We test the hypothesis that strategies which are long on portfolios of looser stocks and short on portfolios of winner stocks generate abnormal returns in Brazil. This type of evidence for the US stock market was interpreted by The Bondt and Thaler (1985) as reflecting systematic evaluation mistakes caused by investors overreaction to news related to the firm performance. We found evidence of contrarian strategies profitability for horizons from 3 months to 3 years in a sample of stock returns from BOVESPA and SOMA from 1986 to 2000. The strategies are more profitable for shorter horizons. Therefore, there was no trace of the momentum effect found by Jagadeesh and Titman (1993) for the same horizons with US data. There are remaing unexplained positive returns for contrarian strategies after accounting for risk, size, and liquidity. We also found that the strategy profitability is reduced after the Real Plan, which suggests that the Brazilian stock market became more efficient after inflation stabilization.