assist-bias.R

## Scorekeeper bias in LEB Oro

library(tidyverse)

data_file <- "https://github.com/solmos/feb-data/raw/master/assists-leb-oro.csv"
assists_leb_oro <- read_csv(data_file)

# Players -----------------------------------------------------------------

ast_player <- assists_leb_oro %>%
  group_by(season, team, player, home) %>%
  summarise(
    ast_total = sum(ast),
    ast_avg = mean(ast)
  ) %>%
  ungroup()

# Players per season with more than 100 total assists in one season
ast_player_dif <- ast_player %>%
  mutate(home = ifelse(home, "home", "away")) %>%
  pivot_wider(
    id_cols = c(season, team, player),
    names_from = home,
    values_from = c(ast_total, ast_avg)
  ) %>%
  mutate(
    ast_total = ast_total_home + ast_total_away,
    ast_total_dif = ast_total_home - ast_total_away,
    ast_total_ratio = ast_total_home / ast_total_away
  ) %>%
  filter(ast_total > 100)

# Percentage of players with more assists at home
sum(ast_player_dif$ast_total_dif > 0) / nrow(ast_player_dif)

# Ratio
players_table <- ast_player_dif %>%
  arrange(desc(ast_total_ratio)) %>% 
  mutate(
    last_name = str_extract(player, "[A-ZÀ-ÿ]+"),
    first_name = str_remove(str_extract(player, ", [A-ZÀ-ÿ]+"), ", "),
    player_name = paste(first_name, last_name)
  ) %>% 
  select(
    season, player_name, team,
    ast_total, ast_total_home, ast_total_away,
    ast_total_ratio
  )
players_table


# Team --------------------------------------------------------------------

ast_team <- assists_leb_oro %>%
  group_by(game_code, team) %>%
  summarise(
    season = unique(season),
    home = unique(home),
    m_fg = sum(m_fg),
    ast = sum(ast)
  ) %>%
  ungroup() %>%
  mutate(fg_ast_pct = ast / m_fg)

ast_team

ast_diff <- ast_team %>%
  group_by(season, team, home) %>%
  summarise(
    fg_ast_pct = mean(fg_ast_pct)
  ) %>%
  ungroup() %>%
  pivot_wider(names_from = home, values_from = fg_ast_pct) %>%
  rename(away = `FALSE`, home = `TRUE`) %>%
  mutate(d = home - away)

ast_diff %>%
  group_by(season) %>%
  summarise(d = mean(d) * 100) %>%
  ggplot(aes(season, d)) +
  geom_point() +
  geom_line() +
  ylim(c(0, 13))