About Project
A neural network that predicts the outcomes of international football matches, built specifically for the FIFA World Cup 2026.
The model factors in squad strength, form and Elo, and outputs win, draw and loss probabilities along with the expected scoreline. For knockout games, a separate learned penalty-shootout model decides ties based on each team's specific shootout strength.
Try it live below: single predictions, World Cup groups, and the full bracket up to the final, where you can enter real results.
Key Responsibilities & Highlights
Neural network (PyTorch) trained on 49,000 international matches
Features from squad strength (FIFA ratings), Elo, form and head-to-head
Custom learned penalty-shootout model for knockout games
Interactive World Cup bracket, runs entirely in the browser (React)