import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import scipy

from collections import Counter

from scipy.stats import ttest_1samp, ttest_ind
from scipy.stats import levene

df = pd.read_csv('games.csv')

df.columns = df.columns.map(lambda x: x.lower())

df = df.dropna(subset=['name']).reset_index(drop=True)
df['year_of_release'] = df.groupby('name')['year_of_release'].transform(lambda x: x.fillna(x.mean()))

df['critic_score'] = df.groupby('name')['critic_score'].transform(lambda x: x.fillna(x.mean()))

df.loc[df.user_score=='tbd', 'user_score'] = np.nan
df.user_score = df.user_score.astype('float')
df['user_score'] = df.groupby('name')['user_score'].transform(lambda x: x.fillna(x.mean()))

# создаем столбец когорт
df['user_score_cat'] = df['user_score'].apply(lambda x: x//1)

# Заполняем пропуски
df.loc[df['user_score_cat']==0, 'critic_score'] = df.loc[df['user_score_cat']==0, 'critic_score'].fillna(52.250000)
df.loc[df['user_score_cat']==1, 'critic_score'] = df.loc[df['user_score_cat']==1, 'critic_score'].fillna(43.340860)
df.loc[df['user_score_cat']==2, 'critic_score'] = df.loc[df['user_score_cat']==2, 'critic_score'].fillna( 48.46782)
df.loc[df['user_score_cat']==3, 'critic_score'] = df.loc[df['user_score_cat']==3, 'critic_score'].fillna( 50.38503)
df.loc[df['user_score_cat']==4, 'critic_score'] = df.loc[df['user_score_cat']==4, 'critic_score'].fillna(56.278505)
df.loc[df['user_score_cat']==5, 'critic_score'] = df.loc[df['user_score_cat']==5, 'critic_score'].fillna( 61.10962)
df.loc[df['user_score_cat']==6, 'critic_score'] = df.loc[df['user_score_cat']==6, 'critic_score'].fillna( 65.62730)
df.loc[df['user_score_cat']==7, 'critic_score'] = df.loc[df['user_score_cat']==7, 'critic_score'].fillna( 71.28083)
df.loc[df['user_score_cat']==8, 'critic_score'] = df.loc[df['user_score_cat']==8, 'critic_score'].fillna( 76.69920)
df.loc[df['user_score_cat']==9, 'critic_score'] = df.loc[df['user_score_cat']==9, 'critic_score'].fillna(  80.4016)

df.drop('user_score_cat', axis='columns', inplace=True)

df['rating'] = df.groupby('name')['rating'].transform(lambda x: x.fillna(Counter(x).most_common()[0][0]))

df.rating.fillna('No rating', inplace=True)

df['year_of_release'] =  df.groupby('platform')['year_of_release'].transform(lambda x: x.fillna(np.mean(x)))

df.year_of_release = df.year_of_release.astype('int')

#суммарные продажи
df['world_sales'] = df['na_sales'] + df['eu_sales'] + df['jp_sales'] + df['other_sales']

#Суммарные продажи у каждой платформы
df.groupby('platform').sum()['world_sales']

most_popular = df.groupby('platform').sum()['world_sales'].sort_values(ascending=False).index.values[:10]

popular_df = df[df.platform.isin(most_popular)]
popular_df = popular_df.pivot_table(index='year_of_release', columns='platform', values='world_sales', aggfunc='sum')
popular_df.plot(figsize=(8,6))

temp = popular_df.T.fillna(0).reset_index()
temp[temp[2016]==0]

temp.set_index('platform',inplace=True)

life_of_platform = temp[temp!=0].count(axis=1)

print('Среднее время жизни платформы:',life_of_platform.drop('PC').mean())

df_striped = df[(df['year_of_release']>=2008)].copy()

df_striped.pivot_table(index='platform',
                                                               columns='year_of_release',
                                                               values='world_sales',
                                                               aggfunc='sum').T.plot(figsize=(10,7))

df_popular = df_striped[df_striped.platform.isin(most_popular)]

df_popular_grouped = df_popular.groupby('platform').sum()
df_popular_grouped

popular_platforms_list = df_popular_grouped.index.values
popular_platforms_list

fig, ax = plt.subplots(nrows=1, ncols=8, sharey=True, figsize=(30,6))

ax[0].scatter(data=df_popular[df_popular.platform=='DS'], x='user_score',y='world_sales')
ax[1].scatter(data=df_popular[df_popular.platform=='PC'], x='user_score',y='world_sales')
ax[2].scatter(data=df_popular[df_popular.platform=='PS2'], x='user_score',y='world_sales')
ax[3].scatter(data=df_popular[df_popular.platform=='PS3'], x='user_score',y='world_sales')
ax[4].scatter(data=df_popular[df_popular.platform=='PS4'], x='user_score',y='world_sales')
ax[5].scatter(data=df_popular[df_popular.platform=='PSP'], x='user_score',y='world_sales')
ax[6].scatter(data=df_popular[df_popular.platform=='Wii'], x='user_score',y='world_sales')
ax[7].scatter(data=df_popular[df_popular.platform=='X360'], x='user_score',y='world_sales')



fig.suptitle('Зависимость между user score и продажами у каждой платформы')
ax[0].set_ylabel('World sales')
for i,e in enumerate(ax[:]):
    e.set_xlabel('User score')
    platform_name = popular_platforms_list[i]
    e.set_title(platform_name)

plt.show()