In forecasting, the challenge faced by analysts is to select the "optimal" forecast and forecasting procedure. One approach to selecting such forecasts is to use composite forecasting methods. This paper views the conceptual framework for forming optimal composite forecasts and attempts to identify circumstances when composite forecasts might work best. Then, Monte Carlo simulations and real world examples are used to compare composite forecasts with individual forecasts and to compare alternative methods of making composite forecasts. The appraisal is intended to provide insights into the potential benefits of composite forecasts and thereby provide some practical guides to the use of composite methods.