Comparing WNBA Players

Player Comparison Framework

Comparing WNBA players requires understanding positional differences, era adjustments, and the unique characteristics of the women's game. Statistical comparison should account for pace, playing time, team quality, and role-specific responsibilities.

Python: Multi-Dimensional Player Comparison

import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler
from scipy.spatial.distance import euclidean
import matplotlib.pyplot as plt
import seaborn as sns

# Sample WNBA player data (normally loaded from API)
players_data = pd.DataFrame({
    'player': ['Breanna Stewart', 'A\'ja Wilson', 'Sabrina Ionescu',
               'Jewell Loyd', 'Alyssa Thomas', 'Napheesa Collier',
               'Kelsey Plum', 'Candace Parker', 'Jonquel Jones'],
    'position': ['F', 'F', 'G', 'G', 'F', 'F', 'G', 'F', 'C'],
    'minutes': [34.2, 33.8, 35.1, 33.5, 32.8, 33.2, 31.5, 28.9, 30.5],
    'points': [21.8, 22.8, 19.2, 19.6, 13.5, 20.5, 17.8, 13.2, 14.6],
    'rebounds': [8.5, 9.5, 4.8, 4.2, 9.6, 9.8, 2.9, 8.3, 8.9],
    'assists': [3.5, 2.2, 6.3, 3.8, 7.8, 3.4, 4.2, 5.1, 2.8],
    'steals': [1.5, 1.9, 1.2, 1.3, 1.7, 1.3, 1.0, 1.1, 0.8],
    'blocks': [1.2, 1.5, 0.3, 0.2, 0.8, 1.6, 0.1, 0.6, 1.4],
    'fg_pct': [0.460, 0.515, 0.447, 0.432, 0.542, 0.488, 0.438, 0.475, 0.502],
    'three_pct': [0.372, 0.315, 0.375, 0.385, 0.125, 0.345, 0.412, 0.328, 0.300],
    'ft_pct': [0.825, 0.815, 0.855, 0.882, 0.710, 0.843, 0.890, 0.788, 0.821],
    'turnovers': [2.8, 2.5, 3.2, 2.1, 3.5, 2.3, 2.4, 2.9, 1.9]
})

# =============================================================================
# 1. Calculate Advanced Metrics
# =============================================================================

def calculate_advanced_stats(player_stats, league_avg_pace=80):
    """Calculate advanced statistics for player comparison"""

    # True Shooting Percentage
    player_stats['ts_pct'] = (
        player_stats['points'] /
        (2 * (player_stats['points'] / player_stats['fg_pct'] +
              0.44 * player_stats['points'] / player_stats['ft_pct']))
    )

    # Per-36 minute stats
    min_factor = 36 / player_stats['minutes']
    player_stats['pts_per_36'] = player_stats['points'] * min_factor
    player_stats['reb_per_36'] = player_stats['rebounds'] * min_factor
    player_stats['ast_per_36'] = player_stats['assists'] * min_factor

    # Usage Rate (simplified - normally needs team data)
    # USG% = [(FGA + 0.44*FTA + TOV) * Team_Min] / [Min * (Team_FGA + 0.44*Team_FTA + Team_TOV)]
    # Simplified approximation
    player_stats['usage_rate'] = (
        (player_stats['points'] + player_stats['turnovers']) /
        player_stats['minutes'] * 100
    )

    # Assist-to-Turnover Ratio
    player_stats['ast_to_ratio'] = (
        player_stats['assists'] / player_stats['turnovers']
    )

    # Player Efficiency Rating (simplified)
    player_stats['per'] = (
        player_stats['points'] +
        player_stats['rebounds'] +
        player_stats['assists'] +
        player_stats['steals'] +
        player_stats['blocks'] -
        (player_stats['points'] / player_stats['fg_pct'] - player_stats['points']) -
        player_stats['turnovers']
    ) / player_stats['minutes'] * 36

    return player_stats

players_data = calculate_advanced_stats(players_data)

print("=== Advanced Player Statistics ===")
print(players_data[['player', 'position', 'ts_pct', 'per',
                     'usage_rate', 'ast_to_ratio']].round(3))

