How to make new columns out of every second row in a pandas df

user_13143177 0
Viewed 178

I have a data frame for NBA data that I am having a hard time manipulating. I would like to change df1 to df2 by having both teams and their scores in a game along the same row twice to resemble the games outcome from both teams' standpoints:

df1

GameID     TeamID     TeamAbb     PTS
   0        1001        TOR        99
   0        1023        ATL        86
   1        1004        DAL        102
   1        1003        POR        100
   2        1015        LAL        96
   2        1029        MIL        85

df2

GameID     Team1ID     Team2ID     Team1Abb      Team2Abb    Team1PTS    Team2PTS
   0        1001        1023         TOR           ATL          99          86
   0        1023        1001         ATL           TOR          86          99
   1        1004        1003         DAL           POR          102         100
   1        1003        1004         POR           DAL          100         102

So in essence, a sort of widening of the data frame.

pandas data-cleaning data-manipulation python
1 Answers

Try:

df2 = df1.set_index(['GameID', df1.groupby('GameID').cumcount()+1]).unstack()
df2.columns=[f'{i}_{j}' for i, j in df2.columns]
df2.reset_index()

Output:

  GameID  TeamID_1  TeamID_2 TeamAbb_1 TeamAbb_2  PTS_1  PTS_2
0       0      1001      1023       TOR       ATL     99     86
1       1      1004      1003       DAL       POR    102    100
2       2      1015      1029       LAL       MIL     96     85

Details:

  • Use 'GameID' to groupby and cumcount getting 1 and 2.
  • Then, flatten multiindex column headers created by groupby using list comprehension
  • Lastly, reset_index

Update per comment below:

#Create home team and visiting team records
g = df.groupby('GameID').cumcount()
dfh = df.set_index(['GameID', g + 1])
dfv = df.set_index(['GameID', 2 - g])

dfh = dfh.unstack()
dfh.columns = [f'{i}_{j}' for i, j in dfh.columns]

dfv = dfv.unstack()
dfv.columns = [f'{i}_{j}' for i, j in dfv.columns]

# concatenate home and visiting records
pd.concat([dfh, dfv]).sort_index().reset_index()

Output:

   GameID  TeamID_1  TeamID_2 TeamAbb_1 TeamAbb_2  PTS_1  PTS_2
0       0      1001      1023       TOR       ATL     99     86
1       0      1023      1001       ATL       TOR     86     99
2       1      1004      1003       DAL       POR    102    100
3       1      1003      1004       POR       DAL    100    102
4       2      1015      1029       LAL       MIL     96     85
5       2      1029      1015       MIL       LAL     85     96
user_63615310
Related