Capstone - Complete Analytics System

library(tidyverse)
library(StatsBombR)

# ============================================
# LEAGUE COUNTER-ATTACK ANALYSIS SYSTEM
# ============================================

# Identify counter-attack sequences
identify_counter_attacks <- function(events) {
  events %>%
    filter(play_pattern.name == "From Counter") %>%
    mutate(
      sequence_id = paste(match_id, possession, sep = "_")
    )
}

# Calculate team counter-attack metrics
calculate_team_counter_metrics <- function(events, matches) {

  counter_events <- identify_counter_attacks(events)
  teams <- unique(c(matches$home_team.home_team_name, matches$away_team.away_team_name))
  n_matches <- matches %>% count(home_team.home_team_name) %>% pull(n) %>% max()

  # Offensive metrics
  offensive <- counter_events %>%
    group_by(team.name) %>%
    summarise(
      counter_sequences = n_distinct(sequence_id),
      counter_shots = sum(type.name == "Shot"),
      counter_xg = sum(shot.statsbomb_xg, na.rm = TRUE),
      counter_goals = sum(type.name == "Shot" & shot.outcome.name == "Goal", na.rm = TRUE),
      .groups = "drop"
    ) %>%
    mutate(
      counters_per_match = counter_sequences / n_matches,
      xg_per_counter = counter_xg / counter_sequences,
      counter_conversion = counter_goals / counter_shots * 100
    )

  # Defensive vulnerability (counters conceded)
  defensive <- counter_events %>%
    group_by(possession_team.name) %>%
    summarise(
      counters_conceded = n_distinct(sequence_id),
      shots_conceded = sum(type.name == "Shot"),
      xg_conceded = sum(shot.statsbomb_xg, na.rm = TRUE),
      goals_conceded = sum(type.name == "Shot" & shot.outcome.name == "Goal", na.rm = TRUE),
      .groups = "drop"
    ) %>%
    rename(team.name = possession_team.name) %>%
    mutate(
      conceded_per_match = counters_conceded / n_matches,
      xg_conceded_per_counter = xg_conceded / counters_conceded
    )

  # Combine offensive and defensive
  combined <- offensive %>%
    full_join(defensive, by = "team.name", suffix = c("_off", "_def")) %>%
    replace_na(list(
      counter_sequences = 0, counter_xg = 0,
      counters_conceded = 0, xg_conceded = 0
    )) %>%
    mutate(
      transition_balance = counter_xg - xg_conceded,
      transition_ratio = ifelse(xg_conceded > 0, counter_xg / xg_conceded, Inf),
      overall_effectiveness = (
        (counters_per_match * 20) +
        (xg_per_counter * 50) -
        (conceded_per_match * 15) -
        (xg_conceded_per_counter * 40)
      )
    ) %>%
    arrange(desc(overall_effectiveness))

  return(combined)
}

# Visualize counter-attack analysis
visualize_counter_analysis <- function(metrics) {

  # Scatter: Offensive vs Defensive
  p1 <- ggplot(metrics, aes(x = counter_xg, y = xg_conceded)) +
    geom_point(aes(size = counter_sequences, color = transition_balance), alpha = 0.7) +
    geom_text(aes(label = team.name), vjust = -0.8, size = 3, check_overlap = TRUE) +
    geom_abline(intercept = 0, slope = 1, linetype = "dashed", alpha = 0.5) +
    scale_color_gradient2(low = "red", mid = "gray", high = "green", midpoint = 0) +
    labs(
      title = "Counter-Attack Balance",
      subtitle = "Above line = vulnerable, Below line = effective",
      x = "Counter xG Generated", y = "Counter xG Conceded",
      color = "Balance", size = "Counters"
    ) +
    theme_minimal()

