Youngjun Woo

Youngjun Woo

M.S. in Data Science | University of Michigan

2 - 6) EDA - Team information

12 minute read 

import numpy as np 
import pandas as pd 

import matplotlib.pyplot as plt 
import seaborn as sns 
import missingno as msno 
from utils import my_histogram

from sklearn.experimental import enable_iterative_imputer
from sklearn.impute import IterativeImputer

import warnings
warnings.filterwarnings('ignore')

pd.set_option('display.max_columns', None)

df_team = pd.read_csv("../data/df_team.csv")


df_team
team_api_id team_long_name league_name team_measured_date buildUpPlaySpeed buildUpPlaySpeedClass buildUpPlayDribbling buildUpPlayDribblingClass buildUpPlayPassing buildUpPlayPassingClass buildUpPlayPositioningClass chanceCreationPassing chanceCreationPassingClass chanceCreationCrossing chanceCreationCrossingClass chanceCreationShooting chanceCreationShootingClass chanceCreationPositioningClass defencePressure defencePressureClass defenceAggression defenceAggressionClass defenceTeamWidth defenceTeamWidthClass defenceDefenderLineClass
0 9930 FC Aarau Switzerland Super League 2010-02-22 60 Balanced NaN Little 50 Mixed Organised 60 Normal 65 Normal 55 Normal Organised 50 Medium 55 Press 45 Normal Cover
1 9930 FC Aarau Switzerland Super League 2014-09-19 52 Balanced 48.0 Normal 56 Mixed Organised 54 Normal 63 Normal 64 Normal Organised 47 Medium 44 Press 54 Normal Cover
2 9930 FC Aarau Switzerland Super League 2015-09-10 47 Balanced 41.0 Normal 54 Mixed Organised 54 Normal 63 Normal 64 Normal Organised 47 Medium 44 Press 54 Normal Cover
3 8485 Aberdeen Scotland Premier League 2010-02-22 70 Fast NaN Little 70 Long Organised 70 Risky 70 Lots 70 Lots Organised 60 Medium 70 Double 70 Wide Cover
4 8485 Aberdeen Scotland Premier League 2011-02-22 47 Balanced NaN Little 52 Mixed Organised 53 Normal 48 Normal 52 Normal Organised 47 Medium 47 Press 52 Normal Cover
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
1453 10000 SV Zulte-Waregem Belgium Jupiler League 2011-02-22 52 Balanced NaN Little 52 Mixed Organised 52 Normal 48 Normal 53 Normal Organised 46 Medium 48 Press 53 Normal Cover
1454 10000 SV Zulte-Waregem Belgium Jupiler League 2012-02-22 54 Balanced NaN Little 51 Mixed Organised 47 Normal 52 Normal 50 Normal Organised 44 Medium 55 Press 53 Normal Cover
1455 10000 SV Zulte-Waregem Belgium Jupiler League 2013-09-20 54 Balanced NaN Little 51 Mixed Organised 47 Normal 52 Normal 32 Little Organised 44 Medium 58 Press 37 Normal Cover
1456 10000 SV Zulte-Waregem Belgium Jupiler League 2014-09-19 54 Balanced 42.0 Normal 51 Mixed Organised 47 Normal 52 Normal 32 Little Organised 44 Medium 58 Press 37 Normal Cover
1457 10000 SV Zulte-Waregem Belgium Jupiler League 2015-09-10 54 Balanced 42.0 Normal 51 Mixed Organised 47 Normal 52 Normal 32 Little Organised 44 Medium 58 Press 37 Normal Cover

1458 rows × 25 columns

team_num_attr = ["team_api_id", "team_measured_date", "buildUpPlaySpeed", "buildUpPlayDribbling", "buildUpPlayPassing", 
                 "chanceCreationPassing", "chanceCreationCrossing", "chanceCreationShooting", "defencePressure", "defenceAggression",
                 "defenceTeamWidth"]

df_team_num_attr = df_team[team_num_attr]

