python - 可视化面对面记录的最佳方式是什么？

Question

我有一个数据框如下所示：

A = pd.DataFrame({'team':[1,2,3,2,1,1,3,4,1,2], 'opp_team':[2,1,2,3,3,4,1,1,2,1], 'result':[1,0,1,0,1,1,0,0,1,0]})

结果列有 1 代表胜利，0 代表失败。我想找出在图表上呈现头对头记录的最佳方式。

我在想一对情节，但我认为它不会奏效，因为它不会显示胜利和失败的计数。在上面的例子中，球队 1 与球队 2 交手 2 次，并且都赢了。因此，图表也应该显示计数。

我可以尝试什么来解决这个问题？

标签： pythonpandasgraphvisualization

Answer 1

只需将它们显示在 2 个单独的图表中，1 个图表显示对战记录的总数，另一个显示一个团队对另一个团队的总胜率（另一个团队的胜率 = 团队的输球率）。

为此，我认为需要对数据框进行重组，使其每行仅显示 1 个游戏 ID。为了便于分组，对 and 进行排序team，opp_team使得索引team总是小于索引opp_team

我生成了一个样本数据集，我将如何总结和绘制它，供您参考：

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np

def generate_data(num_data=100, num_teams=4):

    team = np.random.randint(1,1 + num_teams,num_data)
    opp_team = np.random.randint(1,1 + num_teams,num_data)

    while len(opp_team[opp_team == team]) > 0:
        opp_team[opp_team == team] = np.random.randint(1,1 + num_teams,len(opp_team[opp_team == team]))

    results = np.round(np.random.rand(100))    

    return team, opp_team, results

def reorder_team(team, opp_team, result):

    if team > opp_team:
        team, opp_team = opp_team, team
        result = 1 - result

    return team, opp_team, result

# Generate data and get summary by team match-ups
team, opp_team, results = generate_data()

df = pd.DataFrame(data={'team':team,'opp_team':opp_team,'results':results}, dtype='int')
df = pd.DataFrame(df.apply(lambda x : reorder_team(x['team'], x['opp_team'], x['results']),axis=1).to_list(), 
                  columns=df.columns)
df[['team','opp_team']] = 'Team ' + df[['team','opp_team']].astype(str)

df_summary = df.groupby(['team','opp_team']).agg(['sum','count'])
df_summary.columns = ['wins', 'total']
df_summary.reset_index(inplace=True)
df_summary['team_winrate'] = (df_summary['wins'] / df_summary['total'])
df_summary['opp_team_winrate'] = 1 - df_summary['team_winrate']

产生：

您可以使用以下脚本绘制它们（或使用您喜欢的库编写自己的脚本）：

fig, (ax_count, ax_win) = plt.subplots(1,2, figsize=(12,6))

y_locs = list(range(len(df_summary)))

ax_count.barh(y_locs, width=df_summary['total'], color='tab:gray')
ax_count.set_yticks(y_locs)
ax_count.set_yticklabels(df_summary['team'] + ' VS ' + df_summary['opp_team'])
ax_count.set_title('Total No. of Match Ups')
ax_count.set_xticks([])

for loc in ['top','left','right','bottom']:
    ax_count.spines[loc].set_visible(False)

for p in ax_count.patches:
    ax_count.annotate(f'{p.get_width()}',
                      (p.get_x() + p.get_width(), p.get_y() + p.get_height()/2.), 
                      ha='right', va='center', xytext=(-5,0), textcoords='offset points', 
                      color='white',fontweight='heavy')

ax_win.barh(y_locs, width=df_summary['team_winrate'], color='tab:blue')
ax_win2 = ax_win.twinx()
ax_win2.barh(y_locs, width=df_summary['opp_team_winrate'], 
             left=df_summary['team_winrate'], color='tab:red')

ax_win.set_yticks(y_locs)
ax_win.set_yticklabels(df_summary['team'])
ax_win2.set_yticks(y_locs)
ax_win2.set_yticklabels(df_summary['opp_team'])

ax_win.set_xlim(0,1)
ax_win.set_title('Winning Rate')
ax_win.set_xticks([])

for loc in ['top','left','right','bottom']:
    ax_win.spines[loc].set_visible(False)
    ax_win2.spines[loc].set_visible(False)

for p in ax_win.patches:
    ax_win.annotate(f'{p.get_width() * 100 :.0f} %',
                      (0, p.get_y() + p.get_height()/2.), 
                      ha='left', va='center', xytext=(10,0), textcoords='offset points', 
                      color='white',fontweight='heavy')

    ax_win2.annotate(f'{(1 - p.get_width()) * 100 :.0f} %',
                      (1, p.get_y() + p.get_height()/2.), 
                      ha='right', va='center', xytext=(-10,0), textcoords='offset points', 
                      color='white',fontweight='heavy')

plt.show()