Exploring NCAA Postseason Awards

Author

Kyle Bresko, Charles Lane, Kaleb Treangen 

library(tidyverse)
library(ggplot2)
library(plotly)
library(DT)
library(leaflet)
library(qrcode)

Introduction

In recent years, NCAA athletes have had more freedom to transfer schools to play sports thanks to NIL (Name, Image and Likeness) without any restrictions or repercussions. This change to the landscape of college sports has transformed college athletics and has resulted in team rosters having a large increase in turnover rates. To put this in perspective, there are some reports that show that over 50% of all college athletes who enter the portal may end up not even finding a new school. That is just how many athletes are entering the portal. The only real restriction for all sports is the ‘transfer window’ which is the only time an athlete can transfer and is currently only open for 15 days. This makes it important for coaches to know which athletes to target before the window opens and closes. While season-ending awards can signal high performance, these recognitions occur after the season. With this in mind, our goal is to model high performers using previous season statistics to be able to forecast awards won by current season players.

Data

Data Sources

Load Data

schools <- tibble(
  School = c(
    "FIU", "Jacksonville St", "LA Tech", "Liberty", "Mid Tennessee",
    "New Mexico St", "Sam Houston", "UTEP", "W Kentucky",
    "Kennesaw State", "Delaware", "Missouri State"
  ),
  `23-24` = c(
    TRUE, TRUE, TRUE, TRUE, TRUE,
    TRUE, TRUE, TRUE, TRUE,
    FALSE, FALSE, FALSE
  ),
  `24-25` = c(
    TRUE, TRUE, TRUE, TRUE, TRUE,
    TRUE, TRUE, TRUE, TRUE,
    TRUE, FALSE, FALSE
  ),
  `25-26` = c(
    TRUE, TRUE, TRUE, TRUE, TRUE,
    TRUE, TRUE, TRUE, TRUE,
    TRUE, TRUE, TRUE
  )
)

import_season <- function(season, schools) {
  active_schools <- schools %>% filter(.data[[season]]) %>% pull(School)
  boxscores_list <- map(active_schools, function(school) {
    path <- str_glue("../data/boxscores/{school}_20{season}.csv")
    df <- read_csv(path, show_col_types = FALSE)
    message(str_glue('Season {season} for "{school}" imported.'))
    df
  })
  bind_rows(boxscores_list)
}

boxscores_24 <- import_season("23-24", schools)
boxscores_25 <- import_season("24-25", schools)
boxscores_26 <- import_season("25-26", schools)

awards_24 <- read_csv("../data/awards/conferenceusa_awards_2023-24.csv", show_col_types = FALSE)
awards_25 <- read_csv("../data/awards/conferenceusa_awards_2024-25.csv", show_col_types = FALSE)
awards_26 <- read_csv("../data/awards/conferenceusa_awards_2025-26.csv", show_col_types = FALSE)

Data Summary

for (info in list(
  list(df = boxscores_24, label = "24"),
  list(df = boxscores_25, label = "25"),
  list(df = boxscores_26, label = "26")
)) {
  cat(str_glue("---------- Boxscores '{info$label} ----------"), "\n")
  cat(str_glue("Features:\t{ncol(info$df)}"), "\n")
  cat(str_glue("Observations:\t{nrow(info$df)}"), "\n\n")
}
---------- Boxscores '24 ---------- 
Features:   23 
Observations:   3049 

---------- Boxscores '25 ---------- 
Features:   23 
Observations:   3325 

---------- Boxscores '26 ---------- 
Features:   23 
Observations:   3866 
for (info in list(
  list(df = awards_24, label = "24"),
  list(df = awards_25, label = "25"),
  list(df = awards_26, label = "26")
)) {
  cat(str_glue("---------- Awards '{info$label} ----------"), "\n")
  cat(str_glue("Features:\t{ncol(info$df)}"), "\n")
  cat(str_glue("Observations:\t{nrow(info$df)}"), "\n\n")
}
---------- Awards '24 ---------- 
Features:   6 
Observations:   36 

