This paper tests the contrarian investment strategy, which predicts that stocks that consistently underperform (outperform) the market would in subsequent periods outperform (underperform) those stocks that have previously outperformed (underperformed) the market, using a revised nonparameteric estimator of excess returns and risk coefficients, specified in a time-varying risk multi-factor CAPM model. The conventional parametric approach is used as the control estimator for comparing the effectiveness of this nonparametric approach. Using bootstrap simulations, conventional CAPM estimates reveal that there exists a significant price reversal effect between the formation and test periods, as did the nonparametric estimates. However, one striking difference was that the nonparametric approach revealed more conservative but still significant estimates than did conventional parametric approaches. The multi-factor model reveals weaker results of price reversals and the results dissipate over time. Therefore, the contrarian strategy is only weakly supported and it is concluded that, ceteris paribus , the nonparametric approach yields significantly better estimates than do parametric approaches in estimating the parameters of both the single-factor and multi-factor CAPM.