Lecture 22 – Data 100, Fall 2024¶

Data 100, Fall 2024

Acknowledgments Page

In [1]:
import numpy as np
import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
import plotly.io as pio
pio.templates["plotly"].layout.colorway = px.colors.qualitative.Vivid
px.defaults.width = 800

from scipy.optimize import minimize
import sklearn.linear_model as lm
from sklearn.metrics import r2_score

In this lecture, we will work with the basketball dataset, which contains information about basketball games played in the NBA. In the cell below, we perform data cleaning to transform the data into a useful form, which we store as the DataFrame games.

Our goal in this portion of the lecture is to predict whether or not a team wins their game ("WON") given their "GOAL_DIFF". The variable "GOAL_DIFF" represents the difference in successful field goal rates between the two teams competing in a game. A positive value for "GOAL_DIFF" means that a team made more field goals than their opponent; a negative value means that the opponent made more field goals.

In [2]:
basketball = pd.read_csv("data/nba.csv")
basketball.head()
Out[2]:
SEASON_ID TEAM_ID TEAM_ABBREVIATION TEAM_NAME GAME_ID GAME_DATE MATCHUP WL MIN FGM ... DREB REB AST STL BLK TOV PF PTS PLUS_MINUS VIDEO_AVAILABLE
0 22017 1610612744 GSW Golden State Warriors 21700002 2017-10-17 GSW vs. HOU L 240 43 ... 35 41 34 5 9 17 25 121 -1 1
1 22017 1610612745 HOU Houston Rockets 21700002 2017-10-17 HOU @ GSW W 240 47 ... 33 43 28 9 5 13 16 122 1 1
2 22017 1610612738 BOS Boston Celtics 21700001 2017-10-17 BOS @ CLE L 240 36 ... 37 46 24 11 4 12 24 99 -3 1
3 22017 1610612739 CLE Cleveland Cavaliers 21700001 2017-10-17 CLE vs. BOS W 240 38 ... 41 50 19 3 4 17 25 102 3 1
4 22017 1610612750 MIN Minnesota Timberwolves 21700011 2017-10-18 MIN @ SAS L 240 37 ... 31 42 23 7 4 13 16 99 -8 1

5 rows × 29 columns

In [3]:
basketball = pd.read_csv("data/nba.csv")
first_team = basketball.groupby("GAME_ID").first()
second_team = basketball.groupby("GAME_ID").last()
games = first_team.merge(second_team, left_index = True, right_index = True, suffixes = ["", "_OPP"])
games['GOAL_DIFF'] = games["FG_PCT"] - games["FG_PCT_OPP"]
games['WON'] = (games['WL'] == "W").astype(int)
games = games[['TEAM_NAME', 'TEAM_NAME_OPP', 'MATCHUP', 'WON', 'WL', 'GOAL_DIFF']]
games
Out[3]:
TEAM_NAME TEAM_NAME_OPP MATCHUP WON WL GOAL_DIFF
GAME_ID
21700001 Boston Celtics Cleveland Cavaliers BOS @ CLE 0 L -0.049
21700002 Golden State Warriors Houston Rockets GSW vs. HOU 0 L 0.053
21700003 Charlotte Hornets Detroit Pistons CHA @ DET 0 L -0.030
21700004 Indiana Pacers Brooklyn Nets IND vs. BKN 1 W 0.041
21700005 Orlando Magic Miami Heat ORL vs. MIA 1 W 0.042
... ... ... ... ... ... ...
21701226 New Orleans Pelicans San Antonio Spurs NOP vs. SAS 1 W 0.189
21701227 Oklahoma City Thunder Memphis Grizzlies OKC vs. MEM 1 W 0.069
21701228 LA Clippers Los Angeles Lakers LAC vs. LAL 0 L 0.017
21701229 Utah Jazz Portland Trail Blazers UTA @ POR 0 L -0.090
21701230 Houston Rockets Sacramento Kings HOU @ SAC 0 L -0.097

1230 rows × 6 columns

Logistic Regression¶

If we visualize our data, we see a very different result from the scatter plots we have worked with for linear regression. Because a team can only win or lose a game, the only possible values of "WON" are 1 (if the team won the game) or 0 (if the team lost).

In [4]:
px.scatter(games, 
           x="GOAL_DIFF", y="WON", color="WL", 
           hover_data=['TEAM_NAME', 'TEAM_NAME_OPP'])