Tackle Value

library(dplyr)
library(ggplot2)

# Join Tackles data with Play data to get relevant information
tackles_with_play_info <- left_join(tackles_data, play_data, by = c("gameId", "playId"))

# Filter out plays where playNullifiedByPenalty is Y, and where penaltyYards is not NA
tackles_with_play_info <- tackles_with_play_info %>%
  filter((playNullifiedByPenalty != "Y"))

tackles_with_play_info <- tackles_with_play_info %>%
  filter((is.na(penaltyYards)) | (penaltyYards == 0))

tackles_with_play_info <- tackles_with_play_info %>%
  filter((pff_missedTackle != 1))

tackles_with_play_info <- tackles_with_play_info %>%
  filter((forcedFumble != 1))

tackles_with_play_info <- tackles_with_play_info %>%
  filter((playResult >= 0))

tackles_with_play_info <- tackles_with_play_info %>%
  filter((passResult < 20 & !(is.na(passResult))) | (is.na(passResult) & playResult < 10))

# Join with Tracking data to get additional player tracking information
tackles_with_tracking_info <- left_join(tackles_with_play_info, tracking_data, by = c("gameId", "playId", "nflId"))

# Join with Player data to get player positions
tackles_with_positions <- left_join(tackles_with_tracking_info, player_data %>% select(nflId, position), by = "nflId")

# Join with Game data to map home and visiting teams
tackles_with_team_info <- left_join(tackles_with_positions, game_data %>% select(gameId, homeTeamAbbr, visitorTeamAbbr), by = "gameId") %>%
  mutate(
    defensiveTeam = ifelse(possessionTeam == homeTeamAbbr, visitorTeamAbbr, homeTeamAbbr),
    win_prob_added = ifelse(possessionTeam == homeTeamAbbr, visitorTeamWinProbilityAdded, homeTeamWinProbabilityAdded)
  )

# Filter for ILB or MLB players
filtered_tackles <- tackles_with_team_info %>%
  filter(position %in% c("ILB", "MLB"))



# Check if 'dis' column exists in filtered_tackles
if (!"dis" %in% colnames(filtered_tackles)) {
  stop("Required variable 'dis' not found in filtered_tackles.")
}



# Calculate Nonexplosive Tackle Value for each play separately and remove duplicates
nonexplosive_Butkus_tackle_value <- filtered_tackles %>%
  group_by(gameId, playId, nflId, defensiveTeam, possessionTeam) %>%
  summarise(
    win_prob_added = last(win_prob_added),
    expected_points_added = last(expectedPointsAdded),
    yards_gained = last(playResult),
    tackle_range = sqrt((last(x) - first(x))^2 + (last(y) - first(y))^2),
    tackle_value = (0.15*win_prob_added) - (0.30*expected_points_added) - (0.40*yards_gained) + (0.15*tackle_range)
  ) 



# Normalize the nonexplosive_tackle_value column to a scale of -1 to 1
nonexplosive_Butkus_tackle_value$normalized_tackle_value <- scales::rescale(nonexplosive_Butkus_tackle_value$tackle_value, 
                                                                            to = c(-0.5, 0.5))

# View the results
head(nonexplosive_Butkus_tackle_value)


library(dplyr)


# Join Tackles data with Play data to get relevant information
explosive_tackles <- left_join(tackles_data, play_data, by = c("gameId", "playId"))

# Filter out plays where playNullifiedByPenalty is Y, and where penaltyYards is not NA
explosive_tackles <- explosive_tackles %>%
  filter((playNullifiedByPenalty != "Y"))

explosive_tackles <- explosive_tackles %>%
  filter((is.na(penaltyYards)) | (penaltyYards == 0))

explosive_tackles <- explosive_tackles %>%
  filter((pff_missedTackle != 1))

explosive_tackles <- explosive_tackles %>%
  filter((forcedFumble == 1) | (playResult < 0)) #plays where tackler forces a fumble or loss


# Join with Tracking data to get additional player tracking information
explosive_tackles_tracking <- left_join(explosive_tackles, tracking_data, by = c("gameId", "playId", "nflId"))

# Join with Player data to get player positions
explosive_tackles_players <- left_join(explosive_tackles_tracking, player_data %>% select(nflId, position), by = "nflId")

