Loading...
Not Loading? Load a simpler page!
← Back to Projects
Data Science & AI
AMPA Predictor
A Next-Generation Statistical Football Predictor for the 2026 FIFA World Cup.
Overview
AMPA (Agnel Match Predicting Algorithm) is a state-of-the-art simulation framework engineered specifically for the 2026 FIFA World Cup. It processes over 49,000 historical international matches dating back to the 1870s, computes dynamic attacking and defensive strengths, and runs a complex Bivariate Poisson Distribution to simulate every possible scoreline of every match in the tournament.
The Challenge
Unlike basic predictors that rely purely on static FIFA rankings or simple win/loss ratios, AMPA required a robust mathematical model. I had to resolve the classical overdispersion problem in football goal modeling where the variance of goals is greater than the mean.
The Solution
I implemented a Negative Binomial Distribution with Dixon-Coles correction to prevent 0-0 inflation, along with a custom Elo rating system. The project is split into AMPE (Python Engine) which acts as the brain, and AMPI (Web Interface) which visually renders the tournament bracket.
The Statistical Model
Custom Elo Rating System
Goal difference multiplier and tournament weighting ensure accurate relative strength tracking.
Negative Binomial Distribution
Dixon-Coles correction prevents 0-0 inflation, with a custom bivariate urgency factor for highly realistic match outcomes.
Form, Streaks & Momentum
Tracks a team's last 10-15 matches to trigger momentum multipliers based on streaks and recent head-to-head records.
Live Data Integration
Fetches real-time, in-progress World Cup 2026 scores. Real outcomes overwrite predictions dynamically.