Vibe 游戏： 4人乒乓球与游戏数据分析

AI
游戏
数据分析
Author

Tony D

Published

February 10, 2026

游戏简介

VibePong 是一款简单而有趣的乒乓类游戏，支持最多 4 名玩家同台竞技，也支持与 CPU 对手作战。游戏会记录各种指标，如球速、游戏时长和玩家表现，并将这些数据保存到 CSV 文件中以便进行分析。

立即游玩：https://jcwinning.github.io/vibepong

游戏数据分析

在本文档中，我们分析了 VibePong 生成的游戏数据。分析内容涵盖数据提取、清洗以及带有可视化效果的探索性数据分析 (EDA)。

第一步：读取数据

首先，我们在 game_data 目录中搜索所有的摘要 (summary) 和动作 (action),遥感(telemetry) CSV 文件，并将其加载到 pandas DataFrame 中。

显示代码 
import pandas as pd
import glob
import os

# 加载 game_data 目录下的所有 CSV 文件
data_dir = './game_data'
def load_combined(pattern):
    files = glob.glob(os.path.join(data_dir, f'vibepong_{pattern}_*.csv'))
    return pd.concat([pd.read_csv(f) for f in files], ignore_index=True) if files else pd.DataFrame()

df_summary = load_combined('summary')
df_actions = load_combined('actions')
df_telemetry = load_combined('telemetry')

print(f"已加载 {len(df_summary)} 行摘要数据，{len(df_actions)} 条动作记录，以及 {len(df_telemetry)} 条遥测记录。")
已加载 75 行摘要数据，366 条动作记录，以及 658 条遥测记录。

逐场游戏数据

显示代码 
# Display the first few rows of summary data
df_summary.head()
Game ID Date Duration (s) Winner Ball Speed Theme Language Player Lives Hits CPU Difficulty
0 G1770713902861-393 2026-02-10T08:58:27.879Z 5.02 CPU 6.56 light en Player 1 0 0 NaN
1 G1770713902861-393 2026-02-10T08:58:27.879Z 5.02 CPU 6.56 light en CPU 1 0 NaN
2 G1770783544585-550 2026-02-11T04:20:01.889Z 57.30 Player 1 20.38 dark en Player 1 1 14 NaN
3 G1770783544585-550 2026-02-11T04:20:01.889Z 57.30 Player 1 20.38 dark en CPU 0 14 NaN
4 G1770713913739-52 2026-02-10T08:58:46.511Z 12.77 Player 3 7.08 light en Player 1 0 0 NaN

逐个动作数据

显示代码 
# Display the first few rows of summary data
df_actions.head()
Game ID Timestamp (ms) Player Action Details
0 G1770783544585-550 4000 System Ball served towards CPU NaN
1 G1770783544585-550 4919 CPU Hit Ball NaN
2 G1770783544585-550 6787 Player 1 Hit Ball NaN
3 G1770783544585-550 8653 CPU Hit Ball NaN
4 G1770783544585-550 10519 Player 1 Hit Ball NaN

时间序列数据

显示代码 
# Display the first few rows of summary data
df_telemetry.head()
Game ID Timestamp (ms) Ball X Ball Y Ball VX Ball VY Ball Speed p1 X p1 Y p1 Center ... p3 Center p3 Active p3 Lives p3 Distance p4 X p4 Y p4 Center p4 Active p4 Lives p4 Distance
0 G1770793570881-726 4011 343.50 349.77 -6.496 -0.2292 6.5010 20.0 300.0 350.0 ... 350.0 1.0 1.0 0.0 NaN NaN NaN NaN NaN NaN
1 G1770793570881-726 4112 304.53 348.40 -6.496 -0.2292 6.5073 20.0 300.0 350.0 ... 350.0 1.0 1.0 0.0 NaN NaN NaN NaN NaN NaN
2 G1770793570881-726 4227 259.06 346.79 -6.496 -0.2292 6.5146 20.0 300.0 350.0 ... 350.0 1.0 1.0 0.0 NaN NaN NaN NaN NaN NaN
3 G1770793570881-726 4328 220.08 345.42 -6.496 -0.2292 6.5208 20.0 300.0 350.0 ... 350.0 1.0 1.0 0.0 NaN NaN NaN NaN NaN NaN
4 G1770793570881-726 4428 181.11 344.04 -6.496 -0.2292 6.5271 20.0 300.0 350.0 ... 350.0 1.0 1.0 0.0 NaN NaN NaN NaN NaN NaN

5 rows × 31 columns

第二步：数据清洗

我们将通过将时间戳转换为 datetime 对象、处理数值类型并确保一致性来清洗数据。

显示代码 
# 清洗摘要数据
df_summary['Date'] = pd.to_datetime(df_summary['Date'])
numeric_summary = ['Duration (s)', 'Ball Speed', 'Lives', 'Hits']
df_summary[numeric_summary] = df_summary[numeric_summary].apply(pd.to_numeric, errors='coerce')
df_summary = df_summary.dropna(subset=['Player'])

