The root-mean-squared error (RMSE) and mean absolute error (MAE) are widely used metrics for evaluating model performance. However, there is ongoing debate about which is more appropriate. RMSE is optimal for normally distributed errors, while MAE is optimal for Laplacian errors. When errors deviate from these distributions, other metrics may be more suitable. The choice of error metric should align with the expected probability distribution of the errors, as this affects the validity of inferences. This paper reviews the justification for choosing between RMSE and MAE and discusses alternatives better suited for complex error distributions. It emphasizes the importance of understanding the underlying error distribution when selecting a metric. The paper also highlights that while RMSE and MAE are useful, they are not always the best choices, and alternatives like robust statistics or likelihood-based methods may be more appropriate. The discussion includes examples from hydrology and rainfall-runoff modeling, and it concludes that the choice of error metric should be guided by the specific characteristics of the data and the goals of the analysis.The root-mean-squared error (RMSE) and mean absolute error (MAE) are widely used metrics for evaluating model performance. However, there is ongoing debate about which is more appropriate. RMSE is optimal for normally distributed errors, while MAE is optimal for Laplacian errors. When errors deviate from these distributions, other metrics may be more suitable. The choice of error metric should align with the expected probability distribution of the errors, as this affects the validity of inferences. This paper reviews the justification for choosing between RMSE and MAE and discusses alternatives better suited for complex error distributions. It emphasizes the importance of understanding the underlying error distribution when selecting a metric. The paper also highlights that while RMSE and MAE are useful, they are not always the best choices, and alternatives like robust statistics or likelihood-based methods may be more appropriate. The discussion includes examples from hydrology and rainfall-runoff modeling, and it concludes that the choice of error metric should be guided by the specific characteristics of the data and the goals of the analysis.