StatPilot

StatPilot is a web application that predicts NBA player prop outcomes—such as points, rebounds, and assists—using historical player performance data. It provides probabilities that a player will go over or under a given stat line against specific opponents.

🌐 Live Demo: https://statpilot-6cg9.onrender.com/

---

Project Overview

StatPilot leverages a linear regression model trained on recent game data to estimate player performance. The model incorporates:

• Opponent-specific averages
• Season-long averages
• Recent game averages (rolling window)

Using these features, it predicts the expected stat value and calculates the probability of exceeding or falling below user-defined stat lines, assuming a normal distribution of performance.

---

Backend Architecture

• Built with FastAPI serving both API endpoints and frontend HTML pages.
• Reads and cleans an NBA dataset (LeGamble_Dataset - Sheet1.csv).
• Core prediction logic in get_stat_probability:
  • Filters player data by opponent and stat category.
  • Fits a linear regression model on opponent, season, and recent averages.
  • Adjusts predictions based on data availability and variability.
  • Returns detailed stats and probability estimates.
• API endpoints:
  • / — homepage
  • /stats — main dashboard
  • /chart — chart visualization
  • /input — accepts user inputs and returns prediction JSON.

---

Frontend Details

• Static assets (images, CSS) served from /static.
• HTML pages rendered by FastAPI and display the interactive UI.
• User input forms send requests to the backend /input endpoint to fetch prediction results dynamically.

---

Technologies Used

• Python 3.9
• FastAPI
• Uvicorn
• Pandas & NumPy
• Scikit-learn (Linear Regression)
• SciPy (statistical functions)
• Docker (containerization)

---

Deployment

StatPilot is deployed as a live web application using Docker, allowing seamless scaling and easy access for users without requiring local setup.

🌍 Check it out live: https://statpilot-6cg9.onrender.com/