mavericks.py

# -*- coding: utf-8 -*-
"""Mavericks.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1JjesiHM2DVV5BE6xS170mjvhIxIer8pg

# AWS Maverick Dataset

Retrieval and Instantiation
"""

import os

requisite='boto3'
os.system(f"pip install {requisite}")

import boto3
import io
import pandas as pd

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

# Set Buffer
buffer_pbp = io.BytesIO()
buffer_players = io.BytesIO()

# Create connection to S3
s3 = boto3.resource('s3', aws_access_key_id = 'AKIAWNNDBSXELJDB2NPI', aws_secret_access_key = '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')

# Read Players Data from S3
players_object = s3.Object('utd-hackathon', 'game_players.parquet')
players_object.download_fileobj(buffer_players)
df_players = pd.read_parquet(buffer_players)

print(df_players.head(5))
print(df_players.columns)
print(df_players.shape)

# Extract the following data from the dataFrame
df_players = df_players[["season", "name", "fgm", "fga"]] # TODO: expand this
print(df_players.head(5))

# Merging the data so find the total stats a player has in a season and specific season type
aggregate_functions = {"fgm": "sum", "fga" : "sum"}
df_players = df_players.groupby(['season', "name"]).agg(aggregate_functions).reset_index()
print(df_players.head(5))
print(df_players.tail(5))
print(pd.unique(df_players['season']))

"""# Kaggle Dataset

Instantiation and Visualization
"""

df_mvp = pd.read_csv("mvp_votings.csv", index_col=0)
print(df_mvp.columns)
# Example of renaming the column. Don't rename here tho, rename at the end when everything is finished
#column_order = ['field goal attempts(fga)', 'field goal 3 pointer attempts(fg3a)', 'free throw attempts(fta)', 'player efficiency rating(per)', 'true shooting percentage(ts_pct)', 'usage percentage(usg_pct)', 'box plus-minus(bpm)', 'season', 'player', 'win_pct', 'votes_first', 'points_won', 'points_max', 'award_share','g', 'mp_per_g', 'pts_per_g', 'total rebound percentage(trb_per_g)', 'ast_per_g', 'stl_per_g', 'blk_per_g', 'fg_pct', 'fg3_pct', 'ft_pct', 'win shares(ws)', 'ws_per_48']
#df_mvp.columns = column_order
print(df_mvp.head(10))

import matplotlib.pyplot as plt
import seaborn as sns
plt.figure(figsize=(20,10))
df_mvp_visual = df_mvp
df_mvp_visual['Beat Threshold']=abs(df_mvp_visual['award_share']) > 0.5 #maybe adjust threshold
def scatter(attribute):
    p1=sns.lmplot(x= attribute, y="award_share", data= df_mvp_visual, hue='Beat Threshold', fit_reg=False,height=8,aspect=4)
    ax = p1.axes[0,0]
    plt.show()
scatter("fga") # Obvious trend
scatter("fg3a") # No obvious trend
scatter("fta") # Somewhat obvious trend
#scatter("per") # Obvious trend
#scatter("ts_pct") # Obvious trend
#scatter("usg_pct") # Obvious trend
#scatter("bpm") # Obvious trend
#scatter("win_pct") # Obvious trend
#scatter("g") # Obvious trend
#scatter("mp_per_g") # Obvious trend
#scatter("pts_per_g") # Obvious trend
#scatter("trb_per_g") # No Obvious trend
#scatter("ast_per_g") # No Obvious trend
#scatter("stl_per_g") # No Obvious trend
#scatter("blk_per_g") # No Obvious trend
#scatter("fg_pct") # Obvious trend
#scatter("fg3_pct") # Obvious trend (?)
#scatter("ft_pct") # Obvious trend
#scatter("ws") # Obvious trend
#scatter("ws_per_48") # Obvious trend

df_mvp['MVP'] = "No"
for season in df_mvp['season'].unique():
  # Isolate the season's dataframe
  season_df=df_mvp.loc[df_mvp['season'] == season]
  # Find the 3 largest award share's player of that season
  list_MVP = season_df["award_share"].nlargest(3)
  # Annotate the MVP position for the player
  df_mvp['MVP'][list_MVP.index[0]] = "First"
  df_mvp['MVP'][list_MVP.index[1]] = "Second"
  df_mvp['MVP'][list_MVP.index[2]] = "Third"

import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GroupShuffleSplit
from sklearn.ensemble import RandomForestRegressor
from sklearn.preprocessing import StandardScaler
'''
features = ['fga', 'fta', 'per', 'ts_pct', 'usg_pct', 'bpm', 'win_pct', 'g', 'mp_per_g', 'pts_per_g', 'fg_pct', 'fg3_pct', 'ft_pct', 'ws', 'ws_per_48']

test_columns = df_mvp.pop('award_share')
#gs = GroupShuffleSplit(n_splits=2, train_size=.7, random_state=665)