---------- Awards '25 ---------- 
Features:   6 
Observations:   36 

---------- Awards '26 ---------- 
Features:   6 
Observations:   37

Data Wrangling / Transformation

Cleaning

# Combine the seasons for EDA and later transformations
# but must retain season for train/validate/test later
boxscores_24 <- boxscores_24 %>% mutate(season = "2023-24")
boxscores_25 <- boxscores_25 %>% mutate(season = "2024-25")
boxscores_26 <- boxscores_26 %>% mutate(season = "2025-26")

boxscores <- bind_rows(boxscores_24, boxscores_25, boxscores_26)
print("Boxscores")
[1] "Boxscores"
cat(str_glue("Features:\t{ncol(boxscores)}"), "\n")
Features:   24 
cat(str_glue("Observations:\t{nrow(boxscores)}"), "\n")
Observations:   10240 
awards <- bind_rows(awards_24, awards_25, awards_26) %>%
  rename(season = `Year Won`)

print("Awards")
[1] "Awards"
cat(str_glue("Features:\t{ncol(awards)}"), "\n")
Features:   6 
cat(str_glue("Observations:\t{nrow(awards)}"), "\n")
Observations:   109

Feature Engineering

Check Duplicates

check_duplicates <- function(df) {
  dupes <- duplicated(df)
  n <- sum(dupes)
  if (n > 0) {
    cat(str_glue("{n} duplicated rows found"), "\n")
  } else {
    cat("No duplicated rows found\n")
  }
  df[dupes, ]
}
print("Boxscore duplicate check:")
[1] "Boxscore duplicate check:"
check_duplicates(boxscores)
No duplicated rows found
# A tibble: 0 × 24
# ℹ 24 variables: player_id <dbl>, player <chr>, MIN <dbl>, PTS <dbl>,
#   FGM <dbl>, FGA <dbl>, 3PTM <dbl>, 3PTA <dbl>, FTM <dbl>, FTA <dbl>,
#   REB <dbl>, AST <dbl>, TO <dbl>, STL <dbl>, BLK <dbl>, OREB <dbl>,
#   DREB <dbl>, PF <dbl>, team <chr>, opponent <chr>, home <lgl>,
#   starter <lgl>, game_id <dbl>, season <chr>
print("Award duplicate check:")
[1] "Award duplicate check:"
check_duplicates(awards)
No duplicated rows found
# A tibble: 0 × 6
# ℹ 6 variables: Player Name <chr>, School <chr>, Conference <chr>,
#   Award Name <chr>, Subgroup <chr>, season <chr>

Remove Coach Awards

cat("Before removing coach awards\n")
Before removing coach awards
cat(str_glue("Features:\t{ncol(awards)}"), "\n")
Features:   6 
cat(str_glue("Observations:\t{nrow(awards)}"), "\n\n")
Observations:   109 
awards <- awards %>%
  filter(!str_detect(`Award Name`, regex("coach", ignore_case = TRUE)))

cat("After removing coach awards\n")
After removing coach awards
cat(str_glue("Features:\t{ncol(awards)}"), "\n")
Features:   6 
cat(str_glue("Observations:\t{nrow(awards)}"), "\n")
Observations:   105

Normalize Player Names

normalize <- function(name) {
  m <- str_match(name, "^\\s*([A-Za-z])[A-Za-z''.-]*\\s+([A-Za-z-]+)")
  if (is.na(m[1, 1])) {
    message(str_glue("Unable to handle {name}"))
    return(NA_character_)
  }
  str_to_lower(str_c(m[1, 2], m[1, 3]))
}

award_players <- unique(awards$`Player Name`)
awards_players_normal <- character()
cat("Name Conflicts for Awards (like a jr and sr):\n")
Name Conflicts for Awards (like a jr and sr):
for (name in award_players) {
  norm <- normalize(name)
  if (norm %in% awards_players_normal) {
    cat(str_glue("Conflict with {name}"), "\n")
  } else {
    awards_players_normal <- c(awards_players_normal, norm)
  }
}

