— GLOSSARY
How to judge a forecasting model's evidence
Short, plain-English explanations of the terms behind a trustworthy track record. Each one shows how ApolloAlpha reports it.
Brier scoreThe Brier score measures how close probability forecasts are to what actually happened. Lower is better. How to read it, with a worked example.Walk-forward backtestA walk-forward backtest retrains a model on past data and scores it only on the period that follows. Why it avoids look-ahead bias, and what it cannot prove.Calibrated forecastA forecast is calibrated when events given a 70% chance happen about 70% of the time. How calibration is checked and how to read a calibration table.Out-of-sample testingOut-of-sample testing scores a model only on data it never trained on. What it means, why it matters for trading models, and how to spot results that skip it.