Do NBA teams with greater team salaries win more?

import matplotlib.pyplot as plt
%matplotlib inline
plt.style.use('fivethirtyeight')

import numpy as np
import pandas as pd
from sklearn import linear_model
from sklearn.linear_model import LinearRegression

# blog html visualization 
from IPython.display import HTML

# plt.rcParams['figure.dpi'] = 50
# %config InlineBackend.figure_format = 'retina'
%config InlineBackend.figure_format = 'svg'

raw=pd.read_csv('./data/nba_team_salary.csv')

# reorganize columns in chronological order
salary = pd.concat([raw["team"], raw[raw.columns[:0:-1]]], axis=1)

# change NaN to zero for no salary expansion teams: Charlotte 2003, 2004; New Orleans 2001, 2002
salary.fillna("$0", inplace=True)

# rename city to mascot name to merge later with games dataset
teams = {"team":["Hawks","Celtics","Nets","Hornets","Bulls","Cavaliers","Mavericks","Nuggets",
                 "Pistons","Warriors","Rockets","Pacers","Clippers","Lakers","Grizzlies","Heat",
                 "Bucks","Timberwolves","Pelicans","Knicks","Thunder","Magic","Sixers","Suns",
                 "Trailblazers","Kings","Spurs","Raptors","Jazz","Wizards"]}

salary = pd.concat([pd.DataFrame(teams),salary.iloc[:,1:16]],axis=1)

# transform wide to long data for plotting
# 450 columns: 30 teams x 15 years
long_salary = pd.melt(salary, id_vars='team', var_name='year', value_name='salary')
long_salary.head()

# year into int
long_salary["year"] = long_salary["year"].apply(lambda x: int(x))

# salary into int
long_salary["salary"] = long_salary["salary"].apply(lambda x: int(x.replace("$","").replace(",","")))

def z_salary(df,year,salary):
    year_mean = df[(df['year']==year)]['salary'].mean()
    year_std = df[(df['year']==year)]['salary'].std()
    return (salary - year_mean)/year_std

long_salary['z_salary'] = long_salary.apply(lambda row: z_salary(long_salary, row["year"], row["salary"]), axis=1)
HTML(long_salary.head().to_html(classes="table table-responsive table-striped table-bordered"))

# export to csv 
long_salary.to_csv("./data/nba_team_salary_final.csv")

games = pd.read_csv('./data/nba_elo.csv')

# I'll only look at the regular season to keep n consistent across teams
games = games[games["is_playoffs"] == 0]

# This dataset has a lot of extraneous information, so I'll focus on the columns of interest. 
games = games[["year_id","date_game","team_id","fran_id","opp_id","opp_fran","game_result"]]

# This dataset is from 1947-2019, so I'll filter for years 2001 to 2015 to match our Salary data
games = games[(games["year_id"] >= 2001) & (games["year_id"] <= 2015)]

HTML(games.head().to_html(classes="table table-responsive table-striped table-bordered"))

# reset record to beginning with list comphrension
record = {team:{"G":0,"W":0,"L":0} for team in teams}

# Each game shows up for the home and away team, so I'll only need to run control flow for when they lose
# Repeat for each year

def create_record(year):
    for i, row in games[games["year_id"]==year].iterrows():
        if row.game_result == "L":
            record[row.fran_id]["G"] += 1
            record[row.opp_fran]["G"] += 1
            record[row.fran_id]["L"] += 1
            record[row.opp_fran]["W"] += 1
    return record

# set up list of each team and each year
teams = sorted(list(games.fran_id.unique()))
all_years = list(games.year_id.unique())

# now repeat for all the years in a dictionary object
seasons = {}
for year in all_years:
    record = {team:{"G":0,"W":0,"L":0} for team in teams}
    seasons[year] = create_record(year)

create_record(2001)