df_team_num_attr
team_api_id team_measured_date buildUpPlaySpeed buildUpPlayDribbling buildUpPlayPassing chanceCreationPassing chanceCreationCrossing chanceCreationShooting defencePressure defenceAggression defenceTeamWidth
0 9930 2010-02-22 60 NaN 50 60 65 55 50 55 45
1 9930 2014-09-19 52 48.0 56 54 63 64 47 44 54
2 9930 2015-09-10 47 41.0 54 54 63 64 47 44 54
3 8485 2010-02-22 70 NaN 70 70 70 70 60 70 70
4 8485 2011-02-22 47 NaN 52 53 48 52 47 47 52
... ... ... ... ... ... ... ... ... ... ... ...
1453 10000 2011-02-22 52 NaN 52 52 48 53 46 48 53
1454 10000 2012-02-22 54 NaN 51 47 52 50 44 55 53
1455 10000 2013-09-20 54 NaN 51 47 52 32 44 58 37
1456 10000 2014-09-19 54 42.0 51 47 52 32 44 58 37
1457 10000 2015-09-10 54 42.0 51 47 52 32 44 58 37

1458 rows × 11 columns

df_team_num_attr.isna().sum()

team_api_id                 0
team_measured_date          0
buildUpPlaySpeed            0
buildUpPlayDribbling      969
buildUpPlayPassing          0
chanceCreationPassing       0
chanceCreationCrossing      0
chanceCreationShooting      0
defencePressure             0
defenceAggression           0
defenceTeamWidth            0
dtype: int64
  • Since there are missing values in only the buildUpPlayDribbling column, let’s just drop the column.
df_team_num_attr.drop("buildUpPlayDribbling", axis = 1, inplace = True)
  • Since there are only 2 teams in each match, there will be total 2 teams x 10 team attributes = 20 columns for each match.
  • So the original team attributes columns will be used for modeling unlike the player attribute columns that will be used after dimensionality reduction by PCA.
team_num_collected = pd.DataFrame(df_team_num_attr.groupby("team_api_id").count().team_measured_date)
team_num_collected.columns = ["num_collected"]
team_num_collected
num_collected
team_api_id
1601 6
1773 3
1957 6
2033 5
2182 6
... ...
10281 6
108893 6
158085 2
208931 2
274581 1

288 rows × 1 columns

team_num_collected.describe()
num_collected
count 288.000000
mean 5.062500
std 1.540255
min 1.000000
25% 5.000000
50% 6.000000
75% 6.000000
max 6.000000 
my_histogram(team_num_collected, "num_collected")

png

  • For most teams, team attributes were collected 5 to 6 times.
  • So, we will use team attributes collected closest to match dates.
df_match_basic = pd.read_csv("../data/df_match_basic.csv")

df_match_team_num_attr = df_match_basic[["match_api_id", "match_date", "home_team_api_id", "away_team_api_id"]]

home_team_info = df_match_team_num_attr.merge(df_team_num_attr, how = "left", left_on = "home_team_api_id", right_on = "team_api_id") \
                                       .sort_values(["team_api_id", "match_date", "team_measured_date"])

home_team_info["date_diff"] = (pd.to_datetime(home_team_info.team_measured_date) - pd.to_datetime(home_team_info.match_date)).dt.days
home_team_info = home_team_info[home_team_info.date_diff < 0]

home_team_info = home_team_info.sort_values(["team_api_id", "match_date", "date_diff"], ascending = [True, True, False]).groupby(["team_api_id", "match_date"]).first()

home_team_info = home_team_info.drop(["home_team_api_id", "away_team_api_id", "team_measured_date", "date_diff"], axis = 1) \
                    .rename(columns = {"buildUpPlaySpeed": "home_team_buildUpPlaySpeed", "buildUpPlayPassing": "home_team_buildUpPlayPassing",
                                       "chanceCreationPassing": "home_team_chanceCreationPassing", "chanceCreationCrossing": "home_team_chanceCreationCrossing",
                                       "chanceCreationShooting": "home_team_chanceCreationShooting", "defencePressure": "home_team_defencePressure", 
                                       "defenceAggression": "home_team_defenceAggression", "defenceTeamWidth": "home_team_defenceTeamWidth"})

away_team_info = df_match_team_num_attr.merge(df_team_num_attr, how = "left", left_on = "away_team_api_id", right_on = "team_api_id") \
                                       .sort_values(["team_api_id", "match_date", "team_measured_date"])

away_team_info["date_diff"] = (pd.to_datetime(away_team_info.team_measured_date) - pd.to_datetime(away_team_info.match_date)).dt.days
away_team_info = away_team_info[away_team_info.date_diff < 0]

