All projects
04Probabilistic modellingIn progressIn progress

Champions League Forecasting Lab

A transparent forecasting lab planned around sequential ratings, score models, leakage-safe evaluation, and tournament simulation.

01RateSequential Elo
02ModelPoisson scores
03EvaluateWalk-forward
04SimulateMonte Carlo

The context

A rigorous plan—without pretending it is finished.

The project is designed to make changing football beliefs inspectable: every forecast should exist before the match, evaluation should move chronologically, and tournament odds should be reproducible from the underlying match model.

My contribution

What I built

  1. 01

    Designed a historical match-ingestion and validation workflow.

  2. 02

    Specified sequential Elo ratings and Poisson score modelling, with a possible Dixon–Coles extension.

  3. 03

    Planned walk-forward evaluation using Brier score, log loss, calibration, and error analysis.

  4. 04

    Planned immutable pre-match snapshots and Monte Carlo tournament simulations.

Architecture

How the work moves

01Historical fixtures
02Sequential strength model
03Walk-forward evaluation
04Tournament simulation

Evidence

Planned proof points

01Sequential team-strength design
02Probabilistic score and outcome targets
03Leakage-safe evaluation plan
04Immutable forecast-history requirement

Honest evaluation

Limits and trade-offs

  • No working implementation, output, metric, test suite, repository, or deployment was available during the portfolio audit.
  • Every technical item on this page describes planned scope, not completed functionality.
  • The project is analytical and must not be framed as betting advice.

Inspect the work

Stack and reproduction

  • Python
  • Elo
  • Poisson
  • Calibration
  • Monte Carlo
  • Streamlit
  1. 1Reproduction instructions will be published with the first verified implementation.