# =============================================================================
# 2. Multi-Dimensional Player Similarity
# =============================================================================

def find_similar_players(target_player, all_players, n_similar=3):
    """Find most similar players using Euclidean distance"""

    # Select comparison features
    features = ['pts_per_36', 'reb_per_36', 'ast_per_36',
                'ts_pct', 'usage_rate', 'steals', 'blocks']

    # Standardize features
    scaler = StandardScaler()
    scaled_features = scaler.fit_transform(all_players[features])

    # Find target player index
    target_idx = all_players[all_players['player'] == target_player].index[0]
    target_vector = scaled_features[target_idx]

    # Calculate distances
    distances = []
    for idx, player_vector in enumerate(scaled_features):
        if idx != target_idx:
            dist = euclidean(target_vector, player_vector)
            distances.append({
                'player': all_players.iloc[idx]['player'],
                'distance': dist
            })

    # Sort by similarity (smallest distance)
    similar_df = pd.DataFrame(distances).sort_values('distance').head(n_similar)

    return similar_df

# Example: Find players similar to Breanna Stewart
similar_to_stewart = find_similar_players('Breanna Stewart', players_data, n_similar=3)
print("\n=== Players Most Similar to Breanna Stewart ===")
print(similar_to_stewart)

# =============================================================================
# 3. Position-Specific Rankings
# =============================================================================

def rank_by_position(players_df, stat='per'):
    """Rank players within their position"""

    position_rankings = players_df.copy()
    position_rankings['position_rank'] = (
        position_rankings.groupby('position')[stat]
        .rank(ascending=False, method='min')
    )

    return position_rankings.sort_values(['position', 'position_rank'])

position_rankings = rank_by_position(players_data, stat='per')
print("\n=== Position Rankings by PER ===")
print(position_rankings[['player', 'position', 'per', 'position_rank']])

# =============================================================================
# 4. Player Comparison Radar Chart
# =============================================================================

def create_comparison_radar(player1, player2, all_players):
    """Create radar chart comparing two players"""

    categories = ['Points', 'Rebounds', 'Assists', 'TS%', 'Steals', 'Blocks']

    # Normalize stats to 0-100 scale
    p1_data = all_players[all_players['player'] == player1].iloc[0]
    p2_data = all_players[all_players['player'] == player2].iloc[0]

    # Calculate percentiles
    p1_values = [
        (p1_data['points'] / all_players['points'].max()) * 100,
        (p1_data['rebounds'] / all_players['rebounds'].max()) * 100,
        (p1_data['assists'] / all_players['assists'].max()) * 100,
        (p1_data['ts_pct'] / all_players['ts_pct'].max()) * 100,
        (p1_data['steals'] / all_players['steals'].max()) * 100,
        (p1_data['blocks'] / all_players['blocks'].max()) * 100,
    ]

    p2_values = [
        (p2_data['points'] / all_players['points'].max()) * 100,
        (p2_data['rebounds'] / all_players['rebounds'].max()) * 100,
        (p2_data['assists'] / all_players['assists'].max()) * 100,
        (p2_data['ts_pct'] / all_players['ts_pct'].max()) * 100,
        (p2_data['steals'] / all_players['steals'].max()) * 100,
        (p2_data['blocks'] / all_players['blocks'].max()) * 100,
    ]

    # Close the plot
    p1_values += p1_values[:1]
    p2_values += p2_values[:1]

    angles = np.linspace(0, 2 * np.pi, len(categories), endpoint=False).tolist()
    angles += angles[:1]

    fig, ax = plt.subplots(figsize=(8, 8), subplot_kw=dict(projection='polar'))

    ax.plot(angles, p1_values, 'o-', linewidth=2, label=player1, color='blue')
    ax.fill(angles, p1_values, alpha=0.25, color='blue')

    ax.plot(angles, p2_values, 'o-', linewidth=2, label=player2, color='red')
    ax.fill(angles, p2_values, alpha=0.25, color='red')