# Join with Game data to map home and visiting teams
explosive_tackles_teams <- left_join(explosive_tackles_players, game_data %>% select(gameId, homeTeamAbbr, visitorTeamAbbr), by = "gameId") %>%
  mutate(
    defensiveTeam = ifelse(possessionTeam == homeTeamAbbr, visitorTeamAbbr, homeTeamAbbr),
    win_prob_added = ifelse(possessionTeam == homeTeamAbbr, visitorTeamWinProbilityAdded, homeTeamWinProbabilityAdded)
  )

# Filter for ILB or MLB players
explosive_tackles_LB <- explosive_tackles_teams %>%
  filter(position %in% c("ILB", "MLB"))


# Check if 'dis' column exists in filtered_tackles
if (!"dis" %in% colnames(explosive_tackles_LB)) {
  stop("Required variable 'dis' not found in filtered_tackles.")
}

# Calculate Nonexplosive Tackle Value for each play separately and remove duplicates
explosive_Butkus_tackle_value <- explosive_tackles_LB %>%
  group_by(gameId, playId, nflId, defensiveTeam, possessionTeam) %>%
  summarise(
    win_prob_added = last(win_prob_added),
    expected_points_added = last(expectedPointsAdded),
    yards_gained = last(playResult),
    tackle_range = sqrt((last(x) - first(x))^2 + (last(y) - first(y))^2),
    tackle_value = (0.15*win_prob_added) - (0.30*expected_points_added) - (0.40*yards_gained) + (0.15*tackle_range)
  ) 

# Normalize the nonexplosive_tackle_value column to a scale of -1 to 1
explosive_Butkus_tackle_value$normalized_tackle_value <- scales::rescale(explosive_Butkus_tackle_value$tackle_value, 
                                                                         to = c(0.5, 1))
library(dplyr)


# Join Tackles data with Play data to get relevant information
offensive_explosive_plays <- left_join(tackles_data, play_data, by = c("gameId", "playId"))

# Filter out plays where playNullifiedByPenalty is Y, and where penaltyYards is not NA
offensive_explosive_plays <- offensive_explosive_plays %>%
  filter((playNullifiedByPenalty != "Y"))

offensive_explosive_plays <- offensive_explosive_plays %>%
  filter((is.na(penaltyYards)) | (penaltyYards == 0))

offensive_explosive_plays <- offensive_explosive_plays %>%
  filter((pff_missedTackle != 1))

offensive_explosive_plays <- offensive_explosive_plays %>%
  filter((forcedFumble != 1))

offensive_explosive_plays <- offensive_explosive_plays %>%
  filter((passLength >= 20) | (is.na(passLength)) & (playResult >= 10))


# Join with Tracking data to get additional player tracking information
offensive_explosive_players <- left_join(offensive_explosive_plays, tracking_data, by = c("gameId", "playId", "nflId"))

# Join with Player data to get player positions
offensive_explosive_positions <- left_join(offensive_explosive_players, player_data %>% select(nflId, position), by = "nflId")

# Join with Game data to map home and visiting teams
offensive_explosive_teams <- left_join(offensive_explosive_positions, game_data %>% select(gameId, homeTeamAbbr, visitorTeamAbbr), by = "gameId") %>%
  mutate(
    defensiveTeam = ifelse(possessionTeam == homeTeamAbbr, visitorTeamAbbr, homeTeamAbbr),
    win_prob_added = ifelse(possessionTeam == homeTeamAbbr, visitorTeamWinProbilityAdded, homeTeamWinProbabilityAdded)
  )

# Filter for ILB or MLB players
offensive_explosive_LBs <- offensive_explosive_teams %>%
  filter(position %in% c("ILB", "MLB"))


# Check if 'dis' column exists in filtered_tackles
if (!"dis" %in% colnames(offensive_explosive_LBs)) {
  stop("Required variable 'dis' not found in filtered_tackles.")
}

# Calculate Tackle Value for each play separately and remove duplicates where offense has explosive play
explosive_Butkus_tackle_value_bad <- offensive_explosive_LBs %>%
  group_by(gameId, playId, nflId, defensiveTeam, possessionTeam) %>%
  summarise(
    win_prob_added = last(win_prob_added),
    expected_points_added = last(expectedPointsAdded),
    yards_gained = last(playResult),
    tackle_range = sqrt((last(x) - first(x))^2 + (last(y) - first(y))^2),
    tackle_value = (0.15*win_prob_added) - (0.30*expected_points_added) - (0.40*yards_gained) + (0.15*tackle_range)
  ) 

