Time series forecasting is a challenging field for machine learning. In most built- in prediction of machine learning tools, it is one-step forecasting from the perspective of time series. In most predictive equations, one-step forecasts are usually good but useless in practice. This paper compares the performance of multistep forecasting of machine learning methods and statistical time series models. We use datasets of two frequencies: US monthly unemployment rate and daily volatility data of exchange- traded fund (ETF) price; and our results show that:First, for one-step forecasting, all models exhibit good out-of-sample predictability; among them, ML methods are not apex forecasters, and several nonlinear statistical time series models outperform them.Second, MLs are not good forecasters when it comes to multistep forecasting. Except the deep learning method known as long short-term memory (LSTM), other ML methods and econometric time series models show themselves as poor forecasters. Moreover, although LSTM outdid itself in forecasting, its performance requires a deeply and properly trained pattern recognition from the network loops, it is not only time-consuming, but also unpredictable.