The challenge in making optimal use of survey forecasts lies in the frequent entry and exit of individual forecasters. An equal weighted average of the forecasts is generally used. Recently, Capistran and Timmermann (2009) explored a variety of methods to improve the mean forecast of the Survey of Professional Forecasters. They document that a bias-adjusted average forecast marginally improves the benchmark and conclude, in line with previous research, that the “equal-weighted forecast turns out to be extraordinarily difficult to beat.” In this paper we aim to challenge this conclusion. The main strategy we employ is to rank forecasters based on their previous forecasting errors and to take a weighted average of the best ranked forecasters. This simple approach can be carried out in a variety of ways: one can vary the amount of prediction errors that are considered when ranking the forecasters, one can vary the number of best-ranked experts that receive a non-zero weight, etc. The novelty of our approach is that we produce a large amount of such variations and construct an automated system which dynamically selects and combines the best performing restrictions. We demonstrate that the benchmark-models can be beaten for macroeconomic variables like inflation and government spending.