This paper studies the performance of a sample of funds of hedge funds (FoHFs) from January 1994 to August 2009. We apply the false discoveries (FD) technique of Barras, Scaillet and Wermers (2010) to separate the FoHFs into skilled, zero-alpha and unskilled. We measure the alpha of the FoHFs using two models – (1) a 16-factor model with a combination of factors from Fung and Hsieh (2004) and Capocci, Corhay and Hubner (2005) and (2) a 13-factor model of hedge fund indices from Dow Jones Credit Suisse. Applying the FD procedure to the first model, we find that, after fees, the majority of FoHFs do not channel alpha from single-manager hedge funds. Applying the FD procedure to the second model, we find that only a very small fraction of FoHFs deliver after-fees alpha per se, i.e. on top of the alpha of the hedge fund indices. A series of robustness checks confirms the results of the FD procedure. We also compare the performance of our sample of FoHFs to artificial FoHFs constructed by randomly picking hedge funds. The lack of significant differences in the average performance of the real and artificial FoHFs confirms the results obtained by the FD procedure.