boxscore_players <- unique(boxscores$player)
boxscore_players_normal <- character()
cat("Name Conflicts for Boxscores (like a jr and sr):\n")
Name Conflicts for Boxscores (like a jr and sr):
for (name in boxscore_players) {
  norm <- normalize(name)
  if (norm %in% boxscore_players_normal) {
    cat(str_glue("Conflict with {name}"), "\n")
  } else {
    boxscore_players_normal <- c(boxscore_players_normal, norm)
  }
}
Conflict with T. Horton III 
missing_in_boxscores <- setdiff(awards_players_normal, boxscore_players_normal)
for (name in missing_in_boxscores) {
  cat(str_glue("Award name '{name}' not found in boxscores"), "\n")
}
Award name 'ypowell' not found in boxscores 
boxscores <- boxscores %>%
  mutate(player = if_else(player == "Z. Powell", "Y. Powell", player))
normalize_col <- function(x) {
  map_chr(x, normalize)
}

awards <- awards %>% mutate(pname = normalize_col(`Player Name`))
boxscores <- boxscores %>% mutate(pname = normalize_col(player))

Train/Test Split

train_boxscores <- boxscores %>% filter(season %in% c("2023-24", "2024-25"))
test_boxscores  <- boxscores %>% filter(season == "2025-26")

cat("boxscores shape:", nrow(boxscores), "x", ncol(boxscores), "\n")
boxscores shape: 10240 x 25 
cat("train_boxscores shape:", nrow(train_boxscores), "x", ncol(train_boxscores), "\n")
train_boxscores shape: 6374 x 25 
cat("test_boxscores shape:", nrow(test_boxscores), "x", ncol(test_boxscores), "\n")
test_boxscores shape: 3866 x 25 
train_awards <- awards %>% filter(season %in% c("2023-24", "2024-25"))
test_awards  <- awards %>% filter(season == "2025-26")

cat("awards shape:", nrow(awards), "x", ncol(awards), "\n")
awards shape: 105 x 7 
cat("train_awards shape:", nrow(train_awards), "x", ncol(train_awards), "\n")
train_awards shape: 70 x 7 
cat("test_awards shape:", nrow(test_awards), "x", ncol(test_awards), "\n")
test_awards shape: 35 x 7

Aggregate by Player-Season

aggregate_player_season <- function(df) {
  # Helper to get the mode (most frequent value)
  stat_mode <- function(x) {
    ux <- unique(x)
    ux[which.max(tabulate(match(x, ux)))]
  }

  aggr <- df %>%
    group_by(pname, season) %>%
    summarise(
      min_total       = sum(MIN),
      min_pg          = mean(MIN),
      pts_pg          = mean(PTS),
      fg_made_pg      = mean(FGM),
      fg_attempted_pg = mean(FGA),
      tp_made_pg      = mean(`3PTM`),
      tp_attempted_pg = mean(`3PTA`),
      ft_made_pg      = mean(FTM),
      ft_attempted_pg = mean(FTA),
      ast_pg          = mean(AST),
      tov_pg          = mean(TO),
      stl_pg          = mean(STL),
      blk_pg          = mean(BLK),
      oreb_pg         = mean(OREB),
      dreb_pg         = mean(DREB),
      pf_pg           = mean(PF),
      team            = stat_mode(team),
      home_rate       = mean(home),
      starter_rate    = mean(starter),
      .groups = "drop"
    ) %>%
    mutate(
      fg_percentage_pg = if_else(fg_attempted_pg == 0, 0, fg_made_pg / fg_attempted_pg),
      tp_percentage_pg = if_else(tp_attempted_pg == 0, 0, tp_made_pg / tp_attempted_pg),
      ft_percentage_pg = if_else(ft_attempted_pg == 0, 0, ft_made_pg / ft_attempted_pg)
    )

  aggr
}
train_player_season <- aggregate_player_season(train_boxscores)
test_player_season <- aggregate_player_season(test_boxscores)
cat("train_player_season shape:", nrow(train_player_season), "x", ncol(train_player_season), "\n")
train_player_season shape: 266 x 24 
cat("test_player_season shape:", nrow(test_player_season), "x", ncol(test_player_season), "\n")
test_player_season shape: 166 x 24 
aggregate_awards <- function(df) {
  df %>%
    filter(`Award Name` != "Men's Conference USA All-Freshman Winners") %>%
    group_by(pname, season) %>%
    summarise(award_count = n(), .groups = "drop")
}

train_player_season_awards <- aggregate_awards(train_awards)
cat("train awards shape:", nrow(train_player_season_awards), "x", ncol(train_player_season_awards), "\n")
train awards shape: 40 x 3 
test_player_season_awards <- aggregate_awards(test_awards)
cat("test awards shape:", nrow(test_player_season_awards), "x", ncol(test_player_season_awards), "\n")
test awards shape: 23 x 3