# Normalize the nonexplosive_tackle_value column to a scale of -1 to 1
explosive_Butkus_tackle_value_bad$normalized_tackle_value <- scales::rescale(explosive_Butkus_tackle_value_bad$tackle_value, 
                                                                         to = c(-1, -0.5))

# View the results
head(explosive_Butkus_tackle_value_bad)



# Assuming you have the necessary dataframes loaded: tackle_value, play_data, tackles_data, and player_data

# Merge tackle_value with play_data on gameId and playId
#combination <- left_join(tackle_value, play_data, by = c("gameId", "playId"))

# Merge with tackles_data on gameId and playId
#combination <- left_join(combination, tackles_data, by = c("gameId", "playId"))

# Merge with player_data on nflId
#combination <- left_join(combination, player_data, by = "nflId")

#combination <- combination %>%
 # filter((pff_missedTackle != 1))

# View the combined dataset
#head(combination)


#Graph of average tackle value nonexplosive vs nonexplosive tackles frequency
library(ggplot2)
library(ggrepel)

# Filter players with 20 or more tackles from nonexplosive_Butkus_tackle_value
nonexplosive_tackle_by_player <- nonexplosive_Butkus_tackle_value %>%
  group_by(nflId) %>%
  summarise(
    avg_nonexplosive_tackle_value = mean(tackle_value),
    normalized_avg_nonexplosive_tackle_value = scales::rescale(mean(tackle_value)),
    total_attempts = n()
  ) %>%
  left_join(player_data, by = "nflId") %>%
  left_join(tackle_efficiency %>% filter(position %in% c("ILB", "MLB") & total_tackles_attempted >= 20), by = "nflId")
nonexplosive_tackle_by_player <- inner_join(nonexplosive_tackle_by_player, tackle_efficiency, by = "nflId")

# Join with existing nonexplosive_tackle_by_player (if it already exists)
if (exists("nonexplosive_tackle_by_player")) {
  nonexplosive_tackle_by_player <- left_join(nonexplosive_tackle_by_player, nonexplosive_tackle_by_player, by = c("nflId" = "nflId"))
}

# Calculate snap count for each player
snap_count_data <- tracking_data %>%
  filter(frameId == 1) %>%
  group_by(nflId, displayName) %>%
  summarise(snap_count = n())

# Join snap count data with nonexplosive_Butkus_tackle_value
nonexplosive_tackle_with_snap_count <- left_join(nonexplosive_Butkus_tackle_value, snap_count_data, by = "nflId")

# Calculate total nonexplosive tackles by player
nonexplosive_tackle_by_player <- nonexplosive_tackle_with_snap_count %>%
  group_by(nflId, displayName, snap_count) %>%
  summarise(
    avg_normalized_tackle_value = mean(normalized_tackle_value, na.rm = TRUE),
    total_nonexplosive_tackles = sum(!is.na(tackle_value & tackle_value >= 0)))

# Calculate tackle frequency
nonexplosive_tackle_by_player <- nonexplosive_tackle_by_player %>%
  mutate(tackle_frequency = total_nonexplosive_tackles / snap_count)

# Assuming there's a common identifier, adjust the key accordingly
common_key <- "nflId"
# Inner join nonexplosive_tackle_by_player with the top 50 players from tackle_efficiency
mapped_nonexplosive_tackle <- inner_join(
  nonexplosive_tackle_by_player,
  tackle_efficiency,
  by = common_key
)

# Calculate medians
median_x3 <- median(mapped_nonexplosive_tackle$tackle_frequency)
median_y3 <- median(mapped_nonexplosive_tackle$avg_normalized_tackle_value)

# Create a new variable for quadrant using medians
mapped_nonexplosive_tackle <- mapped_nonexplosive_tackle %>%
  mutate(quadrantC = case_when(
    tackle_frequency >= median_x3 & avg_normalized_tackle_value >= median_y3 ~ "Make a lot of valuable nonexplosive tackles",
    tackle_frequency < median_x3 & avg_normalized_tackle_value >= median_y3 ~ "Make little of valauable nonexplosive tackles",
    tackle_frequency >= median_x3 & avg_normalized_tackle_value < median_y3 ~ "Make a lot of less valuable nonexplosive tackles",
    tackle_frequency < median_x3 & avg_normalized_tackle_value < median_y3 ~ "Make little of less valuable nonexplosive tackles",
    TRUE ~ NA_character_
  ))

