Decoding the NBA MVP: A Machine Learning Approach

Every NBA season sparks heated debates about the Most Valuable Player (MVP) award, a prestigious recognition in basketball. Yet, understanding the criteria for this accolade can be perplexing for those unfamiliar with the sport. The MVP isn't simply awarded to the best player but to the one who has the greatest positive impact on their team during the regular season. To unravel the intricacies of this award and explore its patterns, I embarked on a unique journey applying machine learning techniques.

In this project, I aimed to decipher the logic behind MVP selection by leveraging machine learning. By analyzing data from the 2006–07 to 2021–22 NBA seasons, encompassing 16 years, I sought to predict the MVP before the official announcement. To tackle the prediction task, I employed the concept of MVP Share, which normalizes the voting points received by each player relative to the maximum possible points. This approach ensured fair comparisons across seasons with varying numbers of voters. With 71 variables per player, including individual and team statistics, the dataset was comprehensive. Utilizing various regression algorithms such as Support Vector Machines, Random Forest, I evaluated model performance. The goal was to identify the most effective model for MVP prediction.This project transcended its initial aim of MVP prediction, and also  offered me valuable insights into the historical trends and nuances of the NBA MVP award. While the models demonstrated efficacy, there remains room for enhancement by incorporating additional variables and refining model architectures.