  # Rankings bar chart
  p2 <- ggplot(metrics %>% head(15),
               aes(x = reorder(team.name, overall_effectiveness), y = overall_effectiveness)) +
    geom_bar(stat = "identity", aes(fill = overall_effectiveness > 0)) +
    coord_flip() +
    scale_fill_manual(values = c("FALSE" = "#d73027", "TRUE" = "#1a9850"), guide = "none") +
    labs(
      title = "Transition Effectiveness Rankings",
      x = "", y = "Effectiveness Score"
    ) +
    theme_minimal()

  return(list(scatter = p1, rankings = p2))
}

# Generate comprehensive report
generate_counter_report <- function(metrics) {

  cat("\n", rep("=", 65), "\n", sep = "")
  cat("LEAGUE COUNTER-ATTACK ANALYSIS\n")
  cat(rep("=", 65), "\n\n", sep = "")

  cat("TOP 5 COUNTER-ATTACKING TEAMS (by xG generated):\n")
  cat("-", rep("-", 50), "\n", sep = "")
  metrics %>%
    arrange(desc(counter_xg)) %>%
    head(5) %>%
    select(team.name, counter_sequences, counter_xg, counter_goals, xg_per_counter) %>%
    print()

  cat("\n\nTOP 5 MOST VULNERABLE TEAMS (by xG conceded on counters):\n")
  cat("-", rep("-", 50), "\n", sep = "")
  metrics %>%
    arrange(desc(xg_conceded)) %>%
    head(5) %>%
    select(team.name, counters_conceded, xg_conceded, goals_conceded) %>%
    print()

  cat("\n\nBEST TRANSITION BALANCE:\n")
  cat("-", rep("-", 50), "\n", sep = "")
  metrics %>%
    arrange(desc(transition_balance)) %>%
    head(5) %>%
    select(team.name, counter_xg, xg_conceded, transition_balance) %>%
    print()
}

# Main execution
comps <- FreeCompetitions()
matches <- FreeMatches(Competitions = comps %>% filter(competition_id == 11))
events <- free_allevents(MatchesDF = matches, Atea = TRUE)
events <- allclean(events)

metrics <- calculate_team_counter_metrics(events, matches)
generate_counter_report(metrics)
plots <- visualize_counter_analysis(metrics)
print(plots$scatter)
print(plots$rankings)

library(tidyverse)
library(StatsBombR)
library(viridis)

# ============================================
# OPTIMAL PRESSING ZONE ANALYSIS
# ============================================

# Identify ball recoveries and subsequent outcomes
analyze_recovery_outcomes <- function(events, team_name, grid_size = 20) {

  team_events <- events %>% filter(team.name == team_name)

  # Ball recovery events
  recoveries <- team_events %>%
    filter(
      type.name %in% c("Ball Recovery", "Interception", "Tackle") |
      (type.name == "Duel" & duel.outcome.name %in% c("Won", "Success"))
    ) %>%
    filter(!is.na(location.x), !is.na(location.y)) %>%
    mutate(
      recovery_index = index,
      recovery_match = match_id,
      recovery_x = location.x,
      recovery_y = location.y,
      grid_x = floor(location.x / grid_size) * grid_size,
      grid_y = floor(location.y / grid_size) * grid_size,
      recovery_zone = case_when(
        location.x > 80 ~ "High",
        location.x > 40 ~ "Mid",
        TRUE ~ "Low"
      )
    )

  # Track what happens in the 10 seconds after recovery
  recovery_outcomes <- map_dfr(1:nrow(recoveries), function(i) {
    recovery <- recoveries[i, ]

    # Get subsequent events (within 10 seconds / ~30 events)
    subsequent <- events %>%
      filter(
        match_id == recovery$recovery_match,
        index > recovery$recovery_index,
        index <= recovery$recovery_index + 30,
        team.name == team_name
      )

    # Check outcomes
    tibble(
      grid_x = recovery$grid_x,
      grid_y = recovery$grid_y,
      recovery_zone = recovery$recovery_zone,
      recovery_x = recovery$recovery_x,
      recovery_y = recovery$recovery_y,
      led_to_shot = any(subsequent$type.name == "Shot"),
      shot_xg = sum(subsequent$shot.statsbomb_xg, na.rm = TRUE),
      led_to_goal = any(subsequent$type.name == "Shot" &
                        subsequent$shot.outcome.name == "Goal", na.rm = TRUE),
      sequence_length = nrow(subsequent),
      max_progression = ifelse(nrow(subsequent) > 0,
                               max(subsequent$location.x, na.rm = TRUE) - recovery$recovery_x,
                               0)
    )
  })

