计算由两列过滤的平均数据框

Question

我有这个数据框

    Unnamed: 0  Datetime    HomeTeam    AwayTeam    Ball PossessionMatch_H  Ball PossessionMatch_A
0   0   2021-05-24 02:30:00 U. De Chile Everton                         68      32
1   1   2021-05-23 21:00:00 Huachipato  Colo Colo                       48      52
2   2   2021-05-23 18:30:00 Melipilla   Antofagasta                     47      53
3   3   2021-05-23 02:30:00 U. Espanola U. Catolica                     37      63
4   4   2021-05-23 00:00:00 S. Wanderers    O'Higgins                   29      71
... ... ... ... ... ... ...
57  57  2021-03-28 15:45:00 Palestino   Antofagasta                     58      42
58  58  2021-03-28 01:00:00 U. Espanola S. Wanderers                    50      50
59  59  2021-03-27 22:30:00 Colo Colo   Union La Calera                 58      42
60  60  2021-03-27 20:00:00 Everton O'Higgins                           54      46
61  61  2021-03-27 15:00:00 Curico Unido    Melipilla                   41      59

我想将其拆分为多个数据帧并在“HomeTeam”和“AwayTeam”中应用两个标准，然后计算 Ball Possession 的平均值并将其放入新列“Ball PossessionMatch_H/MP”（如果球队在） HomeTeam”和“Ball PossessionMatch_A/MP”，如果球队在“AwayTeam”中

代码：

teams = pd.unique(df[['HomeTeam', 'AwayTeam']].values.ravel('K'))
df_list3 = []
for team in teams:
    if team in df['HomeTeam']:
        current_df = df[(df.HomeTeam == team) | (df.AwayTeam == team)].copy()
        current_df["Ball PossessionMatch_H/MP"] = current_df.apply(
            lambda x: (
                current_df[(current_df.HomeTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_H"].sum()
                + current_df[(current_df.AwayTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_A"].sum()
            )
            / (
                current_df[(current_df.HomeTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_H"].size
                + current_df[(current_df.AwayTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_A"].size
            ),
            axis=1,
        )
    else:
        current_df = df[(df.HomeTeam == team) | (df.AwayTeam == team)].copy()
        current_df["Ball PossessionMatch_A/MP"] = current_df.apply(
            lambda x: (
                current_df[(current_df.HomeTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_H"].sum()
                + current_df[(current_df.AwayTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_A"].sum()
            )
            / (
                current_df[(current_df.HomeTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_H"].size
                + current_df[(current_df.AwayTeam == team)].loc[x.name + 1 :, "Ball PossessionMatch_A"].size
            ),
            axis=1,
        )
        df_list3.append(current_df)

df = pd.concat(df_list3)
print(df)

它不考虑if 并仅创建具有计算值的“Ball PossessionMatch_A/MP”列并复制行，我会将append 的结果添加到原始数据框。

预期结果：

    Unnamed: 0  Datetime    HomeTeam    AwayTeam    Ball PossessionMatch_H  Ball PossessionMatch_A  Ball PossessionMatch_H/MP   Ball PossessionMatch_A/MP
0   0   2021-05-24 02:30:00 U. De Chile Everton               68           32         50.33      43.5

Answer 1

下面的呢？

hometeam_count = df.groupby("HomeTeam")["Ball PossessionMatch_H"].count()
hometeam_sum = df.groupby("HomeTeam")["Ball PossessionMatch_H"].sum()
awayteam_count = df.groupby("AwayTeam")["Ball PossessionMatch_A"].count()
awayteam_sum = df.groupby("AwayTeam")["Ball PossessionMatch_A"].sum()

df["Ball PossessionMatch_H/MP"] = df["HomeTeam"].apply(lambda x: ((hometeam_sum.loc[x] if x in hometeam_sum.index else 0) + (awayteam_sum.loc[x] if x in awayteam_sum.index else 0)) / ((hometeam_count.loc[x] if x in hometeam_count.index else 0)  + (awayteam_count.loc[x] if x in awayteam_count.index else 0)))
df["Ball PossessionMatch_A/MP"] = df["AwayTeam"].apply(lambda x: ((hometeam_sum.loc[x] if x in hometeam_sum.index else 0) + (awayteam_sum.loc[x] if x in awayteam_sum.index else 0)) / ((hometeam_count.loc[x] if x in hometeam_count.index else 0)  + (awayteam_count.loc[x] if x in awayteam_count.index else 0)))

上面的代码利用groupby 函数分别计算每个团队的sum 和count 以及Ball PossessionMatch_A 和Ball PossessionMatch_H。
代码的第二部分基本上分别查找（loc）每个HomeTeam 和AwayTeam 对应的sum 和count 值，并按预期执行平均计算。

如您所知，HomeTeam 和 AwayTeam 的代码几乎相同，它们也将被清理，以便我们 Dont' repeat yourself。我这样做纯粹是为了便于阅读。

虽然这并不能解决问题（至于为什么没有在您创建的列中填写值），但我提供了该问题的替代解决方案，我认为该解决方案更直观且更具可读性。

Answer 2

免责声明：只是帮助您朝着某个方向前进，不提供复制/粘贴解决方案。

出于演示目的，我对您的数据框进行了一些编辑：

import pandas as pd

df = pd.DataFrame(
    {
     "Datetime": [pd.Timestamp(x) for x in ["2021-05-24 02:30:00", "2021-05-23 21:00:00", 
                                            "2021-05-23 18:30:00", "2021-05-23 02:30:00", 
                                            "2021-05-23 00:00:00"]],
     "HomeTeam": ["Huachipato", "Huachipato", "Huachipato", "U. Espanola", "U. Espanola"],
     "AwayTeam": ["Everton", "Colo Colo", "Antofagasta", "Everton",  "Colo Colo"],
     "Ball PossessionMatch_A": [68, 48, 47, 37, 29],
    }     
)


print(df.groupby("HomeTeam")["Ball PossessionMatch_A"].mean())

输出是：

HomeTeam
Huachipato     54.333333
U. Espanola    33.000000
Name: Ball PossessionMatch_A, dtype: float64

现在，如果您需要过滤一些其他信息，比如对于给定的“AwayTeam”，例如“Everton”，您可以：

print(df[df["AwayTeam"].eq("Everton")].groupby("HomeTeam")["Ball PossessionMatch_A"].mean())

然后你会得到：

HomeTeam
Huachipato     68
U. Espanola    37
Name: Ball PossessionMatch_A, dtype: int64