{'Bucks': {'G': 164, 'W': 93, 'L': 71},
 'Bulls': {'G': 164, 'W': 65, 'L': 99},
 'Cavaliers': {'G': 164, 'W': 83, 'L': 81},
 'Celtics': {'G': 164, 'W': 76, 'L': 88},
 'Clippers': {'G': 164, 'W': 87, 'L': 77},
 'Grizzlies': {'G': 164, 'W': 78, 'L': 86},
 'Hawks': {'G': 164, 'W': 85, 'L': 79},
 'Heat': {'G': 164, 'W': 87, 'L': 77},
 'Hornets': {'G': 82, 'W': 33, 'L': 49},
 'Jazz': {'G': 164, 'W': 91, 'L': 73},
 'Kings': {'G': 164, 'W': 84, 'L': 80},
 'Knicks': {'G': 164, 'W': 65, 'L': 99},
 'Lakers': {'G': 164, 'W': 77, 'L': 87},
 'Magic': {'G': 164, 'W': 68, 'L': 96},
 'Mavericks': {'G': 164, 'W': 103, 'L': 61},
 'Nets': {'G': 164, 'W': 64, 'L': 100},
 'Nuggets': {'G': 164, 'W': 70, 'L': 94},
 'Pacers': {'G': 164, 'W': 79, 'L': 85},
 'Pelicans': {'G': 164, 'W': 91, 'L': 73},
 'Pistons': {'G': 164, 'W': 64, 'L': 100},
 'Raptors': {'G': 164, 'W': 96, 'L': 68},
 'Rockets': {'G': 164, 'W': 101, 'L': 63},
 'Sixers': {'G': 164, 'W': 74, 'L': 90},
 'Spurs': {'G': 164, 'W': 113, 'L': 51},
 'Suns': {'G': 164, 'W': 90, 'L': 74},
 'Thunder': {'G': 164, 'W': 89, 'L': 75},
 'Timberwolves': {'G': 164, 'W': 63, 'L': 101},
 'Trailblazers': {'G': 164, 'W': 101, 'L': 63},
 'Warriors': {'G': 164, 'W': 84, 'L': 80},
 'Wizards': {'G': 164, 'W': 65, 'L': 99}}

# rename index column to team
df_seasons = pd.DataFrame(seasons).reset_index().rename(columns={'index':'team'})

# wide to long data for plotting
long_seasons = pd.melt(df_seasons, id_vars='team', var_name='year', value_name='seasons')

# Break out the G,W,L into columns
items = list(long_seasons["seasons"])

long_seasons["G"] = pd.DataFrame([item["G"] for item in items])
long_seasons["L"] = pd.DataFrame([item["L"] for item in items])
long_seasons["W"] = pd.DataFrame([item["W"] for item in items])

# Calculate winning % for each team per year, and add as a new column
long_seasons['win_%'] = round(long_seasons['W']/long_seasons['G'],2)

# drop seasons, redundant column with G, L, W
long_seasons = long_seasons.drop(columns=['seasons'])

# make year into int so it can be merged later
long_seasons["year"] = long_seasons["year"].apply(lambda x: int(x))

HTML(long_seasons.head().to_html(classes="table table-responsive table-striped table-bordered"))

# merge datasets together
final = long_salary.merge(long_seasons, on=["team","year"])
final

# add divisions data
divisions_dict = {"Atlantic": ["Celtics","Nets","Sixers","Raptors","Knicks"],
             "Central": ["Bulls","Cavaliers","Pistons","Bucks","Pacers"],
             "Southeast": ["Hawks","Hornets","Heat","Magic","Wizards"],
             "Northwest": ["Timberwolves","Nuggets","Trailblazers","Thunder","Jazz"],
             "Pacific" : ["Warriors","Clippers","Lakers","Kings","Suns"],
             "Southwest" : ["Mavericks", "Rockets","Grizzlies","Pelicans","Spurs"]}

divisions = ["Atlantic","Central","Southeast","Northwest","Pacific","Southwest"]

def team_to_division(team):
    for division in divisions:
        if team in divisions_dict[division]:
            return division

final["division"] = final["team"].apply(lambda x: team_to_division(x))

# add conference data
conferences = ["East","West"]
conferences_dict = {"East": ["Celtics","Nets","Sixers","Raptors","Knicks","Bulls","Cavaliers","Pistons","Bucks","Pacers","Hawks","Hornets","Heat","Magic","Wizards"],
                   "West": ["Timberwolves","Nuggets","Trailblazers","Thunder","Jazz","Warriors","Clippers","Lakers","Kings","Suns","Mavericks", "Rockets","Grizzlies","Pelicans","Spurs"]}

def division_to_conference(division):
    for conference in conferences:
        if division in conferences_dict[conference]:
            return conference

final["conference"] = final["team"].apply(lambda x: division_to_conference(x))

