As being in a lognormal-securities market, this study develops a simple rule in constructing optimal portfolios with regard to the situation that the probability distribution of portfolio returns does not have finite moments. By means of asymptotic properties when short sales are not allowed, the simple rule incorporating estimation risk can be derived accordingly. Our numerical example specifies optimal portfolios with estimation risk are not equivalent to those without estimation risk considered. In addition, portfolios constructed based on the simple rule are examined to present a better out-of-sample investment performance relative to its counterparty and a naïve benchmark.