Accurate forecasts are essential for a great number of applications ye
t large errors and considerable uncertainty characterize most of our a
ttempts for predicting the future. This article surveys empirical stud
ies on forecasting accuracy, summarizing their conclusions and examini
ng the implications involved. It then discusses the effects of system
complexity on forecasting accuracy and the related uncertainty while p
redicting future events. Finally, it provides suggestions for forecast
ing as accurately and realistically as possible when large and complex
systems are involved.