Using a very large dataset that spans the period 1996 to 2007 and contains transaction prices for all reported auctions at the Chicago Wine Company, we analyze how the prices of high-end wines have evolved over the last decade. We characterize the returns for different wine categories and show that characteristics like vintage, rating and ranking have an impact not only on the pricing of wines but also on their subsequent returns. The best wines according to these characteristics earn higher returns and tend to have either a lower or a similar variance than less good wines. Nevertheless, the different categories of wines seem to follow a similar trend over the long run. This essentially means that the market for Bordeaux wines is not segmented. To address the issue of diversification, we consider a realistic setting that accounts for covariance between equities and wines and also for coskewness and cokurtosis. We employ a polynomial goal programming model to investigate how investor preferences affect the portfolio allocation and the return distribution. Wine returns are only slightly correlated with other assets and as such they can be used to reduce the risk of an equity portfolio. Wines look even more attractive when the investor also has concerns about the skewness of his portfolio. However, the part to be invested in wines is reduced if we include the kurtosis into the analysis. Finally, it seems advisable to diversify across the range of wine categories as their moves in the short-run are partially independent of each other. First growths and wines rated as extraordinary by Robert Parker deliver the best tradeoff in terms of portfolio expected returns, variance, skewness and kurtosis for most investor preference settings considered.