Backtesting Without Fooling Yourself

How to test a football model honestly: out-of-sample data, proper scoring rules and the mistakes that make backtests look better than reality.

A backtest shows how a model would have done in the past — and it is easy to fool yourself. This section covers out-of-sample testing, scoring rules such as the Brier score and log loss, and the common mistakes that inflate results.

No backtest guarantees future results.

Tutorials in this module

Poisson match predictorEnter expected goals for both teams and get every scoreline, 1X2, over/under 2.5 and BTTS.