app.R

library(dplyr)
library(shiny)
library(DT)
library(RColorBrewer)
library(scales)
library(readr)
library(reactable)
library(fmsb)

#This project has the goal of creating a shiny app to give the most similar current NBA player to current college players, 
#with the hopes of determining logical player comps for draft prospects.



stats25 <- read_csv("stats25.csv")
stats_past <- read_csv("stats_past.csv")


rd_bu_colors <- colorRampPalette(rev(brewer.pal(11, "RdBu")))
palette_100 <- rd_bu_colors(100) 
ui <- fluidPage(
  titlePanel("Find the Most Similar College Player"),
  
  sidebarLayout(
    sidebarPanel(
      selectizeInput("selected_player", "Select a Player:", choices = NULL, multiple = FALSE, options = list(maxOptions = 1000)),
      actionButton("find_button", "Find Most Similar Players")
    ),
    mainPanel(
      tabsetPanel(
        tabPanel("Similar Players", reactableOutput("comparison_table")),
        tabPanel("Radar Plot", 
                 plotOutput("radar_plot", height = "600px"))
      )
    )
  )
)

server <- function(input, output, session) {
  
  observe({
    updateSelectizeInput(session, "selected_player", choices = unique(stats25$name), server = TRUE)
  })
  
  find_most_similar_players <- function(input_player_name) {
    
    stat_cols <- c("ppg", "apg", "rpg", "spg", "bpg", "fg_pct", "fg3_pct", "to", "fgm", "fga", "fg3m", "fg3a")
    
    available_cols <- intersect(c("name", "season", "team", stat_cols), colnames(stats25))
    available_cols_past <- intersect(c("name", "season", "team", stat_cols), colnames(stats_past))
    
    input_player <- stats25 %>%
      filter(name == input_player_name) %>%
      select(all_of(available_cols))
    
    scaled_past_stats <- stats_past %>%
      mutate(across(all_of(stat_cols), ~ (.-mean(., na.rm = TRUE)) / sd(., na.rm = TRUE))) %>%
      select(name, all_of(stat_cols))
    
    input_player_scaled <- (input_player[, stat_cols] - colMeans(stats_past[stat_cols], na.rm = TRUE)) / 
      apply(stats_past[stat_cols], 2, sd, na.rm = TRUE)
    input_player_scaled[is.na(input_player_scaled)] <- 0
    
    input_player_scaled <- as.numeric(input_player_scaled)
    
    euclidean_distances <- rowSums((as.matrix(scaled_past_stats[, stat_cols]) - 
                                      matrix(input_player_scaled, nrow = nrow(scaled_past_stats), 
                                             ncol = length(input_player_scaled), byrow = TRUE))^2)
    
    top_5_indices <- order(euclidean_distances)[1:5]
    top_5_similar_players <- stats_past[top_5_indices, available_cols_past]  # Use actual stats
    
    comparison_df <- bind_rows(input_player, top_5_similar_players) %>%
      mutate(Player_Type = ifelse(name == input_player_name, "Selected", "Similar")) %>%
      mutate(across(c("ppg", "apg", "rpg", "spg", "bpg", "to", "fgm", "fga", "fg3m", "fg3a"), ~ round(., 1))) %>%
      mutate(across(c("fg_pct", "fg3_pct"), ~ round(., 3))) %>%
      select(name, season, team, Player_Type, everything()) %>%  # Reorder columns here
      rename(PPG = ppg,
             APG = apg,
             RPG = rpg,
             SPG = spg,
             BPG = bpg,
             `FG%` = fg_pct,
             `3P%` = fg3_pct,
             TO = to,
             FGM = fgm,
             FGA = fga,
             FG3M = fg3m,
             FG3A = fg3a)
    
    return(comparison_df)
  }
  
  observeEvent(input$find_button, {
    
    stat_cols <- c("PPG", "APG", "RPG", "SPG", "BPG", "FG%", "3P%", "TO", "FGM", "FGA", "FG3M", "FG3A")
    
    comparison_df <- find_most_similar_players(input$selected_player)
    
    output$comparison_table <- renderReactable({
      reactable(
        comparison_df,
        columns = c(
          list(
            Player_Type = colDef(show = FALSE),
            name = colDef(name = "Player"),
            season = colDef(name = "Season"),
            team = colDef(name = "Team")
          ),
          setNames(lapply(stat_cols, function(stat) {
            colDef(
              style = function(value, index) {
                if (comparison_df$Player_Type[index] == "Selected") return(NULL)
                
                selected_value <- comparison_df %>% filter(Player_Type == "Selected") %>% pull(!!sym(stat))
                diff <- value - selected_value
                
                if (stat %in% c("FG%", "3P%")) {
                  rescaled_value <- rescale(diff, from = c(-0.1, 0.1), to = c(0, 1))
                } else {
                  rescaled_value <- rescale(diff, from = c(-2, 2), to = c(0, 1))
                }
                
                rescaled_value <- pmin(pmax(rescaled_value, 0), 1)
                color <- palette_100[round(rescaled_value * 99) + 1]
                
                rgb_values <- col2rgb(color) / 255
                luminance <- 0.2126 * rgb_values[1] + 0.7152 * rgb_values[2] + 0.0722 * rgb_values[3]
                
                text_color <- ifelse(luminance > 0.5, "black", "white")
                
                list(background = color, color = text_color)
              }
            )
          }), stat_cols)
        ),
        bordered = TRUE,
        striped = TRUE,
        highlight = TRUE
      )
    })
    output$radar_plot <- renderPlot({
      req(input$selected_player)
      
      stats_columns <- c("mpg", "ppg", "apg", "rpg", "spg", "bpg", "fg_pct", "fg3_pct")
      
      percentile_data <- stats25 %>%
        ungroup() %>% 
        mutate(across(all_of(stats_columns), ~ percent_rank(.) * 100))
      
      player_percentiles <- percentile_data %>%
        filter(name == input$selected_player) %>%
        select(all_of(stats_columns))
      
      if (nrow(player_percentiles) == 0) {
        return(NULL)
      }
      
      radar_df <- rbind(rep(100, length(stats_columns)), 
                        rep(0, length(stats_columns)), 
                        player_percentiles)
      colnames(radar_df) <- stats_columns
      
      radar_df <- radar_df %>% 
        rename(MPG = mpg,
               PPG = ppg,
               APG = apg,
               RPG = rpg,
               SPG = spg,
               BPG = bpg,
               `FG%` = fg_pct,
               `3P%` = fg3_pct)
      
      radarchart(
        radar_df,
        axistype = 1,
        pcol = "darkgreen", 
        pfcol = rgb(0, 0.5, 0, 0.3), 
        plwd = 2,
        cglcol = "gray", 
        cglty = 1, 
        axislabcol = "black",
        cglwd = 0.8, 
        vlcex = 1.5,
        title = paste("Radar Plot for", input$selected_player)
      )
    })
  })
}


shinyApp(ui, server)