Prior studies challenge the practical usefulness of Markowitz portfolio optimization in improving the return-risk tradeoff in portfolio management. We approach this question from a unique angle by examining whether one can improve the performance of a large sample of actual mutual fund portfolios by re-optimizing the holdings using simple mean-variance optimization methods. Our analyses produce compelling evidence of the benefits from Markowitz optimization. Simple portfolio optimization improves mutual fund portfolios’ risk-adjusted performance despite noisy expected return estimates inferred from mutual fund portfolio weights. Several alternative optimization strategies, including the risk-parity portfolio, minimum variance portfolio, mean-variance portfolio and Sharpe ratio maximization portfolio all outperform actual mutual fund portfolios in terms of the Sharpe ratio and other risk-adjusted performance measures. Moreover, the results are robust to subsamples partitioned on various dimensions. In contrast to DeMiguel et al. (2009), we find that the 1/N portfolio performs the worst.