# Plot the graph with text repel
ggplot(mapped_nonexplosive_tackle, aes(x = tackle_frequency, y = avg_normalized_tackle_value, size = total_nonexplosive_tackles)) +
  geom_point(aes(color = quadrantC)) +
  geom_text_repel(aes(label = displayName.x), box.padding = 0.5, point.padding = 0.2, size = 2.5, max.overlaps = 1000) +
  labs(title = "Butkus Tackle Value, Nonexplosive Plays",
       x = "Tackle Frequency, nonexplosive plays",
       y = "Average Tackle Value, -0.5 to 0.5") +
  theme_minimal() +
  scale_x_continuous(limits = c(0.03, 0.15), breaks = seq(0.03, 0.15, 0.03)) +
  scale_y_continuous(limits = c(-0.2, 0), breaks = seq(-0.2, 0, 0.05)) +
  geom_vline(xintercept = median_x3, linetype = "solid", color = "purple") +
  geom_hline(yintercept = median_y3, linetype = "solid", color = "purple") +
  theme(legend.position = "none")


# Filter players with 20 or more tackles from explosive_Butkus_tackle_value
explosive_tackle_by_player <- explosive_Butkus_tackle_value %>%
  group_by(nflId) %>%
  summarise(
    avg_explosive_tackle_value = mean(tackle_value),
    normalized_avg_explosive_tackle_value = scales::rescale(mean(tackle_value)),
    total_attempts = n()
  ) %>%
  left_join(player_data, by = "nflId") %>%
  left_join(tackle_efficiency %>% filter(position %in% c("ILB", "MLB") & total_tackles_attempted >= 20), by = "nflId")

# Calculate snap count for each player
snap_count_data_explosive <- tracking_data %>%
  filter(frameId == 1) %>%
  group_by(nflId, displayName) %>%
  summarise(snap_count = n())

# Join snap count data with explosive_Butkus_tackle_value
explosive_tackle_with_snap_count <- left_join(explosive_Butkus_tackle_value, snap_count_data_explosive, by = "nflId")

# Calculate total explosive tackles by player
explosive_tackle_by_player <- explosive_tackle_with_snap_count %>%
  group_by(nflId, displayName, snap_count) %>%
  summarise(
    avg_normalized_tackle_value_exp = mean(normalized_tackle_value, na.rm = TRUE),
    total_explosive_tackles = sum(!is.na(tackle_value & tackle_value >= 0))) #TFL or FF

explosive_tackle_by_player <- explosive_tackle_by_player %>%
  mutate(tackle_frequency_exp = total_explosive_tackles / snap_count)

# Filter out rows with missing values
explosive_tackle_by_player <- explosive_tackle_by_player %>%
  filter(!is.na(avg_normalized_tackle_value_exp) & !is.na(total_explosive_tackles))

# Assuming there's a common identifier, adjust the key accordingly
common_key <- "nflId"
# Inner join explosive_tackle_by_player with the top 50 players from tackle_efficiency
mapped_explosive_tackle <- inner_join(
  explosive_tackle_by_player,
  tackle_efficiency,
  by = common_key
)


# Calculate medians
median_x2 <- median(mapped_explosive_tackle$tackle_frequency_exp, na.rm = TRUE)
median_y2 <- median(mapped_explosive_tackle$avg_normalized_tackle_value_exp, na.rm = TRUE)

# Create a new variable for quadrant using medians
mapped_explosive_tackle <- mapped_explosive_tackle %>%
  mutate(quadrantD = case_when(
    tackle_frequency_exp >= median_x2 & avg_normalized_tackle_value_exp >= median_y2 ~ "Many big tackles, more valuable tackles",
    tackle_frequency_exp < median_x2 & avg_normalized_tackle_value_exp >= median_y2 ~ "Less big tackles, more valuable tackles",
    tackle_frequency_exp >= median_x2 & avg_normalized_tackle_value_exp < median_y2 ~ "Many big tackles, less valuable tackles",
    tackle_frequency_exp < median_x2 & avg_normalized_tackle_value_exp < median_y2 ~ "Less big tackles, less valuable tackles",
    TRUE ~ NA_character_
  ))

