ABSTRACT
The Latin phrase caveat emptor means simply “let the buyer beware”. This chapter outlines how one goes about deciding whether a series or group of long-range forecasts has any skill or not, and some of the considerations behind the steps leading to the decisions. The appropriate no-skill reference level depends entirely on how a set of forecasts and observations is distributed. A symmetric normal distribution and an asymmetric distribution are sketched. In the former case the median and the mean are the same. A forecaster will be right nearly all of the time if he forecasts above normal exclusively. However, such forecasts will have little value. The forecast distribution was heavily weighted to near-normal values, but also had twice as many forecasts as below forecasts. On the other hand the observations were heavily skewed toward above normal, with almost as many of these as below and near-normal forecasts combined.