  return(recovery_outcomes)
}

# Calculate zone effectiveness
calculate_zone_effectiveness <- function(recovery_outcomes) {

  zone_stats <- recovery_outcomes %>%
    group_by(grid_x, grid_y) %>%
    summarise(
      n_recoveries = n(),
      shots_generated = sum(led_to_shot),
      total_xg = sum(shot_xg),
      goals = sum(led_to_goal),
      shot_rate = mean(led_to_shot) * 100,
      avg_xg_per_recovery = mean(shot_xg),
      avg_progression = mean(max_progression),
      .groups = "drop"
    ) %>%
    filter(n_recoveries >= 5) %>%  # Minimum sample
    mutate(
      effectiveness_score = (shot_rate * 0.3) +
                            (avg_xg_per_recovery * 100 * 0.4) +
                            (avg_progression * 0.3)
    ) %>%
    arrange(desc(effectiveness_score))

  # Zone-level summary
  zone_summary <- recovery_outcomes %>%
    group_by(recovery_zone) %>%
    summarise(
      n_recoveries = n(),
      shot_rate = mean(led_to_shot) * 100,
      avg_xg = mean(shot_xg),
      goal_rate = mean(led_to_goal) * 100,
      .groups = "drop"
    )

  return(list(grid = zone_stats, zones = zone_summary))
}

# Visualize pressing effectiveness
visualize_pressing_zones <- function(zone_stats, team_name) {

  ggplot(zone_stats, aes(x = grid_x + 10, y = grid_y + 10)) +
    # Pitch background
    annotate("rect", xmin = 0, xmax = 120, ymin = 0, ymax = 80,
             fill = "#228B22", alpha = 0.3) +
    # Zone lines
    annotate("segment", x = 40, xend = 40, y = 0, yend = 80,
             color = "white", linetype = "dashed") +
    annotate("segment", x = 80, xend = 80, y = 0, yend = 80,
             color = "white", linetype = "dashed") +
    # Effectiveness tiles
    geom_tile(aes(fill = effectiveness_score), alpha = 0.8) +
    geom_text(aes(label = paste0(round(shot_rate, 0), "%")),
              color = "white", size = 3) +
    scale_fill_viridis(option = "plasma", name = "Effectiveness") +
    labs(
      title = paste(team_name, "- Pressing Zone Effectiveness"),
      subtitle = "Numbers show shot generation rate from recoveries",
      x = "Pitch Length", y = "Pitch Width"
    ) +
    coord_fixed(xlim = c(0, 120), ylim = c(0, 80)) +
    theme_void() +
    theme(
      plot.title = element_text(hjust = 0.5, face = "bold"),
      legend.position = "right"
    )
}

# Generate tactical recommendations
generate_pressing_recommendations <- function(zone_stats, zone_summary) {

  cat("\n", rep("=", 65), "\n", sep = "")
  cat("PRESSING ZONE TACTICAL ANALYSIS\n")
  cat(rep("=", 65), "\n\n", sep = "")

  cat("ZONE-LEVEL SUMMARY:\n")
  cat("-", rep("-", 50), "\n", sep = "")
  print(zone_summary)

  cat("\n\nTOP 5 PRESSING ZONES (by effectiveness):\n")
  cat("-", rep("-", 50), "\n", sep = "")
  zone_stats %>%
    head(5) %>%
    select(grid_x, grid_y, n_recoveries, shot_rate, avg_xg_per_recovery, effectiveness_score) %>%
    print()

  cat("\n\nTACTICAL RECOMMENDATIONS:\n")
  cat("-", rep("-", 50), "\n", sep = "")

  # Find best zone
  best_zone <- zone_stats %>% slice(1)

  if (best_zone$grid_x > 80) {
    cat("  1. HIGH PRESS recommended - best returns from pressing in final third\n")
  } else if (best_zone$grid_x > 40) {
    cat("  1. MID-BLOCK with pressing triggers - best returns from middle third\n")
  } else {
    cat("  1. LOW BLOCK counter-attack style - winning ball deep creates danger\n")
  }

