NBA Analytics & Visualization Project

Vision

The project aimed to uncover insightful trends and performance patterns in NBA player and team statistics from 2012 to 2018. By leveraging machine learning, statistical modeling, and dynamic visualizations, the goal was to better understand player dynamics, performance forecasting, and team strategies over time.

Approach

• Data Preparation

  • Collected and processed NBA box score data (2012–2018) using custom Python scripts.

  • Aggregated player performance monthly and yearly with sum_player.py, outputting clean CSVs for modeling and visualization.

• Player & Team Comparisons

  • Conducted comparative analysis on Steph Curry vs. other point guards, similarly built players (within 1 standard deviation for height and weight), and all-stars.

  • Utilized curculator.py to drive these analyses, integrating regression and feature engineering pipelines.

• Machine Learning & Simulation

  • Applied classification, clustering, and regression models (via team_change.py) to identify key patterns in player/team performance.

  • Built time-series models to evaluate momentum and simulate team dynamics under hypothetical scenarios.

• Visualization

  • Delivered a polished PowerBI dashboard to present insights on player comparisons, clustering groupings, and regression outcomes in an interactive format.

Challenges

• Managing large datasets with inconsistent player identifiers and missing values.

• Ensuring accurate aggregation across years and teams.

• Balancing interpretability with model complexity in time series and simulation outputs.

Conclusion

This project highlighted the power of data science in sports analytics, combining Python-based machine learning with real-time data exploration through PowerBI. It demonstrated advanced use of statistical models, clustering, and simulation techniques to deliver actionable insights into NBA performance trends.