用每个独特团队获胜的顺序计数器创建列表

Question

Say I have a dataset that contains the home_team, away_team and columns home_win, away_win that tells which team won the game. Like this:

Home_team     Away_Team     Home_Win     Away_Win    gameID
   TB            CLB            1            0         1
   NY            ARZ            0            1         2
   EDM           CAN            1            0         3
   NY            TB             0            1         4
   NY            CLB            1            0         5
   TB            NY             1            0         6

How do you write a sequential counter that counts a Teams Total Wins with respect for previous games and irrespective if the Team was Home or Away. So for gameID:1, each team has a total of 0 total wins. Since TB won the first game they now have a total of 1 wins coming up to their second game agains NY(gameID:4) and NY has a total of 0 previous wins.

So the data would look like this: (AT=Away_Team, HT=Home_Team)

Home_team     Away_Team     Home_Win     Away_Win    gameID    HT'sTotWins      AT'sTotWins
   TB            CLB            1            0         1            0               0
   NY            ARZ            0            1         2            0               0
   EDM           CAN            1            0         3            0               0
   NY            TB             0            1         4            0               1
   NY            CLB            1            0         5            0               0
   TB            NY             1            0         6            2               1

I've read some about GroupBy.cumcount() , but I don't know how to write the conditions. I hope I'm not to unclear about what I want to do, if I am please tell me.

Answer 1

To be more instructive, I extended your source data to 10 games and "shortened" column names to make the printout not so wide.

So the first part of the script, generating source DataFrame is as follows:

import pandas as pd

# Source data
df = pd.DataFrame(data=[
    [ 1, 'TB',  'CLB', 1], [ 2, 'NY',  'ARZ', 0],
    [ 3, 'EDM', 'CAN', 1], [ 4, 'NY',  'TB',  0],
    [ 5, 'NY',  'CLB', 1], [ 6, 'TB',  'NY',  1],
    [ 7, 'ARZ', 'CAN', 1], [ 8, 'ARZ', 'TB',  0],
    [ 9, 'NY',  'EDM', 1], [10, 'TB',  'CAN', 1]],
    columns=['gameID', 'HomeTeam', 'AwayTeam', 'HomeWin']).set_index('gameID')
df['AwayWin'] = 1 - df['HomeWin']

Because winning team can be in both HomeTeam and AwayTeam , there is no simple way to use a single groupby . You have to use it twice, generating each result column.

To generate HTWins (Home Teams's Total wins), use:

hWin = df.HomeTeam.where(df.HomeWin == 1, df.AwayTeam)
hCnt = hWin.groupby(hWin).cumcount()
df['HTWins'] = hCnt.where(df.HomeWin == 1, 0)

And to generate ATWins (Away Teams's Total wins), use:

aWin = df.AwayTeam.where(df.AwayWin == 1, df.HomeTeam)
aCnt = aWin.groupby(aWin).cumcount()
df['ATWins'] = aCnt.where(df.AwayWin == 1, 0)

When you print(df) , you will get:

       HomeTeam AwayTeam  HomeWin  AwayWin  HTWins  ATWins
gameID                                                    
1            TB      CLB        1        0       0       0
2            NY      ARZ        0        1       0       0
3           EDM      CAN        1        0       0       0
4            NY       TB        0        1       0       1
5            NY      CLB        1        0       0       0
6            TB       NY        1        0       2       0
7           ARZ      CAN        1        0       1       0
8           ARZ       TB        0        1       0       3
9            NY      EDM        1        0       1       0
10           TB      CAN        1        0       4       0

To assist in understanding how this script works, run each instruction separately and print the result.

Answer 2

There might be a more "elegant" pandas way of doing this, but I would just break things into for loops and go that way.

import copy
import pandas as pd

df = pd.read_csv('sports_data.csv', header=0, delim_whitespace=True)
df["HT'sTotWins"] = 0
df["AT'sTotWins"] = 0

homeWinsAwayWins = {}
homeAwayCount = {'home':0, 'away':0}

for index, row in df.iterrows():
    homeTeam = row['Home_team']
    awayTeam = row['Away_Team']

    if homeTeam not in homeWinsAwayWins:
        homeWinsAwayWins[homeTeam] = copy.deepcopy(homeAwayCount)
    if awayTeam not in homeWinsAwayWins:
        homeWinsAwayWins[awayTeam] = copy.deepcopy(homeAwayCount)

    df.loc[index,"HT'sTotWins"] = homeWinsAwayWins[homeTeam]['home'] + homeWinsAwayWins[homeTeam]['away']
    df.loc[index,"AT'sTotWins"] = homeWinsAwayWins[awayTeam]['home'] + homeWinsAwayWins[awayTeam]['away']

    homeWin = row['Home_Win']
    awayWin = row['Away_Win']
    if homeWin:
        homeWinsAwayWins[homeTeam]['home'] += 1
    elif awayWin:
        homeWinsAwayWins[awayTeam]['away'] += 1

print(df)

It prints what you want.