# Remove early Charlotte Hornet years, missing data;
# they relocated to New Orleans, and were renamed as the Pelicans
final[final["team"] == 'Hornets'].loc[[3,33,63,93]]
final = final.drop([3,33,63,93])

final.to_csv("./data/final.csv")

raptors = final[final['team']== 'Raptors']

# change way we invoke plot so can manipulate the axis
fig, ax = plt.subplots(figsize=(3, 3))

#ax.scatter(x, y)
ax.scatter(x=raptors['z_salary'], y=raptors['win_%'],alpha=0.5,c="#CE1141")
plt.title('Toronto Raptors', position = (0,1), ha = 'left', fontsize=12)
plt.xlabel("Standardized Salaries", position = (0,0), ha = 'left', color = 'grey')
plt.ylabel('Win %', position = (0, 1), ha = 'right', color = 'grey')

# add linear regression line
X = pd.DataFrame(final[final["team"]=="Raptors"]["z_salary"])
Y = pd.DataFrame(raptors["win_%"])
linreg = linear_model.LinearRegression()
model = linreg.fit(X, Y)
predictions = model.predict(X)

plt.plot(X, predictions, color='#CE1141',linewidth=2)

# remove the top and right borders
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['bottom'].set_visible(False)
ax.spines['left'].set_visible(False)

# add in gridlines
ax.grid(color='grey', linestyle='-', linewidth=0.25, alpha=0.5)

# create cross hairs 
plt.hlines(0.5,-3, 3,alpha=0.5) 
plt.vlines(0,0,1,alpha=0.5)

# reduce ticks on x axis
plt.xticks([-2,-1,0,1,2])

plt.show()

# change way we invoke plot so can manipulate the axis
fig, ax = plt.subplots(figsize=(3, 3))

#ax.scatter(x, y)
scatter_league = (ax.scatter(x=final['z_salary'], y=final['win_%'],alpha=0.5,c="#000000"),
                  plt.title('NBA League', position = (0,1), ha = 'left', fontsize=12),
                  plt.xlabel("Standardized Salaries", position = (0,0), ha = 'left', color = 'grey'),
                  plt.ylabel('Win %', position = (0, 1), ha = 'right', color = 'grey'))

# add linear regression line
X = pd.DataFrame(final["z_salary"])
Y_league = pd.DataFrame(final["win_%"])
linreg = linear_model.LinearRegression()
model = linreg.fit(X, Y_league)
predictions_league = model.predict(X)

# remove the top and right borders
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['bottom'].set_visible(False)
ax.spines['left'].set_visible(False)

# add in gridlines
ax.grid(color='grey', linestyle='-', linewidth=0.25, alpha=0.5)

# create cross hairs 
plt.hlines(0.5,-3, 3,alpha=0.5) 
plt.vlines(0,0,1,alpha=0.5)

# reduce ticks on x axis
plt.xticks([-2,-1,0,1,2])

league_plot = plt.plot(X, predictions_league, color='#000000',linewidth=2,alpha=0.25)
plt.show()

# Create a dictionary with the team names, and team plot colour. I'll do this with Atlantic, 
# then loop through each division on each row

atlantic = {"Celtics": "#007A33",
            "Nets": "#000000",
            "Sixers": "#006BB6",
            "Raptors": "#CE1141",
            "Knicks": "#F58426"}

def plot_team_row(team_df, ax,team_color='#00000'):

    # linear regression
    X = pd.DataFrame(team_df["z_salary"])
    Y = pd.DataFrame(team_df["win_%"])
    linreg = linear_model.LinearRegression()
    model = linreg.fit(X, Y)
    predictions = model.predict(X)
    ax.plot(X, predictions, color=team_color,alpha=0.8,linewidth=2)    

    # plot league average
    ax.plot(pd.DataFrame(final['z_salary']), predictions_league, color='#C4CED4',alpha=0.5,linewidth=2);

    # scatterplot
    ax.scatter(x=team_df['z_salary'], y=team_df['win_%'],alpha=0.5,c=team_color,s=10)

    # title for each team subplot
    ax.set_title(f"{team_df['team'].values[0]}",fontsize=10)

    # create cross hairs
    ax.hlines(0.5,-3, 3,alpha=0.5,linewidth=2)
    ax.vlines(0,.10,.90,alpha=0.5,linewidth=2)

    # ticks
    plt.xticks([-2,0,2])

    # remove borders
    ax.spines['top'].set_visible(False)
    ax.spines['right'].set_visible(False)
    ax.spines['bottom'].set_visible(False)
    ax.spines['left'].set_visible(False)

