This paper presents a comparison of three portfolio selection models, Mean-Variance (MV), Mean Absolute Deviation (MAD), and Minimax, as applied to the Brazilian Stock Market (BOVESPA). For this comparison, we used BOVESPA data from three different 12 month time periods: 1999 to 2000, 2001, and 2002 to 2003. Each model generated three optimal portfolios for each period, with performance determined by monthly returns over the period. In general, the accumulated returns from the Minimax modeled portfolios were superior to the BOVESPA’s principal index, the IBOVESPA. The MV model was the least efficient for portfolio selection.