The goal of this project is to predict the best NBA players based on their statistics.
The data is scraped from the Basketball Reference website using the BeautifulSoup library. Player stats are then cleaned, preprocessed using the Pandas library, and stored in CSV files in the data directory. Addional properties are added to the data:
- ANBA: if the player was an All-NBA player,
- RK: if the player was a rookie,
- ANBARK: if the player was an All-NBA player in their rookie year,
- Year: the year the data was collected.
The following features are available for each player:
Player Pos Age Tm G MP PER ORB% DRB% TRB% AST% STL% BLK% USG% OWS DWS WS WS/48 OBPM DBPM BPM VORP ANBA RK ANBARK Year
0 Ömer Aşık C 31 TOT 18 182 3.1 5.5 22.2 14.0 2.1 0.5 1.9 8.7 -0.2 0.2 -0.1 -0.021 -6.5 -1.5 -8.0 -0.3 0 0 0 2018
1 Luke Babbitt SF 28 TOT 50 715 9.8 1.1 13.9 7.5 6.5 0.5 0.8 15.0 0.5 0.4 0.9 0.060 -0.9 -1.5 -2.4 -0.1 0 0 0 2018
2 Marco Belinelli SG 31 TOT 80 1947 14.1 1.1 7.3 4.3 11.9 1.6 0.5 20.7 2.4 1.3 3.7 0.092 0.4 -0.8 -0.5 0.8 0 0 0 2018
The data is then used to train a Random Forest model using the Scikit-learn library.
Only a select number of features are used to train the model. Those were selected based on their importance in predicting the best players.
Feature importances of each feature in the Random Forest model.
Based on the that information, the following features were selected:
selected_features = [
"VORP",
"WS",
"OWS",
"BPM",
"OBPM",
"USG%",
"PER",
"DWS",
"MP",
"WS/48",
]After training the model, the best players are predicted based on the data from the 2024 season. The results are stored in the JSON file.
To run the project, first install the required dependencies:
pip install -r requirements.txtThen run the following command:
python Kajdan_Ignacy.py output.jsonThis repository is licensed under the MIT License - see the LICENSE file for details.