    # add in gridlines
    ax.grid(color='grey', linestyle='-', linewidth=0.25, alpha=0.5)

fig, ax = plt.subplots(nrows=1, ncols=5,figsize=(6,2),sharex=True, sharey=True)

# Division title
plt.figtext(0.48,1, "Atlantic", fontsize=10, color="grey")

# Spacing for subplots
fig.subplots_adjust(hspace = 0.5, wspace=.001, top=0.825)

# pos determines which axes object to pass into the function
#there are 5 pos
col=0 
for team, colour in atlantic.items():
    plot_team_row(final[final['team']==team], ax[col],colour)
    col=col+1 

# force label to appear on first plot rather than the last plot
ax[0].set_xlabel('Standardized Salaries', fontsize=10)
ax[0].set_ylabel('Win %', fontsize=10)
ax[0].tick_params(labelsize=8)
ax[1].tick_params(labelsize=8)
ax[2].tick_params(labelsize=8)
ax[3].tick_params(labelsize=8)
ax[4].tick_params(labelsize=8)

plt.show()

# team colours source: https://teamcolorcodes.com/

atlantic = {"Celtics": "#007A33",
            "Nets": "#000000",
            "Sixers": "#006BB6",
            "Raptors": "#CE1141",
            "Knicks": "#F58426"}

central = {"Bulls": "#CE1141",
            "Cavaliers": "#6F263D",
            "Pistons": "#C8102E",
            "Bucks": "#00471B",
            "Pacers": "#FDBB30"}

southeast = {"Hawks": "#E03A3E",
             "Hornets":"#00788C",
             "Heat": "#98002E",
             "Magic": "#0077C0",
             "Wizards": "#002B5C"}

northwest = {"Timberwolves": "#236192",
            "Nuggets": "#FEC524",
            "Trailblazers": "#E03A3E",
            "Thunder": "#007AC1",
            "Jazz": "#00471B"}

pacific = {"Warriors": "#FDB927",
           "Lakers": "#552583",
           "Clippers": "#C8102E",
           "Kings": "#5A2D81",
           "Suns": "#E56020"}


southwest= {"Mavericks": "#00538C",
            "Rockets": "#CE1141",
            "Grizzlies": "#5D76A9",
            "Pelicans": "#85714D",
            "Spurs": "#C4CED4"}

nba_league = [atlantic, central, southeast, northwest, pacific, southwest]

# 6 rows x 5 columns for all 30 teams
fig, ax = plt.subplots(nrows=6, ncols=5,figsize=(6,6),sharex=True, sharey=True)

# plot title
fig.suptitle('More Money, More Wins', x=0.55, y=1,color = 'black', fontsize=12) 

# Subplots spacing
fig.subplots_adjust(hspace = .5, wspace=.001, top=0.925)

# ax[count][pos]: the axes object contains a numpy array
# count indicates the row
# pos indicates the column
# loop through each division, each team
row = 0 
for division in nba_league:
    col=0 
    for team,color in division.items():
        plot_team_row(final[final['team']==team], ax[row][col],color)
        col=col+1
    row += 1

# labels
ax[5][2].set_xlabel('Standardized Salary', fontsize=8)

ax[0][0].set_ylabel('Win %', fontsize=8)

# ticks
for i in list(range(0,5)):
    ax[5][i].tick_params(labelsize=8)

for i in list(range(0,5)):
    ax[i][0].tick_params(labelsize=8)

plt.figtext(1,0.875, "Atlantic", fontsize=10, color="grey")
plt.figtext(1,0.725, "Central", fontsize=10, color="grey")
plt.figtext(1,0.575, "Southeast", fontsize=10, color="grey")
plt.figtext(1,0.425, "Northwest", fontsize=10, color="grey")
plt.figtext(1,0.275, "Pacific", fontsize=10, color="grey")
plt.figtext(1,0.125, "Southwest", fontsize=10, color="grey") 
plt.show()

# plt.savefig('nba_team_salary.pdf')