Possession adjust metrics for StatsBomb data

possessionAdjust.R

# Compute total minutes played per player -------------------------------

statsbombMinsPlayed <- function(dat) {
  
  dat <- dat %>% 
    group_by(match_id) %>% 
    mutate(end = max(ElapsedTime))
  
  # starting XI
  minutes <- dat %>% 
    group_by(match_id, player.name) %>%  
    filter(row_number() == 1) %>%
    select(team.name, player.name, ElapsedTime, end) %>% 
    na.omit() %>% 
    group_by(match_id) %>% 
    top_n(22, -ElapsedTime) %>% 
    mutate(start = 0) %>% 
    select(-ElapsedTime)
  
  # substitutes
  minutes <- rbind(
    minutes,
    dat %>% 
      filter(type.name == "Substitution") %>% 
      select(match_id, team.name, player.name = substitution.replacement.name, end, start = ElapsedTime)
  )
  
  minutes <- left_join(
    minutes,
    dat %>% 
      filter(type.name == "Substitution") %>% 
      select(match_id, team.name, player.name, end = ElapsedTime),
    by = c("match_id", "team.name", "player.name")
  )
  
  minutes <- minutes %>% 
    mutate(end = if_else(end.y < end.x, end.y, end.x),
           end = if_else(is.na(end.y), end.x, end.y)) %>%
    mutate(mins = (end - start) / 60) %>%
    select(-end.x, -end.y)
  
  return(minutes)

}


# Get opposition time in possession for each player -------------------------------

d <- readRDS("StatsBomb_data.Rdata")

# minutes played per player per match
minutes <- statsbombMinsPlayed(d)

# cumulative time in possession by time elapsed per team
poss <- d %>% 
  group_by(match_id, possession) %>%
  summarise(ElapsedTime = max(ElapsedTime), TimeInPoss = max(TimeInPoss), team.name = possession_team.name[1]) %>%
  ungroup %>% 
  mutate(CumTimeInPoss = cumsum(TimeInPoss))

# cumulative opponent time in possession per player's playing time per match
OppPoss <- lapply(1:nrow(minutes), function(i) {
  
  ss <- minutes[i,]
  
  OppPoss <- poss %>% 
    filter(match_id == ss$match_id, ElapsedTime >= ss$start & ElapsedTime <= ss$end & team.name != ss$team.name) %>% 
    summarise(OppPoss = max(CumTimeInPoss) - min(CumTimeInPoss))
  
  ss %>% 
    select(-start, -end) %>% 
    mutate(OppPoss = OppPoss$OppPoss)

}) %>% 
  plyr::rbind.fill()

# summarise opponent playing time per player
OppPoss <- OppPoss %>% 
  group_by(player.name) %>% 
  summarise(OppPoss = sum(OppPoss)) %>% 
  filter(OppPoss > 0)


# e.g. Padj tackles and interceptions -----------------------------------------

# all tackles and interceptions per player
tackint <- d %>% 
  filter((type.name == "Duel" & duel.type.name == "Tackle" & duel.outcome.name %in% c("Success In Play", "Success Out", "Won")) |
         (type.name == "Interception" & interception.outcome.name %in% c("Success In Play", "Success Out", "Won"))) %>% 
  group_by(team.name, player.name) %>%
  summarise(n = n())
  
# padj, here per 30 minutes opponent possession (around the average duration of any team's possession per match)
ti30 <- left_join(tackint, 
          OppPoss %>% select(player.name, OppPoss),
          by = "player.name") %>% 
  mutate(ti30 = n / OppPoss * 60 * 30) %>% 
  arrange(ti30)