# 清洗动作数据
df_actions['Timestamp (ms)'] = pd.to_numeric(df_actions['Timestamp (ms)'], errors='coerce')

df_summary.head()
Game ID Date Duration (s) Winner Ball Speed Theme Language Player Lives Hits CPU Difficulty
0 G1770713902861-393 2026-02-10 08:58:27.879000+00:00 5.02 CPU 6.56 light en Player 1 0 0 NaN
1 G1770713902861-393 2026-02-10 08:58:27.879000+00:00 5.02 CPU 6.56 light en CPU 1 0 NaN
2 G1770783544585-550 2026-02-11 04:20:01.889000+00:00 57.30 Player 1 20.38 dark en Player 1 1 14 NaN
3 G1770783544585-550 2026-02-11 04:20:01.889000+00:00 57.30 Player 1 20.38 dark en CPU 0 14 NaN
4 G1770713913739-52 2026-02-10 08:58:46.511000+00:00 12.77 Player 3 7.08 light en Player 1 0 0 NaN

第三步：EDA 和可视化

现在我们将查看一些关键绩效指标并可视化游戏结果。

获胜分布

显示代码 
import seaborn as sns
import matplotlib.pyplot as plt

# Set aesthetic style
sns.set_theme(style="whitegrid")
plt.rcParams['font.sans-serif'] = ['Arial Unicode MS']
plt.rcParams['axes.unicode_minus'] = False

谁赢得了最多的游戏？

显示代码 
# 计算每位获胜者的独特游戏场次
unique_games = df_summary.drop_duplicates(subset=['Game ID'])
winner_counts = unique_games['Winner'].value_counts()

plt.figure(figsize=(10, 5))
sns.barplot(x=winner_counts.index, y=winner_counts.values, palette="viridis")
plt.title('玩家/CPU 获胜场次分布', fontsize=14)
plt.show()

游戏时长 vs. 球速

更高的球速会导致游戏时长变短吗？

显示代码 
plt.figure(figsize=(10, 6))
sns.scatterplot(data=unique_games, x='Ball Speed', y='Duration (s)', hue='Winner', s=100)
plt.title('Game Duration vs Ball Speed', fontsize=15)
plt.show()

每个玩家的总击球数

跟踪不同游戏场次的技能（击球数）表现。

显示代码 
plt.figure(figsize=(10, 6))
sns.boxplot(data=df_summary, x='Player', y='Hits', palette="magma")
plt.title('Distribution of Hits per Player', fontsize=15)
plt.show()

动作序列

分析所有游戏中动作的频率。

显示代码 
action_counts = df_actions['Action'].value_counts().head(10)

plt.figure(figsize=(8, 7))
action_counts.plot(kind='barh', color='skyblue')
plt.title('Most Frequent Actions/Events', fontsize=15)
plt.gca().invert_yaxis()
plt.show()

第四步：遥测数据分析

现在让我们分析高频遥测数据，该数据每 100 毫秒捕获一次球和球拍的位置。

随时间变化的球速

球速在所有游戏中是如何演变的？

显示代码 
if not df_telemetry.empty:
    plt.figure(figsize=(10, 5))
    for gid in df_telemetry['Game ID'].unique()[:10]:
        df_g = df_telemetry[df_telemetry['Game ID'] == gid]
        plt.plot(df_g['Timestamp (ms)']/1000, df_g['Ball Speed'], alpha=0.6, label=f'游戏 {gid[-4:]}')
    
    plt.title('球速演变（前 10 场游戏）')
    plt.xlabel('时间 (秒)')
    plt.ylabel('速度')
    plt.legend(bbox_to_anchor=(1.05, 1), loc='upper left', fontsize=8)
    plt.tight_layout()
    plt.show()

球轨迹热图

球在场地的哪个位置停留时间最长？

显示代码 
if not df_telemetry.empty:
    plt.figure(figsize=(10, 6))
    plt.hist2d(df_telemetry['Ball X'], df_telemetry['Ball Y'], bins=50, cmap='hot')
    plt.colorbar(label='采样点')
    plt.title('球轨迹热图')
    plt.show()

球拍移动分析

哪些玩家移动球拍最频繁？

显示代码 
if not df_telemetry.empty:
    dist_cols = [c for c in df_telemetry.columns if 'Distance' in c and 'Ball' not in c]
    movement = []
    for gid in df_telemetry['Game ID'].unique():
        df_g = df_telemetry[df_telemetry['Game ID'] == gid]
        for col in dist_cols:
            movement.append({'Player': col.split()[0], 'Distance': df_g[col].sum()})
    
    plt.figure(figsize=(10, 5))
    sns.boxplot(data=pd.DataFrame(movement), x='Player', y='Distance', palette='coolwarm')
    plt.title('各玩家球拍移动距离分布')
    plt.show()