In reality, in-sample test may have the issue of over-fitting, thus most of investigators tend to focus on the out-of-sample forecasting performance of proposed predictors and models. All the time, they make efforts to avoid the situation in which the out-of-sample results are less than satisfactory with in-sample performance excellent (see, e.g., Rapach et al., 2010). Hence, it is necessary to further explore the out-of-sample statistical predictive results of the SOR. In our analysis, the