  # Check wide vs central
  wide_effectiveness <- zone_stats %>%
    filter(grid_y < 30 | grid_y > 50) %>%
    pull(effectiveness_score) %>%
    mean(na.rm = TRUE)

  central_effectiveness <- zone_stats %>%
    filter(grid_y >= 30 & grid_y <= 50) %>%
    pull(effectiveness_score) %>%
    mean(na.rm = TRUE)

  if (wide_effectiveness > central_effectiveness) {
    cat("  2. Focus pressing on WIDE areas - better transition opportunities\n")
  } else {
    cat("  2. Focus pressing CENTRALLY - better transition opportunities\n")
  }
}

# Main execution
events <- get.matchFree(match_id)
team_name <- "Liverpool"

recovery_outcomes <- analyze_recovery_outcomes(events, team_name, grid_size = 20)
zone_effectiveness <- calculate_zone_effectiveness(recovery_outcomes)

print(visualize_pressing_zones(zone_effectiveness$grid, team_name))
generate_pressing_recommendations(zone_effectiveness$grid, zone_effectiveness$zones)

library(tidyverse)
library(StatsBombR)
library(fmsb)

# ============================================
# PLAYER TRANSITION PROFILE SYSTEM
# ============================================

calculate_player_transition_profiles <- function(events, team_name) {

  team_events <- events %>% filter(team.name == team_name)

  # 1. Ball Winning
  ball_winning <- team_events %>%
    filter(type.name %in% c("Ball Recovery", "Interception", "Tackle", "Duel")) %>%
    mutate(
      won = case_when(
        type.name %in% c("Ball Recovery", "Interception") ~ TRUE,
        type.name == "Tackle" & tackle.outcome.name %in% c("Won", "Success") ~ TRUE,
        type.name == "Duel" & duel.outcome.name %in% c("Won", "Success") ~ TRUE,
        TRUE ~ FALSE
      ),
      high_recovery = !is.na(location.x) & location.x > 60
    ) %>%
    filter(won) %>%
    group_by(player.name, position.name) %>%
    summarise(
      ball_wins = n(),
      high_ball_wins = sum(high_recovery),
      .groups = "drop"
    )

  # 2. Transition Carrying
  carrying <- team_events %>%
    filter(type.name %in% c("Carry", "Dribble")) %>%
    mutate(
      end_x = coalesce(carry.end_location.x, location.x),
      progression = end_x - location.x,
      is_progressive = progression > 10
    ) %>%
    group_by(player.name, position.name) %>%
    summarise(
      total_carries = n(),
      progressive_carries = sum(is_progressive, na.rm = TRUE),
      total_progression = sum(progression[is_progressive], na.rm = TRUE),
      avg_carry_distance = mean(progression[is_progressive], na.rm = TRUE),
      .groups = "drop"
    )

  # 3. Transition Passing
  passing <- team_events %>%
    filter(type.name == "Pass") %>%
    mutate(
      is_successful = is.na(pass.outcome.name),
      progression = pass.end_location.x - location.x,
      is_progressive = progression > 10 & is_successful,
      is_through_ball = pass.technique.name == "Through Ball",
      is_key_pass = pass.shot_assist == TRUE | pass.goal_assist == TRUE
    ) %>%
    group_by(player.name, position.name) %>%
    summarise(
      total_passes = n(),
      progressive_passes = sum(is_progressive, na.rm = TRUE),
      through_balls = sum(is_through_ball & is_successful, na.rm = TRUE),
      key_passes = sum(is_key_pass, na.rm = TRUE),
      .groups = "drop"
    )

  # 4. Transition Finishing
  finishing <- team_events %>%
    filter(
      type.name == "Shot",
      play_pattern.name %in% c("From Counter", "From Throw In", "Regular Play")
    ) %>%
    group_by(player.name, position.name) %>%
    summarise(
      shots = n(),
      xg = sum(shot.statsbomb_xg, na.rm = TRUE),
      goals = sum(shot.outcome.name == "Goal"),
      .groups = "drop"
    )

