metadata

title: NBA Performance Predictor
emoji: 🏀
colorFrom: red
colorTo: blue
sdk: gradio
sdk_version: 5.44.0
app_file: app.py
pinned: false
license: mit

NBA Player Performance Predictor

Model Description

This interactive web application predicts NBA player points per game (PPG) using machine learning. The model analyzes historical player statistics, lag features, and engineered metrics to make predictions.

Features

Interactive Interface: User-friendly sliders and inputs for player statistics
Example Players: Pre-loaded NBA stars (LeBron James, Stephen Curry, etc.)
Real-time Predictions: Instant predictions as you adjust parameters
Player Categories: Automatic classification (Role Player → Superstar)
Mobile Friendly: Works on phones, tablets, and desktops

How to Use

Input Current Season Stats: Use sliders to set age, games played, minutes, etc.
Add Historical Data: Enter previous season performance metrics
Select Position: Choose the player's primary position
Get Prediction: Click "🔮 Predict Performance" for instant results
Try Examples: Use the example player buttons for quick testing

Model Details

Task: Regression (Predicting NBA player points per game)
Method: XGBoost with time-series features
Features: Age, games, minutes, shooting stats, historical performance
Performance: RMSE ~3-5 points per game, R² ~0.6-0.8

Key Features Used

The model considers various factors:

Basic Stats: Age, Games, Minutes Played, Field Goals, etc.
Historical Performance: Previous season statistics
Efficiency Metrics: Points per minute, overall efficiency
Position & Team: Encoded categorical variables
Trend Analysis: Performance changes over time

Prediction Categories

Based on predicted PPG:

🔵 Role Player: < 8 PPG
🟢 Solid Contributor: 8-15 PPG
🟡 Good Scorer: 15-20 PPG
🟠 Star Player: 20-25 PPG
🔴 Superstar: 25+ PPG

Example Players

Try these pre-loaded examples:

LeBron James (Prime): All-around superstar stats
Stephen Curry (Peak): Elite shooting guard numbers
Rookie Player: Typical first-year player stats
Veteran Role Player: Experienced bench contributor

Technical Implementation

Frontend: Gradio for interactive web interface
Backend: Python with XGBoost, scikit-learn, pandas
Deployment: Hugging Face Spaces
Fallback Mode: Simple heuristic when ML model unavailable

Limitations

Works best for players with NBA history (lag features required)
May be less accurate for rookies or players with significant role changes
Predictions based on historical patterns, may not account for injuries or major team changes
Current version runs in fallback mode (simplified predictions)

Future Improvements

Full XGBoost model integration
Additional statistics (advanced metrics, team context)
Multi-target prediction (rebounds, assists, efficiency)
Player comparison features
Historical trend visualization

Usage Examples

Basic Prediction

# Example input for a typical NBA player
player_stats = {
    'age': 27,
    'games': 75,
    'minutes': 32.0,
    'field_goal_pct': 45.0,
    'position': 'Small Forward',
    'pts_last_season': 18.5
}

Star Player Example

# Example for elite player
star_stats = {
    'age': 28,
    'games': 79,
    'minutes': 36.0,
    'field_goal_pct': 50.0,
    'position': 'Point Guard',
    'pts_last_season': 28.5
}

Data Sources

The model was trained on historical NBA player statistics including:

Regular season performance data
Multiple seasons for trend analysis
Various player positions and team contexts

Ethical Considerations

This model is for educational and analytical purposes only. It should not be used for:

Player salary negotiations without additional context
Draft decisions as the sole determining factor
Any form of discrimination or bias in player evaluation

Contact & Feedback

Feel free to provide feedback or suggestions for improvements. This is an educational project demonstrating machine learning applications in sports analytics.

Live Demo: Try the interactive interface above!
Status: Currently running in fallback mode (simplified predictions)
Next Update: Full XGBoost model integration for enhanced accuracy