Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/cyberoctane29/nba-player-career-longevity-prediction-using-naive-bayes-classifier

This project analyzes NBA rookie performance data to predict whether a player will stay in the league for at least five years. Using feature engineering and a Naive Bayes model, it identifies key indicators of career longevity, supporting managers and coaches in player retention decisions.
https://github.com/cyberoctane29/nba-player-career-longevity-prediction-using-naive-bayes-classifier

naive-bayes-classifier naive-bayes-implementation nba-analytics nba-classification nba-stats sports-analytics sports-data

Last synced: 3 months ago
JSON representation

This project analyzes NBA rookie performance data to predict whether a player will stay in the league for at least five years. Using feature engineering and a Naive Bayes model, it identifies key indicators of career longevity, supporting managers and coaches in player retention decisions.

Awesome Lists containing this project

README 

          



# NBA Player Career Longevity Prediction: Using Naive Bayes Classifier



This project analyzes how rookie season performance metrics influence the probability of an NBA player sustaining a career of five years or longer. Using **Gaussian Naive Bayes classification**, I aim to model this relationship to inform data-driven decision-making in player scouting and team management. The project uses Python libraries including **pandas**, **seaborn**, **scikit-learn**, **matplotlib**, and **pickle** for data preprocessing, visualization, modeling, and performance evaluation.



---



## **Project Overview**



The **NBA Player Career Longevity Prediction** project aims to:



* **Identify Predictive Features**: Determine which rookie season statistics best forecast long-term career outcomes

* **Build Classification Model**: Implement and evaluate a Gaussian Naive Bayes classifier for career longevity prediction

* **Validate Model Assumptions**: Assess conditional independence, normality, and class balance for model suitability

* **Support Roster Decisions**: Provide actionable insights for scouting, player development, and team-building strategy



---



## **Dataset Structure**



The dataset contains rookie season statistics from **1,340 NBA players**, capturing their on-court performance across various metrics. Key features include:



**Career Longevity Target:**



* `target_5yrs`: Binary outcome variable indicating career duration (1 = career ≥5 years, 0 = otherwise)



**Performance Metrics:**



* `gp`: Games played in the rookie season (integer)

* `min`: Average minutes played per game (continuous)

* `pts`: Average points per game (continuous)

* `fg`: Field goal percentage per game (continuous)

* `3p`: Three-point field goal percentage per game (continuous)

* `ast`: Average assists per game (continuous)

* `stl`: Average steals per game (continuous)

* `blk`: Average blocks per game (continuous)

* `reb`: Average rebounds per game (continuous)

* `tov`: Average turnovers per game (continuous)

* `ft`: Free throw percentage per game (continuous)



**Engineered Features:**



* `total_points`: Total points scored in the rookie season (derived by multiplying `pts` and `gp`)

* `efficiency`: Points per minute played (derived by dividing `pts` by `min`)



This dataset forms the basis for exploring career longevity patterns and developing predictive models to identify players likely to sustain long-term NBA careers — providing valuable insights for scouting and roster management.



---



## **Data Processing and Analysis Steps**



* **Data Cleaning**:



  * Verified no missing values in all 1,340 observations

  * Removed redundant or less informative features



* **Exploratory Data Analysis (EDA)**:



  * Assessed class balance (62% of players had careers ≥5 years)

  * Visualized feature distributions and correlation patterns within classes



* **Feature Engineering**:



  * Created `total_points` (games played × points per game)

  * Developed `efficiency` as a productivity-per-minute metric



* **Modeling Approach**:



  * Fit a **Gaussian Naive Bayes classifier** using scikit-learn

  * Split data into 75% training and 25% testing sets

  * Evaluated model using accuracy, precision, recall, and F1 score



* **Assumption Validation**:



  * Checked for conditional independence and normality within classes

  * Noted moderate correlations and slight normality deviations, but acceptable for model use



---



## **Key Insights**



### **Predictive Features**



* The engineered `efficiency` metric proved highly informative

* Shooting percentages (`fg`, `3p`) contributed meaningfully to predictions

* Traditional stats like points and assists retained value but were complemented by derived features



### **Model Performance**



* Achieved **84.1% precision** in identifying long-career players

* Delivered balanced **68.96% accuracy**

* Recall of **58.6%** indicates room to improve detection of all long-career cases



---



## **Project Highlights**



* Developed **novel efficiency metrics** to enhance predictive power

* Applied **Gaussian Naive Bayes** with thorough assumption checks

* Achieved **high precision**, valuable for strategic roster decisions

* Provided **clear, actionable recommendations** for basketball operations teams

* Established a framework for future model improvements



---



## **Future Work**



* **Enhance Feature Set**: Incorporate advanced stats like Player Efficiency Rating and usage rates

* **Address Class Imbalance**: Explore resampling or reweighting techniques

* **Test Alternative Models**: Compare performance of Random Forest, Logistic Regression, and Gradient Boosting

* **Build Decision-Support Tools**: Develop interactive dashboards or prediction apps for team use

* **Longitudinal Validation**: Apply the model to recent draft classes for ongoing refinement



This project demonstrates how a structured, data-driven approach can produce **practical, interpretable insights** in professional sports analytics, improving talent evaluation and roster construction through predictive modeling.