Accuracy & evaluation

MASE

Also called mean absolute scaled error

MASE divides your MAE by the MAE of a naive seasonal baseline. Below 1 means you beat 'same as last week' (or last year). It is how you compare series of different sizes without a percent.

The baseline is the point. If you cannot beat last week's value, you are not forecasting; you are formatting. MASE makes that visible.

Define the seasonal period the way the business runs. Daily series often use 7. Monthly use 12. A wrong period makes a weak baseline look strong, and your MASE look heroic.

MASE still ignores bias direction. Pair it with a signed error. And if the naive MAE is tiny (a nearly flat series), MASE becomes jumpy — look at the raw MAE too.