away_team_info = away_team_info.sort_values(["team_api_id", "match_date", "date_diff"], ascending = [True, True, False]).groupby(["team_api_id", "match_date"]).first()

away_team_info = away_team_info.drop(["home_team_api_id", "away_team_api_id", "team_measured_date", "date_diff"], axis = 1) \
                    .rename(columns = {"buildUpPlaySpeed": "away_team_buildUpPlaySpeed", "buildUpPlayPassing": "away_team_buildUpPlayPassing",
                                       "chanceCreationPassing": "away_team_chanceCreationPassing", "chanceCreationCrossing": "away_team_chanceCreationCrossing",
                                       "chanceCreationShooting": "away_team_chanceCreationShooting", "defencePressure": "away_team_defencePressure", 
                                       "defenceAggression": "away_team_defenceAggression", "defenceTeamWidth": "away_team_defenceTeamWidth"})

home_team_info
match_api_id home_team_buildUpPlaySpeed home_team_buildUpPlayPassing home_team_chanceCreationPassing home_team_chanceCreationCrossing home_team_chanceCreationShooting home_team_defencePressure home_team_defenceAggression home_team_defenceTeamWidth
team_api_id match_date
1601.0 2010-03-06 674613 30.0 40.0 50.0 35.0 70.0 65.0 60.0 50.0
2010-03-19 674631 30.0 40.0 50.0 35.0 70.0 65.0 60.0 50.0
2010-04-03 674641 30.0 40.0 50.0 35.0 70.0 65.0 60.0 50.0
2010-04-20 674839 30.0 40.0 50.0 35.0 70.0 65.0 60.0 50.0
2010-04-28 674854 30.0 40.0 50.0 35.0 70.0 65.0 60.0 50.0
... ... ... ... ... ... ... ... ... ... ...
274581.0 2015-12-22 1979948 50.0 50.0 50.0 50.0 50.0 45.0 45.0 50.0
2016-01-23 1980004 50.0 50.0 50.0 50.0 50.0 45.0 45.0 50.0
2016-02-06 1980020 50.0 50.0 50.0 50.0 50.0 45.0 45.0 50.0
2016-02-20 1980036 50.0 50.0 50.0 50.0 50.0 45.0 45.0 50.0
2016-03-05 1980052 50.0 50.0 50.0 50.0 50.0 45.0 45.0 50.0

19913 rows × 9 columns

df_match_team_num_attr = df_match_team_num_attr.merge(home_team_info, how = "left", on = "match_api_id") \
                                               .merge(away_team_info, how = "left", on = "match_api_id") \
                                               .drop(["match_date", "home_team_api_id", "away_team_api_id"], axis = 1)

df_match_team_num_attr
match_api_id home_team_buildUpPlaySpeed home_team_buildUpPlayPassing home_team_chanceCreationPassing home_team_chanceCreationCrossing home_team_chanceCreationShooting home_team_defencePressure home_team_defenceAggression home_team_defenceTeamWidth away_team_buildUpPlaySpeed away_team_buildUpPlayPassing away_team_chanceCreationPassing away_team_chanceCreationCrossing away_team_chanceCreationShooting away_team_defencePressure away_team_defenceAggression away_team_defenceTeamWidth
0 492473 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
1 492474 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
2 492475 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
3 492476 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
4 492477 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
25974 1992091 52.0 48.0 39.0 39.0 37.0 40.0 36.0 52.0 58.0 35.0 64.0 36.0 66.0 47.0 45.0 53.0
25975 1992092 53.0 56.0 38.0 53.0 46.0 42.0 33.0 58.0 63.0 51.0 54.0 40.0 66.0 48.0 53.0 50.0
25976 1992093 50.0 54.0 52.0 54.0 46.0 36.0 43.0 47.0 61.0 37.0 62.0 50.0 53.0 45.0 47.0 54.0
25977 1992094 NaN NaN NaN NaN NaN NaN NaN NaN 62.0 46.0 47.0 50.0 54.0 47.0 43.0 56.0
25978 1992095 52.0 64.0 39.0 66.0 46.0 44.0 34.0 50.0 61.0 46.0 66.0 66.0 54.0 47.0 58.0 53.0

25979 rows × 17 columns

  • Save the table.
df_match_team_num_attr.to_csv("../data/df_match_team_num_attr.csv", index = False)