STAT 451 Project Proposal

Names: Jackson Cramer, Joshua DeRuyter, Yingying Xie, Siti Aleeya Nuha Roslee

Read Data
In [2]:
import pandas as pd

df = pd.read_csv("games.csv")
df.head()
Out[2]:
GAME_DATE_EST GAME_ID GAME_STATUS_TEXT HOME_TEAM_ID VISITOR_TEAM_ID SEASON TEAM_ID_home PTS_home FG_PCT_home FT_PCT_home ... AST_home REB_home TEAM_ID_away PTS_away FG_PCT_away FT_PCT_away FG3_PCT_away AST_away REB_away HOME_TEAM_WINS
0 2022-12-22 22200477 Final 1610612740 1610612759 2022 1610612740 126.0 0.484 0.926 ... 25.0 46.0 1610612759 117.0 0.478 0.815 0.321 23.0 44.0 1
1 2022-12-22 22200478 Final 1610612762 1610612764 2022 1610612762 120.0 0.488 0.952 ... 16.0 40.0 1610612764 112.0 0.561 0.765 0.333 20.0 37.0 1
2 2022-12-21 22200466 Final 1610612739 1610612749 2022 1610612739 114.0 0.482 0.786 ... 22.0 37.0 1610612749 106.0 0.470 0.682 0.433 20.0 46.0 1
3 2022-12-21 22200467 Final 1610612755 1610612765 2022 1610612755 113.0 0.441 0.909 ... 27.0 49.0 1610612765 93.0 0.392 0.735 0.261 15.0 46.0 1
4 2022-12-21 22200468 Final 1610612737 1610612741 2022 1610612737 108.0 0.429 1.000 ... 22.0 47.0 1610612741 110.0 0.500 0.773 0.292 20.0 47.0 0

5 rows × 21 columns

Data Description:

This dataset was acquired via Kaggle, titled "NBA Games Data." Each row of the dataset corresponds to a game and each column represents a statistic related to that game from the 2003 to 2022 seasons.

Link to dataset: https://www.kaggle.com/datasets/nathanlauga/nba-games?resource=download

Descriptions of the Questions

What statistics related to a NBA game are most valuable for predicting whether or not a home team wins?

Can we create a model using these features to predict whether or not a future team wins?

Descriptions of the Variables
  • FG_PCT_home - field goal percentage home team
  • FT_PCT_home - free throw percentage of home team
  • FG3_PCT_home - three point percentage of the home team
  • AST_home - assists of the home team
  • REB_home - rebounds of the home team
  • FG_PCT_away - field goal percentage away team
  • FT_PCT_away - free throw percentage of away team
  • FG3_PCT_away - three point percentage of the away team
  • AST_away - assists of the away team
  • REB_away - rebounds of the away team
  • HOME_TEAM_WINS - if home team won the game
    • Descriptions of the Methods
  • Logistic Regression - we will use this to turn our binary variable HOME_TEAM_WINS into a probability spanning from 0 to 1.
  • Lasso Regression - we will use this to select the features for which we will use to predict the probability that the home team wins.
  • Model Selection - we will apply multiple algorithms (decision tree, logistic regression, SVM) and evaluate which is most effective after feature selection (involving training, validation, and test split).