bat
    slice_max(order_by = index, n = 1) %>% 
    ### Step 2: Runs at Current Moment
    mutate(
      CURRENT.RUNS = case_when(
        # Bottom of Inning == Home Hits, so only home score changes
        about.halfInning == "bottom" ~ result.homeScore,
        # Top of Inning == Away Hits, so only away score changes
        about.halfInning == "top" ~ result.awayScore
      )) %>% 
    ungroup() %>% 
    ### Step 3: Get Runs at end of half inning
    group_by(game_pk, about.inning, about.halfInning) %>% # Not grouped at atbat level
    mutate(
      END.HALF.INNING.RUNS = case_when(
        # Bottom of Inning == Home Hits, so only home score changes
        about.halfInning == "bottom" ~ max(result.homeScore),
        # Top of Inning == Away Hits, so only away score changes
        about.halfInning == "top" ~ max(result.awayScore)
      )) %>% 
    ungroup() %>% 
    ### Step 4: Calculate Runs Yet to Be Scored in 1/2 inning
    mutate(RUNS.REMAINING = END.HALF.INNING.RUNS - CURRENT.RUNS, 
    ### Step 5: Get State of game (runners on and outs)
    RUNNER.ON.1 = case_when(is.na(matchup.postOnFirst.fullName) ~ 0,
                              TRUE ~ 1),
    RUNNER.ON.2 = case_when(is.na(matchup.postOnSecond.fullName) ~ 0,
                         TRUE ~ 1),
    RUNNER.ON.3 = case_when(is.na(matchup.postOnThird.fullName) ~ 0,
                         TRUE ~ 1),
    OUTS = count.outs.end) %>% 
    # Only want to select created columns
    select(starts_with("RUNNER.ON"), OUTS, RUNS.REMAINING) %>% 
    # Can't score any more runs if there are 3 outs
    filter(OUTS != 3) %>% 
  ### Step 6:  Average by state
  # aka Calculate Run Expectancies
    group_by(RUNNER.ON.1, RUNNER.ON.2, RUNNER.ON.3, OUTS) %>%
      summarize(AVG.RUNS = mean(RUNS.REMAINING, na.rm = TRUE)) %>% 
    ungroup() %>% 
    # Make a state variable by pasting runners on
    mutate(game_state = paste0(RUNNER.ON.1, RUNNER.ON.2, RUNNER.ON.3)) %>% 
    select(game_state, OUTS, AVG.RUNS) %>% 
    pivot_wider(names_from = OUTS,
                names_prefix = "outs_",
                values_from = AVG.RUNS)
    
  # For some reason, there was a typo somewhere in the pbp
  # So I had to get rid of a column where outs == 4
  run_expectancies <- run_expectancies %>% 
    select(-outs_4)
  
  write.csv(x = run_expectancies,
            file = file_name,
            row.names = FALSE)
}

# Get all combos of year and league
all_combos <- expand.grid(year = 2015:2021,
                          league = level_ids_key$league)

# Run this create_re function for all leagues back to 2015
create_re(2021, "AAA")