# split the data set into training and test
train_ix, test_ix = next(gs.split(df_mvp, test_columns, groups=df_mvp.season))
X_train = df_mvp.loc[train_ix][features]
y_train = test_columns.loc[train_ix]

X_test = df_mvp.loc[test_ix][features]
y_test = test_columns.loc[test_ix]
'''
X = df_mvp[['fga', 'fta', 'per', 'ts_pct', 'usg_pct', 'bpm', 'win_pct', 'g', 'mp_per_g', 'pts_per_g', 'fg_pct', 'fg3_pct', 'ft_pct', 'ws', 'ws_per_48']]

y = df_mvp['award_share']


# split the data set into training and test
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=665)

# Standardize the dataset
sc = StandardScaler()
X_train_std = sc.fit_transform(X_train)
X_test_std = sc.transform(X_test)

# Train a random forest regressor
rf = RandomForestRegressor(n_estimators=100, max_depth=5, random_state=665)
rf.fit(X_train_std, y_train)

# Make predictions on the test set
y_pred = rf.predict(X_test_std)

# Evaluate the model
'''
mse = mean_squared_error(y_test, y_pred)
rmse = np.sqrt(mse)
r2 = r2_score(y_test, y_pred)
'''

from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import make_pipeline
from sklearn.decomposition import PCA
# Standardize the dataset
sc = StandardScaler()
X_train_std = sc.fit_transform(X_train)
X_test_std = sc.transform(X_test)

from sklearn.tree import DecisionTreeRegressor
basic_model = DecisionTreeRegressor(random_state=1)
basic_model.fit(X_train_std, y_train)
predictions=basic_model.predict(X_test_std)

df=pd.DataFrame(X_test)
df['prediction']=predictions
df['award_share']=y_test
df['season']=[df_mvp['season'][index] for index in df.reset_index()['index']]
df['player']=[df_mvp['player'][index] for index in df.reset_index()['index']]
df['MVP']=[df_mvp['MVP'][index] for index in df.reset_index()['index']]
X = ['fga', 'fta', 'per', 'ts_pct', 'usg_pct', 'bpm', 'win_pct', 'g', 'mp_per_g', 'pts_per_g', 'fg_pct', 'fg3_pct', 'ft_pct', 'ws', 'ws_per_48']
df=df[X + ['player','season','award_share','prediction', 'MVP']]

df['MVP_prediction'] = "No"
for season in df['season'].unique():
  # Isolate the season's dataframe
  season_df=df.loc[df['season'] == season]
  # Find the 3 largest award share's player of that season
  list_MVP = season_df["prediction"].nlargest(3)
  # Annotate the MVP position for the player
  df['MVP_prediction'][list_MVP.index[0]] = "First"
  df['MVP_prediction'][list_MVP.index[1]] = "Second"
  df['MVP_prediction'][list_MVP.index[2]] = "Third"

print(df.head(10))

correct = 0
total = 0
for season in df['season'].unique():
  print('Season ' + season + ' ---------------------------------------- ')
  # Isolate the season's dataframe
  season_df=df.loc[df['season'] == season]
  season_df=season_df[['player', 'award_share', 'MVP', 'prediction', 'MVP_prediction']]
  first_place = season_df.loc[season_df['MVP_prediction'] == 'First']
  total += 1
  print(first_place)
  #print(first_place['MVP'].item())
  #print(first_place['MVP_prediction'].item())
  if(first_place['MVP'].item() == first_place['MVP_prediction'].item()):
    print(' Correctly determine MVP which is ' + first_place['player'].item())
    print('   Predicted award share of ' + str(first_place['prediction'].item()))
    print('   Actual award share of ' + str(first_place['award_share'].item()))
    correct += 1
  else:
    print(' Incorrectly determine MVP which is ' + first_place['player'].item())
    print('   Predicted award share of ' + str(first_place['prediction'].item()))
    print('   Actual award share of ' + str(first_place['award_share'].item()))
    true_first_place = season_df.loc[season_df['MVP'] == 'First']
    print(' Actual MVP is ' + true_first_place['player'].item())
    print('   Predicted award share of ' + str(true_first_place['prediction'].item()))
    print('   Actual award share of ' + str(true_first_place['award_share'].item()))
print('====================================')
print('Accuracy is ' + str(correct) + ' out of ' + str(total) + ", or " + str(correct/total))