Merge Boxscores and Awards

merge_boxscores_awards <- function(boxscores, awards) {
  boxscores %>%
    left_join(awards, by = c("pname", "season")) %>%
    mutate(award_count = replace_na(award_count, 0))
}
train <- merge_boxscores_awards(train_player_season, train_player_season_awards)
cat("train shape:", nrow(train), "x", ncol(train), "\n")
train shape: 266 x 25 
test_full_season <- merge_boxscores_awards(test_player_season, test_player_season_awards)
cat("test_full_season shape:", nrow(test_full_season), "x", ncol(test_full_season), "\n")
test_full_season shape: 166 x 25

Create Efficiency

add_efficiency <- function(df) {
  df %>%
    mutate(
      eff_pg = (pts_pg + dreb_pg + oreb_pg + ast_pg + stl_pg + blk_pg) -
        ((fg_attempted_pg - fg_made_pg) + (ft_attempted_pg - ft_made_pg) + tov_pg)
    )
}

train <- add_efficiency(train)
test_full_season <- add_efficiency(test_full_season)

Create Usage

get_team_totals <- function(df) {
  df %>%
    group_by(team, season) %>%
    summarise(
      team_min_pg = sum(min_pg),
      team_fga_pg = sum(fg_attempted_pg),
      team_fta_pg = sum(ft_attempted_pg),
      team_tov_pg = sum(tov_pg),
      .groups = "drop"
    )
}

add_usage <- function(df, team_totals) {
  df %>%
    left_join(team_totals, by = c("team", "season")) %>%
    mutate(
      player_poss = fg_attempted_pg + 0.44 * ft_attempted_pg + tov_pg,
      team_poss   = team_fga_pg + 0.44 * team_fta_pg + team_tov_pg,
      usg_pct     = 100 * (player_poss * (team_min_pg / 5)) / (min_pg * team_poss)
    ) %>%
    select(-player_poss, -team_poss)
}

train_team_totals <- get_team_totals(train)
train <- add_usage(train, train_team_totals)

test_team_totals <- get_team_totals(test_full_season)
test_full_season <- add_usage(test_full_season, test_team_totals)

Prepare X/y

features <- c(
  "dreb_pg",
  "oreb_pg",
  "ast_pg",
  "stl_pg",
  "blk_pg",
  "fg_attempted_pg",
  "fg_made_pg",
  "ft_attempted_pg",
  "ft_made_pg",
  "tov_pg",
  "team_fga_pg",
  "team_fta_pg",
  "team_tov_pg"
)
features_zero <- c("eff_pg")

X_train <- train %>% select(all_of(features))

# model cant handle two teams not playing in season 24 and 25 but in 26
zero_team_cols <- X_train %>%
  select(starts_with("team_")) %>%
  select(where(~ sum(.) == 0)) %>%
  names()
cat("training:", zero_team_cols, "\n")
training:  
X_train <- X_train %>% select(-any_of(zero_team_cols))

X_test_full <- test_full_season %>%
  select(all_of(features)) %>%
  select(-any_of(zero_team_cols))

y_train <- train$award_count
y_test  <- test_full_season$award_count

Export CSVs

write_csv(X_train, "../data/traintest/X_train.csv")
write_csv(X_test_full, "../data/traintest/X_test.csv")
write_csv(tibble(award_count = y_train), "../data/traintest/y_train.csv")
write_csv(tibble(award_count = y_test), "../data/traintest/y_test.csv")

