Record in One-run Games

So far this season the Seattle Mariners have won a historically high number of one-run games. As of this writing they are 21 - 9, and 1/2 game up on the Astros despite having a run-differential of about 100 runs less. This post looks at records in one-run games over the years.

Libraries

library(RPostgres)
library(DBI)
library(dplyr)
library(ggplot2)
library(magrittr) # for the in-place assignment, %<>%
library(ggrepel)

Data

I’m using the game data from retrosheet.

conn <- dbConnect(RPostgres::Postgres(), 
                  password=Sys.getenv("PSQL_PASS"), 
                  user=Sys.getenv("PSQL_USER"), 
                  port=Sys.getenv("PSQL_PORT"), 
                  dbname='retrosheet')
df1 = dbGetQuery(conn, "select * from game ")

Make two data frames, one where the home team is “the team” and away team is “the opponent”, and visa versa, then rbind them into a single data frame.

dfA = df1 %>% 
  select(game_id, year_id, team_id=away_team_id, opp_id=home_team_id, team_score=away_score_ct, opp_score=home_score_ct) 
dfH = df1 %>% 
  select(game_id, year_id, team_id=home_team_id, opp_id=away_team_id, team_score=home_score_ct, opp_score=away_score_ct)
dfX = dplyr::bind_rows(dfA, dfH)

Compute run differential and an indicator for one-run games.

games = dfX %>% mutate(sc_diff = team_score - opp_score, 
                       is_onerun = abs(sc_diff) == 1, 
                       is_win = team_score > opp_score)

records = games %>% 
  group_by(year_id, team_id) %>% 
  summarise(games=n(), 
            wins=sum(is_win), 
            one_run_games=sum(is_onerun), 
            one_run_wins=sum(is_onerun * is_win), 
            wpct=wins/games, 
            wpct_onerun=one_run_wins/one_run_games) %>% 
  ungroup()

Let’s see which teams had the most one-run wins

top12_wins1 = records %>% 
  arrange(-one_run_wins) %>% 
  head(12) 
  
top12_wins1 %>% knitr::kable()

runs_diff_plot = function(dfX, title='') {
  dfX %>% 
    dplyr::inner_join(games, by=c("team_id", "year_id")) %>% 
    mutate(name=paste(team_id, year_id)) %>% 
    ggplot(aes(x=sc_diff)) + 
    geom_histogram(binwidth = 1) +
    facet_wrap(~name, ncol=3) + 
    theme_minimal(base_size = 16) + 
    labs(x='score differential', 
         title=title)  
}

p = runs_diff_plot(top12_wins1, 'score differential - top 12 1-run win-pct teams')
print(p)

Highest win percentage in 1-run games

top12_wpct1 = records %>% 
  arrange(-wpct_onerun) %>% 
  head(12) 
  
top12_wpct1 %>% knitr::kable()

p = runs_diff_plot(top12_wpct1, 'score differential - top 12 1-run win-pct teams')
print(p)

record in non-one-run games

records %<>% mutate(
  non_onerun_wins = wins-one_run_wins,
  non_onerun_games = games-one_run_games,
  wpct_non_onerun = non_onerun_wins / non_onerun_games
  )

Top 12 in non-one-run games

top12_winX = records %>% 
  arrange(-non_onerun_wins) %>% 
  head(12)
top12_winX %>% knitr::kable()

p = runs_diff_plot(top12_winX, 'score differential - top 12 non-1-run win teams')
print(p)

top12_wpctX = records %>% 
  arrange(-wpct_non_onerun) %>% 
  head(12)
top12_wpctX %>% knitr::kable()

p = runs_diff_plot(top12_wpctX,  'score differential - top 12 non-1-run win-pct teams')
print(p)

Finally, compare winning percentage in 1-run games to winning percentage in non-one-run games

p = records %>% 
  ggplot(aes(x=wpct_non_onerun, y=wpct_onerun)) + 
  geom_point() + 
  theme_minimal(base_size = 16) + 
  geom_hline(yintercept = 0.5, size=0.5, color='steelblue') + 
  geom_vline(xintercept = 0.5, size=0.5, color='steelblue') + 
  labs(x="Win Pct. - Non 1-run", y="Win Pct. - 1-run")
print(p)

Label the ones that are on the edges

records_quads = records %>% 
  mutate(r2=(wpct_onerun - 0.5)**2 + (wpct_non_onerun - 0.5)**2, 
         quad=as.integer(wpct_non_onerun >= 0.5) + 2 * as.integer(wpct_onerun >= 0.5))

lab_df = records_quads %>% 
  group_by(quad) %>% 
  arrange(-r2) %>% 
  mutate(r2_rank=row_number()) %>% 
  filter(r2_rank <= 10) %>% 
  ungroup() %>% 
  mutate(name=paste(team_id, year_id))

p2 = p + 
  geom_text_repel(data = lab_df, aes(x=wpct_non_onerun, y=wpct_onerun, label=name))
print(p2)