In recent years, increasing effort has been devoted to the study of virtual world economies due to their potential of increasing our understanding of the real world economy, and vice versa. Due to a scarce availability of reliable global data, previous virtual world economic studies have been largely limited to qualitative observations. This paper presents novel financial data and is the first to apply a time series approach to the forecasting of virtual commodity prices. The results are assessed against the random walk and, from an efficient markets perspective, evaluates the potential of virtual worlds becoming experimental simulations for the real.