Data Visualization

Plot config

# Combines train + test data for plots
train_df <- cbind(y = y_train, X_train)
test_df  <- cbind(y = y_test,  X_test_full)
full_df <- bind_rows(train_df, test_df)

# Generate Scatter Plots
# Scatter plot predictors
predictors <- c(
  "dreb_pg", "oreb_pg", "ast_pg", "stl_pg", "blk_pg",
  "fg_attempted_pg", "fg_made_pg",
  "ft_attempted_pg", "ft_made_pg",
  "tov_pg",
  "team_fga_pg", "team_fta_pg", "team_tov_pg"
)

# Full labels for titles/axes
Full_names <- c(
  dreb_pg         = "Defensive Rebounds Per Game",
  oreb_pg         = "Offensive Rebounds Per Game",
  ast_pg          = "Assists Per Game",
  stl_pg          = "Steals Per Game",
  blk_pg          = "Blocks Per Game",
  fg_attempted_pg = "Field Goal Attempts Per Game",
  fg_made_pg      = "Field Goals Made Per Game",
  ft_attempted_pg = "Free Throw Attempts Per Game",
  ft_made_pg      = "Free Throws Made Per Game",
  tov_pg          = "Turnovers Per Game",
  team_fga_pg     = "Team Field Goal Attempts Per Game",
  team_fta_pg     = "Team Free Throw Attempts Per Game",
  team_tov_pg     = "Team Turnovers Per Game"
)

scatter_plots = list()
# Scatter plots
for (pred_var in predictors) {
  
  label_name <- Full_names[[pred_var]]
  
  pred_scatter <- ggplot(full_df, aes_string(x = pred_var, y = "y")) +
    geom_point(color = "steelblue", alpha = 0.6, size = 2) +
    geom_smooth(method = "lm", se = TRUE, color = "red", linewidth = 1) +
    labs(
      title = paste("Award Count vs", label_name),
      x = label_name,
      y = "Award Count"
    ) +
    theme_minimal(base_size = 14)
  
  scatter_plots[[pred_var]] <- pred_scatter
}

Figure 1

scatter_plots$dreb_pg
Figure 1

Award count vs Defensive Rebounds Per Game: The regression line slopes upwards showing a positive relationship between the amount of awards won and defensive rebounds, with the higher the defensive rebounds a player has the higher chance of them having an award.

Figure 2

scatter_plots$oreb_pg
Figure 2

Award count vs Offensive Rebounds Per Game: The regression line shows a positive relationship, with the higher the Offensive rebounds a player has the higher chance of them having an award.

Figure 3

scatter_plots$ast_pg
Figure 3

Award count vs Assists Per Game: The regression line shows a positive relationship, with the higher the Assists per games a player has the higher chance of them having an award. Players with multiple awards have a moderate to high assist rate but assists alone do not guarantee awards.

Figure 4

scatter_plots$stl_pg
Figure 4

Award count vs Steal Per Game: The regression line shows a positive relationship. The more steals a player has per game the better chance there is at winning an award but much like assists, steals alone does not guarantee more awards.

Figure 5

scatter_plots$blk_pg
Figure 5

Award Count vs Blocks Per Game: The regression line shows a positive relationship. Blocks alone do not determine award success, but there is a higher chance of getting an award if you have more blocks.

Figure 6

scatter_plots$fg_attempted_pg
Figure 6

Award Count vs Field Goal Attempts Per Game: The regression line shows a positive relationship. Players who take more shots tend to have more awards but just because you take more shots does not guarantee that you will win more awards.

Figure 7

scatter_plots$fg_made_pg
Figure 7

Award Count vs Field Goals Made Per Game: The regression line shows a positive relationship. Players who make more more shots tend to have more awards.

Figure 8

scatter_plots$ft_attempted_pg
Figure 8

Award Count vs Free Throw Attempts Per Game: The regression line shows a positive relationship. The more Free throw attempts per game a player has the higher the chance of them having awards. this most likely coincides with shot attempts because if you are shooting more during a game, there is a higher chance of you getting fouled more.

Figure 9

scatter_plots$ft_made_pg
Figure 9

Award Count vs Free Throws Made Per Game: The regression line shows a positive relationship. Getting to the free throw line and making more free throws is associated with winning more awards likely because it reflects scoring ability which is highly associated with awards won.