# Add other layers and settings
ggplot(mapped_explosive_tackle, aes(x = tackle_frequency_exp, y = avg_normalized_tackle_value_exp, size = total_explosive_tackles)) +
  geom_point(aes(color = quadrantD)) +
  geom_text_repel(aes(label = displayName.x), box.padding = 0.5, point.padding = 0.2, size = 2.5, max.overlaps = 1000) +
  labs(title = "Butkus Tackle Value, Big Plays",
       x = "Explosive Tackle Frequency",
       y = "Average Tackle Value, scaled 0.5 to 1") +
  theme_minimal() +
  scale_x_continuous(limits = c(0, 0.035), breaks = seq(0, 0.035, 0.007)) +
  scale_y_continuous(limits = c(0.55, 0.70), breaks = seq(0.55, 0.70, 0.03)) +
  geom_vline(xintercept = median_x2, linetype = "solid", color = "green") +
  geom_hline(yintercept = median_y2, linetype = "solid", color = "green") +
  theme(legend.position = "none")


# Calculate snap count for each player
snap_count_data_offensive_explosive <- tracking_data %>%
  filter(frameId == 1) %>%
  group_by(nflId, displayName) %>%
  summarise(snap_count = n())

# Join snap count data with explosive_Butkus_tackle_value_bad
explosive_tackle_with_snap_count_bad <- left_join(explosive_Butkus_tackle_value_bad, snap_count_data_offensive_explosive, by = "nflId")

# Filter players with 20 or more tackles from explosive_Butkus_tackle_value_bad
explosive_tackle_by_player_bad <- explosive_Butkus_tackle_value_bad %>%
  group_by(nflId) %>%
  summarise(
    avg_normalized_tackle_value_bad = mean(normalized_tackle_value, na.rm = TRUE),
    total_explosive_tackles_bad = sum(!is.na(tackle_value & tackle_value >= 0))
  ) %>%
  left_join(player_data, by = "nflId") %>%
  left_join(tackle_efficiency %>% filter(position %in% c("ILB", "MLB") & total_tackles_attempted >= 20 & !(is.na(total_tackles_attempted))), by = "nflId")
explosive_tackle_by_player_bad <- inner_join(explosive_tackle_by_player_bad, tackle_efficiency, by = "nflId")

# Calculate tackle frequency using the correct dataset
explosive_tackle_by_player_bad <- explosive_tackle_by_player_bad %>%
  mutate(tackle_frequency_bad = total_explosive_tackles_bad / snap_count.y)

# Calculate medians
median_x1 <- median(explosive_tackle_by_player_bad$tackle_frequency_bad)
median_y1 <- median(explosive_tackle_by_player_bad$avg_normalized_tackle_value_bad)

# Create a new variable for quadrants
explosive_tackle_by_player_bad <- explosive_tackle_by_player_bad %>%
  mutate(quadrantE = case_when(
    tackle_frequency_bad >= median_x1 & avg_normalized_tackle_value_bad >= median_y1 ~ "High rate of tackles being on explosive plays, less explosive plays",
    tackle_frequency_bad < median_x1 & avg_normalized_tackle_value_bad >= median_y1 ~ "Low rate of tackles being on explosive  plays, less explosive plays",
    tackle_frequency_bad < median_x1 & avg_normalized_tackle_value_bad < median_y1 ~ "Low rate of tackles being on explosive plays, more explosive plays",
    tackle_frequency_bad >= median_x1 & avg_normalized_tackle_value_bad < median_y1 ~ "High rate of tackles being on explosive plays, more explosive plays",
    TRUE ~ "Undefined"
  ))

# Plot the graph with text repel
ggplot(explosive_tackle_by_player_bad, aes(x = tackle_frequency_bad, y = avg_normalized_tackle_value_bad, size = total_explosive_tackles_bad)) +
  geom_point(aes(color = quadrantE)) +
  geom_text_repel(aes(label = displayName), box.padding = 0.5, point.padding = 0.2, size = 2.5, max.overlaps = 1000) +
  labs(title = "Butkus Tackle Value, Offensive Explosive Plays",
       x = "Tackle Frequency",
       y = "Average Tackle Value, scaled -0.5 to -1") +
  theme_minimal() +
  scale_x_continuous(limits = c(0, 0.024), breaks = seq(0, 0.024, 0.003)) +
  scale_y_continuous(limits = c(-0.85, -0.55), breaks = seq(-0.85,-0.55, 0.05)) +
  geom_vline(xintercept = median_x1, linetype = "solid", color = "blue") +
  geom_hline(yintercept = median_y1, linetype = "solid", color = "blue") +
  theme(legend.position = "none")

##########