In the preceding article, Konstantinos Katsikopoulos and Aris Syntetos discussed the trade-offs between forecast bias and forecast variance in choosing a suitable forecasting method. Simple methods, they explain, tend to have large bias but low variance, while complexity reduces bias but at the expense of increasing variance. In short, simple methods might be preferable to complex methods, even if the resulting forecasts are biased. Stephan Kolassa now extends their argument to show that even if we know what the correct model is for the data to be forecast - that is, even if we know the seasonal pattern and other influencing factors for a time series - it may still be better to choose a simpler model, one that excludes one or more of these variables. This is a fascinating takeaway.