  # Combine all metrics
  profiles <- ball_winning %>%
    full_join(carrying, by = c("player.name", "position.name")) %>%
    full_join(passing, by = c("player.name", "position.name")) %>%
    full_join(finishing, by = c("player.name", "position.name")) %>%
    replace_na(list(
      ball_wins = 0, high_ball_wins = 0,
      progressive_carries = 0, avg_carry_distance = 0,
      progressive_passes = 0, through_balls = 0, key_passes = 0,
      shots = 0, xg = 0, goals = 0
    )) %>%
    mutate(
      # Composite scores
      ball_winning_score = (ball_wins + high_ball_wins * 0.5),
      carrying_score = (progressive_carries * 2 + avg_carry_distance * 0.1),
      passing_score = (progressive_passes + through_balls * 3 + key_passes * 2),
      finishing_score = (shots + xg * 10 + goals * 5),

      # Total contribution
      transition_contribution = ball_winning_score + carrying_score +
                                passing_score + finishing_score,

      # Dominant role
      dominant_role = case_when(
        ball_winning_score >= pmax(carrying_score, passing_score, finishing_score) ~ "Ball Winner",
        carrying_score >= pmax(ball_winning_score, passing_score, finishing_score) ~ "Carrier",
        passing_score >= pmax(ball_winning_score, carrying_score, finishing_score) ~ "Passer",
        TRUE ~ "Finisher"
      )
    ) %>%
    arrange(desc(transition_contribution))

  return(profiles)
}

# Normalize for radar charts
normalize_for_radar <- function(profiles) {

  profiles %>%
    mutate(
      across(
        c(ball_winning_score, carrying_score, passing_score, finishing_score),
        ~ (. - min(.)) / (max(.) - min(.)) * 100
      )
    )
}

# Create radar chart comparison
create_player_radar <- function(profiles, player_names) {

  radar_data <- profiles %>%
    filter(player.name %in% player_names) %>%
    select(player.name, ball_winning_score, carrying_score,
           passing_score, finishing_score)

  # Prepare for fmsb
  radar_matrix <- radar_data %>%
    select(-player.name) %>%
    as.data.frame()

  rownames(radar_matrix) <- radar_data$player.name
  colnames(radar_matrix) <- c("Ball Winning", "Carrying", "Passing", "Finishing")

  # Add max/min rows
  radar_matrix <- rbind(
    rep(100, 4),
    rep(0, 4),
    radar_matrix
  )

  # Plot
  colors <- c("#E41A1C", "#377EB8", "#4DAF4A", "#984EA3")[1:length(player_names)]

  par(mar = c(1, 1, 2, 1))
  radarchart(
    radar_matrix,
    pcol = colors,
    pfcol = alpha(colors, 0.3),
    plwd = 2,
    plty = 1,
    cglcol = "grey",
    cglty = 1,
    vlcex = 0.9,
    title = "Transition Profile Comparison"
  )

  legend("bottomright",
         legend = player_names,
         col = colors,
         lty = 1,
         lwd = 2,
         bty = "n")
}

# Generate profile report
generate_profile_report <- function(profiles) {

  cat("\n", rep("=", 70), "\n", sep = "")
  cat("PLAYER TRANSITION PROFILE ANALYSIS\n")
  cat(rep("=", 70), "\n\n", sep = "")

  cat("TOP 10 OVERALL TRANSITION CONTRIBUTORS:\n")
  cat("-", rep("-", 55), "\n", sep = "")
  profiles %>%
    head(10) %>%
    select(player.name, dominant_role, transition_contribution,
           ball_winning_score, carrying_score, passing_score, finishing_score) %>%
    print()

  cat("\n\nTOP PLAYERS BY ROLE:\n")
  cat("-", rep("-", 55), "\n", sep = "")

  for (role in c("Ball Winner", "Carrier", "Passer", "Finisher")) {
    cat("\n", role, "s:\n", sep = "")
    profiles %>%
      filter(dominant_role == role) %>%
      arrange(desc(transition_contribution)) %>%
      head(3) %>%
      select(player.name, transition_contribution) %>%
      print()
  }
}

# Main execution
events <- allclean(get.matchFree(match_id))
profiles <- calculate_player_transition_profiles(events, "Liverpool")
profiles_normalized <- normalize_for_radar(profiles)

generate_profile_report(profiles)
create_player_radar(profiles_normalized, head(profiles$player.name, 4))