The size premium, defined as the outperformance of equities of small and medium-sized companies compared with the shares of large companies, is subject to strong cyclical fluctuations over time. This study examines the predictability of this premium for the Swiss stock market. The forecasts used are developed applying a flexible forecasting approach that is based on time variable multi-factor models. Our strategies provide information ratios significantly greater than 1 for a maximum real-time application of a good seven years. The results show that risk variables such as the credit spread and TED spread, the performance of the S&P 500 and statistical variables such as AR(1) terms or trends calculated using the Hodrick-Presscot filter prove to be successful forecasting variables in our algorithm. Furthermore, variables that sum up the consensus estimates of equity analysts (IBES) for various size portfolios can sometimes make valuable forecast contributions.