Figure 10

scatter_plots$tov_pg
Figure 10

Award Count vs Turnovers Per Game: The regression line shows a positive relationship. This shows that having more turnovers per game is highly associated to having more awards. This does not mean turnovers cause awards. higher award players often handle the ball more, create offense, and play larger roles, which naturally can lead to more turnovers.

Figure 11

scatter_plots$team_fga_pg
Figure 11

Award Count vs Team Field Goal Attempts Per Game: The regression line has a slight downward slope, indicating little to no relationship between team shot volume and award count. There is no clear clustering of higher awards only at high or low team attempt values. How many shots a team takes per game does not appear to strongly influence individual award outcomes.

Figure 12

scatter_plots$team_fta_pg
Figure 12

Award Count vs Team Free Throw Attempts Per Game: The regression line has a slight downward slope, indicating little to no relationship between team free throw attempt volume and award count. Overall team free throw attempt totals do not appear to strongly influence individual award success.

Figure 13

scatter_plots$team_tov_pg
Figure 13

Award Count vs Team Turnovers Per Game: The regression line has a slight downward slope, indicating little to no relationship between team turnovers and award count. Overall team turnover totals do not appear to strongly determine individual award outcomes.

Interactive Figures

Interactive Figure 1

p_h <- ggplot(full_df, aes(x = y)) +
  geom_histogram(
    binwidth = 1,
    fill = "steelblue",
    color = "black",
    alpha = 0.8
  ) +
  scale_x_continuous(breaks = seq(min(full_df$y), max(full_df$y), by = 1)) +
  labs(
    title = "Distribution of Award Counts",
    x = "Award Count",
    y = "Frequency"
  ) +
  theme_minimal(base_size = 14)

ggplotly(p_h)

This histogram shows the distribution of our data set, showing that there are excessive zeros which is why for further research we will pursue a zero inflated poisson model.

Interactive Figure 2

p1 <- ggplot(train, aes(x = fg_made_pg, y = award_count, color=pname)) +
  geom_point(alpha = 0.6, size = 2) +
  geom_smooth(method = "lm", se = TRUE, color = "red", linewidth = 1) +
  labs(
    title = paste("Award Count vs Field Goals Made Per Game"),
    x = "Field Goals Made Per Game",
    y = "Award Count"
    ) +
  theme_minimal(base_size = 14) +
  theme(legend.position = "none")
ggplotly(p1, tooltip = c("pname"))

This interactive scatter-plot can be used to view the player that is associated with each data point.

Interactive Tables

num_cols <- names(train)[sapply(train, is.numeric)]

datatable(train) |>
  formatRound(columns = num_cols, digits = 2)

Conclusion

Overall, because of how basketball statistics are generally counts of positive events, then each individual predictor has a positive linear relationship with the response variable.

In contrast, the three team based predictors had slightly negative to no correlation with the response variable.

However, a single predictor or variable does not determine whether a player won an award or not, it is a combination of multiple predictors or variables that determine awards won and how many awards are won.

As a follow up to this exploration, we plan on modeling the interactions with a Zero Poisson Inflation Model.

Team Members

  • Kyle Bresko (kbresko@students.kennesaw.edu)

  • Charles Lane (clane30@students.keensaw.edu)

  • Kaleb Treangen (ktreang1@students.kennesaw.edu)

# KSU coordinates: 34.0382, -84.5819
leaflet() %>%
  addTiles() %>%
  setView(lng = -84.5819, lat = 34.0382, zoom = 15) %>%
  addMarkers(
    lng = -84.5819, lat = 34.0382,
    popup = "Kyle Bresko (kbresko@students.kennesaw.edu)<br>
    Charles Lane (clane30@students.keensaw.edu)<br>
    Kaleb Treangen (ktreang1@students.kennesaw.edu)"
  )
qr <- qr_code("https://rpubs.com/charlestl/r_project_ncaa_awards")
png(file="qr_code.png")
plot(qr)
dev.off()
pdf 
  2 
plot(qr)