R function to compute expected runs for different bases/outs states from event table from statsapi.mlb.com

compute_expected_runs.R

compute_expected_runs <- function(events){
  # function computes expected runs for event table
  # obtained via statsapi.mlb.com
  
  require(dplyr)
  # find the total runs scored and number of outs each half-inning
  events |> 
    group_by(game_id, inning, half_inning) |> 
    summarize(max_runs = sum(runs_on_event),
              total_outs = sum(post_outs - pre_outs),
            .groups = "drop") -> TotalRuns

  # create bases, new_bases, state, new_state variables
  events |> 
     mutate(bases = paste(ifelse(is.na(pre_runner_1b_id), 0, 1),
                       ifelse(is.na(pre_runner_2b_id), 0, 1),
                 ifelse(is.na(pre_runner_3b_id), 0, 1),
                 sep = ""),
         new_bases = paste(ifelse(is.na(post_runner_1b_id), 0, 1),
                       ifelse(is.na(post_runner_2b_id), 0, 1),
                       ifelse(is.na(post_runner_3b_id), 0, 1),
                       sep = ""),
         state = paste(bases, pre_outs),
         new_state = paste(new_bases, post_outs)) -> events

  # for each event, find the runs scored in the half_inning
  # before or at that event
  events |> 
    group_by(game_id, inning, half_inning) |> 
    mutate(cum_runs = cumsum(runs_on_event)) |> 
    ungroup() -> events

  # merge dataset with total runs scored dataset
  # create runs_roi variable
  inner_join(events, TotalRuns, 
           by = c("game_id", "inning", "half_inning")) |> 
    mutate(runs_roi = 
             max_runs - cum_runs + runs_on_event) |> 
    rename(outs = pre_outs) -> events
  
  # only consider situations where total outs is 3 
  # and there is a change in states or runs scored
  events |> 
    filter(state != new_state | runs_on_event > 0) |> 
    filter(total_outs == 3) -> events

  # find runs expectancies
  events |> 
    group_by(bases, outs) |> 
    summarize(ExpectedRuns = mean(runs_roi),
            .groups = "drop") 
}