The NBA Season MVP Predictor project focuses on predicting the NBA player most likely to win the MVP award for the season. The award has been given out since the 1955-56 season, recognizing the top performer of the regular NBA season.
Learn more about the MVP award's history on Wikipedia.
🔍 Explore the predictor's results via our deployed Streamlit application.
With only one MVP each year amidst hundreds of players, the data imbalance is stark.
Solutions:
- Target Variable: Instead of the MVP award, use the MVP share. This allows for a regression model, as multiple players receive votes every season.
- Data Filtering: Set thresholds to exclude non-MVP contenders:
- Players participating in over 50% of season games.
- Players with a minimum of 24 minutes playtime per game.
- Players from teams ranking within the top-10 in their conference.
MVP winners may vary year-on-year, even with consistent stats, depending on the competition.
Solutions:
- Stat Normalization per Season:
- Employ min-max scaling.
- Standardize the data.
All statistical data for this project was meticulously scraped from Basketball Reference. A heartfelt thank you to the creators for such a comprehensive and user-friendly database.