One period portfolio optimisation is investigated by using and comparing several cost functions which measure the portfolio risk as the classical volatility, the conditional VAR, the expected short-fall. We set up a Monte Carlo simulation of prices of a given asset system for the next 3 years, based on the historical multivariate distribution of price fluctuations. One period optimisation is then repeated at the beginning of each year: a strategy is the choice of an optimal portfolio at each of these times. There are several strategies leading to a given final expected return and among these we choose the ``best strategy'' as the one with minimum risk as measured by the realized value of the chosen cost function on the simulated piece of the time-series.