    ax.set_xticks(angles[:-1])
    ax.set_xticklabels(categories)
    ax.set_ylim(0, 100)
    ax.set_title(f"Player Comparison: {player1} vs {player2}", size=14, pad=20)
    ax.legend(loc='upper right', bbox_to_anchor=(1.3, 1.1))
    ax.grid(True)

    return fig

# Example: Compare two elite players
# fig = create_comparison_radar('Breanna Stewart', 'A\'ja Wilson', players_data)
# plt.show()

print("\n=== Player Comparison Tools Available ===")
print("1. Advanced stat calculator (TS%, PER, Per-36)")
print("2. Similarity finder (Euclidean distance)")
print("3. Position-specific rankings")
print("4. Radar chart visualizations")

library(wehoop)
library(tidyverse)
library(scales)
library(ggrepel)

# Load WNBA player box scores for current season
player_stats <- wehoop::load_wnba_player_box(seasons = 2024)

# =============================================================================
# 1. Calculate Season Averages and Advanced Stats
# =============================================================================

season_averages <- player_stats %>%
  group_by(athlete_id, athlete_display_name, team_display_name) %>%
  summarise(
    games = n(),
    minutes = mean(minutes, na.rm = TRUE),
    points = mean(points, na.rm = TRUE),
    rebounds = mean(rebounds, na.rm = TRUE),
    assists = mean(assists, na.rm = TRUE),
    steals = mean(steals, na.rm = TRUE),
    blocks = mean(blocks, na.rm = TRUE),
    turnovers = mean(turnovers, na.rm = TRUE),
    fgm = sum(field_goals_made, na.rm = TRUE),
    fga = sum(field_goals_attempted, na.rm = TRUE),
    fg3m = sum(three_point_field_goals_made, na.rm = TRUE),
    fg3a = sum(three_point_field_goals_attempted, na.rm = TRUE),
    ftm = sum(free_throws_made, na.rm = TRUE),
    fta = sum(free_throws_attempted, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  filter(games >= 10) %>%  # Min 10 games played
  mutate(
    # Advanced metrics
    fg_pct = fgm / fga,
    three_pct = fg3m / fg3a,
    ft_pct = ftm / fta,
    ts_pct = points / (2 * (fga + 0.44 * fta)),

    # Per-36 minute stats
    pts_per_36 = points * (36 / minutes),
    reb_per_36 = rebounds * (36 / minutes),
    ast_per_36 = assists * (36 / minutes),

    # Efficiency ratios
    ast_to_ratio = assists / turnovers,
    usage_rate = (points + turnovers) / minutes * 100,

    # Simplified PER
    per = (points + rebounds + assists + steals + blocks -
           (fga - fgm) - turnovers) / minutes * 36
  )

cat("=== Top Players by PER (min 10 games) ===\n")
print(season_averages %>%
  select(athlete_display_name, team_display_name, games,
         points, rebounds, assists, per) %>%
  arrange(desc(per)) %>%
  head(10))

# =============================================================================
# 2. Find Similar Players Using Statistical Distance
# =============================================================================

find_similar_players <- function(player_name, stats_df, n = 5) {
  # Select comparison features
  features <- c("pts_per_36", "reb_per_36", "ast_per_36",
                "ts_pct", "steals", "blocks")

  # Normalize features
  stats_normalized <- stats_df %>%
    select(athlete_display_name, all_of(features)) %>%
    mutate(across(all_of(features), ~scale(.x)[,1]))

  # Get target player vector
  target <- stats_normalized %>%
    filter(athlete_display_name == player_name)

  if (nrow(target) == 0) {
    cat("Player not found\n")
    return(NULL)
  }

  # Calculate Euclidean distance to all other players
  stats_normalized %>%
    filter(athlete_display_name != player_name) %>%
    rowwise() %>%
    mutate(
      distance = sqrt(sum((c_across(all_of(features)) -
                          target[1, features])^2))
    ) %>%
    ungroup() %>%
    arrange(distance) %>%
    head(n) %>%
    select(athlete_display_name, distance)
}

