The article compares the accuracy of two different forecasts using various measures, even if the errors are not normally distributed or have correlations. They developed tests to see if there is a significant difference in accuracy between the forecasts, both in theory and in practice.