data_plot_maps.py

import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
from pyspark.sql.functions import first

from ea_common import spark, filtered_rounds, db_path

# Filtered rounds needs to grouped up by user_id
filtered_rounds \
    .groupBy("round_id") \
    .agg(
        first("user_id").alias("user_id"),
        first("team").alias("team"),
        first("result").alias("result"),
        first("map_name").alias("map_name"),
        first("server_name").alias("server_name"),
        first("start_time").alias("start_time"),
        first("end_time").alias("end_time")
    ).createOrReplaceTempView("filtered_rounds")

# Map balance:
# - Precomputed by data_analysis_maps.py (win percentage per map, filtered
#   to classic ns_ maps with at least 3 recorded wins for each team) and
#   written to db/map_balance. Just read and plot it here.
map_stats_pd = spark.read.parquet(db_path("map_balance")).toPandas()

# Set the map_name as the index for the DataFrame
map_stats_pd.set_index('map_name', inplace=True)

# Sort by marine win percentage, matching the previous ORDER BY
map_stats_pd.sort_values('marine_win_percentage', ascending=False, inplace=True)

map_stats_pd = map_stats_pd.rename(columns={'marine_win_percentage': 'Marine Win %',
                                            'alien_win_percentage': 'Alien Win %'})

# Create a stacked bar chart with percentages
map_stats_pd[['Marine Win %', 'Alien Win %']].plot(kind='barh',
                                                   stacked=True,
                                                   color=['#0492C2', '#FCAE1E'])

# Add labels and title
plt.xlabel('Map Name')
plt.ylabel('Win Percentage')
plt.title('Marine and Alien Win Percentage by Map')

# Set the x-axis ticks
plt.xticks(np.arange(0, 101, 10))


# Add a vertical line at the 50% mark
plt.axvline(x=50, color='r', linestyle='dotted')

# Map popularity:
# - Calculate the total games for each map
# - Include only classic ns maps (ns_ prefix)

map_popularity_df = spark.sql("""
SELECT
    LOWER(r.map_name) as map_name,
    COUNT(*) as total_games
FROM
    filtered_rounds r
WHERE
    r.map_name IS NOT NULL AND
    LOWER(r.map_name) != 'none' AND
    LOWER(r.map_name) LIKE 'ns_%'
GROUP BY
    LOWER(r.map_name)
ORDER BY
    total_games DESC
""")

map_popularity_df.show()

# Convert the Spark DataFrame to a Pandas DataFrame
map_popularity_pd = map_popularity_df.toPandas()

# Set the map_name as the index for the DataFrame
map_popularity_pd.set_index('map_name', inplace=True)

# Calculate the percentage of total games for each map
map_popularity_pd['percentage'] = map_popularity_pd['total_games'] / map_popularity_pd['total_games'].sum() * 100

# Create a new DataFrame for maps with less than 5% of total games
other_maps = map_popularity_pd[map_popularity_pd['percentage'] < 1.5].sum()

# Remove these maps from the original DataFrame
map_popularity_pd = map_popularity_pd[map_popularity_pd['percentage'] >= 1.5]

# Add the new DataFrame to the original DataFrame
map_popularity_pd = map_popularity_pd._append(pd.DataFrame({'total_games': other_maps['total_games'],
                                                            'percentage': other_maps['percentage']},
                                                            index=['Other maps']))

# Generate a color map with as many colors as there are data points
cmap = plt.get_cmap('Paired') 
colors = cmap(np.linspace(0, 1, len(map_popularity_pd)))

# Create a pie chart with the color map
plt.figure(figsize=(10, 5))
map_popularity_pd['total_games'].plot(kind='pie', autopct='%1.1f%%', colors=colors)

# Some tuning
plt.ylabel('')
plt.title('Map Popularity')

plt.show()