After hosting the 1994 Winter Olympics, the Norwegian national and local authorities expected a 'big boom' in tourism; the actual effects have been less than, and different from, the predictions, and 40% of the full-service hotels in Lillehammer have gone bankrupt. This paper compares ex ante theories and predictions with the ex post reality. Reference areas and time-series analysis are used to clarify the counter-factual and internal validity. International comparisons among Olympic hosts identify general patterns. The aim is to help planners of mega-events and other rare projects to improve their forecasting and decisions. Ex post studies can improve the quality of future ex ante impact assessment of unique projects, but it is important to clarify partial, interaction and cumulative effects. Also, much more careful market and cost–benefit studies are needed.