Evaluating time series models is usually different from most other machine learning tasks as we usually don’t have i.i.d. data.
If the sliding window size is 1, then we have the simplest out-of-sample holdout scenario.
L Ma (2022). 'Evaluating Time Series Models', Datumorphism, 04 April. Available at: https://datumorphism.leima.is/wiki/forecasting/evalutate-time-series-models/.