. Eda Project

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

df=pd.read_csv('./data/fifa_eda.csv')
df.head()

# lets check the percentage null values in each column 
df.info()

# lets check if newely joined player have a higher release clause or old player have a high release clause 
tmp_df=df.groupby(['Joined'])['Release Clause'].mean().reset_index()
plt.figure(figsize=(20,8))
plt.plot(tmp_df['Joined'],tmp_df['Release Clause'])
plt.grid()
plt.xlabel('Joined')
plt.ylabel('Release Clause')  # Player who have join in 2004 have high release clause

# Top 10 players with highest wage 
df.sort_values(by='Wage',ascending=False)[['Name','Wage']].head(10)

# % of left foot players 
print(df['Preferred Foot'].unique())
len(df[df['Preferred Foot']=='Left'])*100/len(df)

# lets now check international reputation , skill moves vs Release clause 
tmp_df1=df.groupby(['International Reputation'])['Release Clause'].mean().reset_index()
tmp_df2=df.groupby(['Skill Moves'])['Release Clause'].mean().reset_index()
plt.figure(figsize=(20,8))
plt.subplot(2,1,1)
plt.barh(y=tmp_df1['International Reputation'],width=tmp_df1['Release Clause'])
plt.xlabel('Release Clause')
plt.ylabel('International Reputation')
plt.subplot(2,1,2)
plt.barh(y=tmp_df2['Skill Moves'],width=tmp_df2['Release Clause'])
plt.xlabel('Release Clause')
plt.ylabel('Skill Moves')

# idetinfy all the contineous features and draw a histogram for the same.
# hint if no of unique values are more then 20 consider that feature as contineous
cont_col=[x for x in df.columns if len(df[x].unique())>20 and df[x].dtype!='O']
cont_col.remove('ID')

for col in cont_col:
    sns.histplot(df[col])
    plt.show()

# oldest player in the dataset
df.sort_values(by=['Age'],ascending=False)[['Age','Name']]

# Total distinct clubs in the data 
len(df['Club'].unique())

# Country wise age distribution of players , only for top 10 countries with highest players 
# first we will find out top 10 countries with highest number of player , since 1 row represents 1 player 
top_10_countries=df.groupby(['Nationality']).size().reset_index()
top_10_countries.columns=['Country','Player_Count']
top_10_countries.sort_values(by='Player_Count',inplace=True,ascending=False)
top_10_countries=top_10_countries.head(10)
top_10_countries

# second method 
# top 10 countries with highest number of players
tmp_df=df.groupby(['Nationality']).size().sort_values(ascending=False).reset_index().head(10)
tmp_df.columns=['Country','No of players']

df.describe()

tmp_df=df[df['Nationality'].isin(top_10_countries['Country'])]

plt.figure(figsize=(20,8))
sns.boxplot(x=tmp_df['Nationality'],y=tmp_df['Age'])

df[df['Wage']<=9].shape[0]/df.shape[0]
import warnings 
warnings.filterwarnings('ignore')
# 1 is 25th quantile value 
# 3 is 50 th quantile

for col in contineous_col:
    plt.figure(figsize=(20,8))
    plt.title(f'Box plot for col {col}')
    sns.boxplot(df[col])
    plt.show()

# is there any correlation between number of player in a club with the average rating of all the player in the club
player_club_rating=df.groupby(['Club']).agg({'ID':'size','Overall':'mean'}).reset_index()
player_club_rating.columns=['Club','no of players','rating']
sns.scatterplot(player_club_rating['no of players'],player_club_rating['rating'])

# does height of the player really effect their rating 
print(df['Height']) # we can see that the height of the player is in decimal upto 6 plays 
# in order to create some buckets of height lets convert this to centimeter and take the int values

# converting height to cm
df['Height_in_cm']=df['Height']*30.48
df['Height_in_cm']

# rounding height 
df['Height_in_cm']=round(df['Height_in_cm']).astype(int)
df['Height_in_cm']

# now we will have atleast few repeated values for each height 
tmp_df=df.groupby(['Height_in_cm'])['Overall'].mean().reset_index()
sns.scatterplot(tmp_df['Height_in_cm'],tmp_df['Overall']) # height does not seem to have much impact of mean rating

# weight vs rating 
tmp_df=df.groupby(['Weight'])['Overall'].mean().reset_index()
sns.scatterplot(tmp_df['Weight'],tmp_df['Overall']) # it has a direct relation with the ratings of player

df