The primary metric score of a model: holdout preferred, else training. R² first for a regression, AUC for a
classification; failing those, the first metric the holdout stores, which a trained model's holdout lists first as the
protocol's primary proper score (all a kept simple baseline or a forecast stores).
The primary metric score of a model: holdout preferred, else training. R² first for a regression, AUC for a classification; failing those, the first metric the holdout stores, which a trained model's holdout lists first as the protocol's primary proper score (all a kept simple baseline or a forecast stores).