BUILT BY

YASH MORI

AI/ML Engineer · Bangalore, India

Built this project to explore the intersection of sports analytics and machine learning — from raw match data to a fully simulated 2026 World Cup using an XGBoost + Random Forest ensemble with ELO and EA FC squad ratings.

PROJECT STATS

Calibration (ECE)0.018
Log Loss0.826
Input Features97
Training Matches35,304
Training Span1884–2024
Simulations10,000

TECH STACK

What powers the predictions, the simulations, and the frontend.

ML & DATA
XGBoostscikit-learnNumPy / PandasJupyter Notebooks
BACKEND
PythonFastAPIPoisson Simulation EngineHuggingFace Spaces
FRONTEND
Next.jsTypeScriptTailwind CSSVercel

DATA SOURCES

Four independent datasets, combined to give the model a complete picture.

International Match Database

35,304 matches from 1884 to 2024 — every FIFA-recognized international result including friendlies, qualifiers, and tournament matches.

ELO Rating System

Historical ELO ratings for every team at the time of each match. Updated after every result, weighted by opponent strength and match importance.

EA Sports FC Ratings

Squad-level player ratings from EA FC (FIFA 15 through FC 26). Scout-assessed attributes covering pace, shooting, passing, defending, and physicality.

Football Manager 2023

Player attributes for 209 nationalities — calibrated to EA scale to fill coverage gaps for nations not in the EA FC database.

FIFA 2026 Predictor · Built by Yash Mori · AI/ML Engineer