The performance of contrarian strategies relies on the time series properties of stock returns (including self- and cross-autocorrelation) as well as on the cross-sectional variation in expected returns of individual securities. Given the return generating process or time series structure, a stock might be identified as a loser and a winner as well. depending on the length of the ranking (formation) period. Henceforth, the contrarian performance over different holding horizons also depends on the time series properties. Based on monthly returns data for all stocks listed on the Tokyo Stock Exchange (TSE), this paper investigates the contrarian performance over various ranking and holding horizons ranging from one month to three years. Empirical results show that contrarian strategies are profitable for all horizons. On average, we find that negative autocorrelation (market overreaction) is the major source of the contrarian profit.