Overview

🏈 NFL Prediction System

Ensemble prediction system combining statistical models and real-time data to predict NFL games with 74.7% accuracy.

🎯 Results

Metric	Value
Average Accuracy	74.7% (over 3 weeks)
Week 1	75% (12/16)
Week 2	68% (11/16)
Week 3	81% (13/16)

Comparison: Most professional models achieve 55-60% accuracy.

🧠 How It Works

Ensemble system combining 4 specialized predictors:

1. Statistical Predictor: Analyzes team performance metrics

   • Win/loss records
   • Point differential
   • Recent form (last 4 games)
   • Home/away splits

2. Weather Analyzer: Real-time weather impact

   • Live data from Open-Meteo & NOAA APIs
   • Team weather performance profiles
   • Historical weather impacts on gameplay

3. News Sentiment: Media coverage analysis

   • Real-time news scraping
   • Team morale indicators
   • Momentum factors

4. Market Analyzer: Betting market patterns

   • Line movements
   • Public vs sharp money
   • Market efficiency indicators

Each predictor votes, weighted by historical accuracy.

🚀 Tech Stack

• Backend: Python, FastAPI, Docker
• Data Sources: ESPN API, Open-Meteo, NOAA, RSS feeds
• Frontend: React dashboard, React Native (Next)
• Database: SQLite

📊 Key Features

• Real team statistics from ESPN
• Live weather data (no API key required)
• Bias elimination (60% home picks vs industry 100%)
• Confidence scoring for each prediction
• Dockerized for easy deployment

🔧 Setup

Live Demo

Link: https://nfl-predictor-system-production.up.railway.app/

• Please Note: This a proof of Concept. React Native App is still in progress.
• Wanted to ensure that the agents are actually working accurately before developing an entire mobile application.

For Developers (Source Code)

git clone https://github.com/Ansh757/nfl-predictor.git

Add .env (to add your OWN WEATHER_API_KEY)

docker-compose up -d