# Example: Find players similar to top scorer
top_scorer <- season_averages %>%
  arrange(desc(points)) %>%
  slice(1) %>%
  pull(athlete_display_name)

cat(sprintf("\n=== Players Similar to %s ===\n", top_scorer))
similar_players <- find_similar_players(top_scorer, season_averages, n = 5)
print(similar_players)

# =============================================================================
# 3. Position-Based Comparison (Guards vs Forwards vs Centers)
# =============================================================================

# Classification by playing style (based on stats)
position_classification <- season_averages %>%
  mutate(
    position_type = case_when(
      assists >= 5 & rebounds < 6 ~ "Guard",
      rebounds >= 7 & assists < 5 ~ "Big",
      TRUE ~ "Wing/Forward"
    )
  )

cat("\n=== Position Type Averages ===\n")
position_averages <- position_classification %>%
  group_by(position_type) %>%
  summarise(
    players = n(),
    avg_points = mean(points),
    avg_rebounds = mean(rebounds),
    avg_assists = mean(assists),
    avg_ts_pct = mean(ts_pct, na.rm = TRUE),
    .groups = "drop"
  )
print(position_averages)

# =============================================================================
# 4. Head-to-Head Player Comparison Function
# =============================================================================

compare_players <- function(player1, player2, stats_df) {
  comparison <- stats_df %>%
    filter(athlete_display_name %in% c(player1, player2)) %>%
    select(athlete_display_name, games, minutes, points, rebounds,
           assists, steals, blocks, ts_pct, per) %>%
    arrange(athlete_display_name)

  return(comparison)
}

# Example comparison
top_2_scorers <- season_averages %>%
  arrange(desc(points)) %>%
  head(2) %>%
  pull(athlete_display_name)

if (length(top_2_scorers) >= 2) {
  cat(sprintf("\n=== Head-to-Head: %s vs %s ===\n",
              top_2_scorers[1], top_2_scorers[2]))
  comparison <- compare_players(top_2_scorers[1], top_2_scorers[2],
                                season_averages)
  print(comparison)
}

# =============================================================================
# 5. Visualization: Scoring vs Efficiency
# =============================================================================

ggplot(season_averages %>% filter(games >= 15),
       aes(x = ts_pct, y = points, label = athlete_display_name)) +
  geom_point(aes(size = usage_rate, color = per), alpha = 0.7) +
  geom_text_repel(data = . %>% filter(points >= 15 | ts_pct >= 0.60),
                  size = 3, max.overlaps = 10) +
  scale_color_gradient(low = "lightblue", high = "darkred",
                       name = "PER") +
  scale_x_continuous(labels = percent_format(accuracy = 1)) +
  scale_size_continuous(name = "Usage Rate") +
  labs(
    title = "WNBA Player Comparison: Scoring Volume vs Efficiency",
    subtitle = "Minimum 15 games played, sized by usage rate",
    x = "True Shooting %",
    y = "Points Per Game"
  ) +
  theme_minimal() +
  theme(legend.position = "right")

# =============================================================================
# 6. All-Around Performance: Points, Rebounds, Assists
# =============================================================================

ggplot(season_averages %>% filter(games >= 15),
       aes(x = rebounds, y = assists, size = points)) +
  geom_point(alpha = 0.6, color = "steelblue") +
  geom_text_repel(aes(label = athlete_display_name),
                  data = . %>% filter(points >= 15),
                  size = 3) +
  labs(
    title = "WNBA All-Around Performance",
    subtitle = "Points (size), Rebounds, and Assists",
    x = "Rebounds Per Game",
    y = "Assists Per Game",
    size = "Points Per Game"
  ) +
  theme_minimal()

cat("\n=== Player Comparison Framework Complete ===\n")
cat("✓ Advanced stat calculations\n")
cat("✓ Similarity analysis\n")
cat("✓ Position-based comparisons\n")
cat("✓ Head-to-head tools\n")
cat("✓ Visualization dashboards\n")

Context Matters in Comparisons

When comparing WNBA players, consider team context, playing style, and role. A high-usage scorer on a weak team may have inflated counting stats but lower efficiency. Conversely, a role player on an elite team may have excellent efficiency but limited volume.