Nº 10of 22Index

ML-Copa

World Cup prediction with XGBoost, adaptive Elo and Dixon-Coles

Year
2026
Status
SHIPPED
Stack
Python · XGBoost · Pandas · Scikit-Learn +2
ml-copa
ML-Copa — screenshot
1 screen
ML-Copa — screenshot

screenshot

01The project

World Cup prediction system combining 49 thousand historical international matches with an XGBoost ensemble, adaptive Elo ratings and Poisson/Dixon-Coles probability models. CRISP-DM pipeline with pre-match feature engineering and calibrated probabilities.

02What was built

  1. 49,071 international matches (1872–2024)

  2. Adaptive Elo rating + Dixon-Coles

  3. Full CRISP-DM pipeline

  4. Validation with log-loss and probability calibration

03Stack

  • Python
  • XGBoost
  • Pandas
  • Scikit-Learn
  • penaltyblog
  • Sportmonks API
Next project →Nº 11

C++ firmware for an ESP32 with a touch screen: 8 LVGL mini-apps and an SDL2 simulator for developing without the board