The research project deals with the development and empirical validation of portfolio strategies. Attention is paid to accounting for the errors that inevitably arise in the econometric estimation of the distributional properties of security returns and in model calibration. Estimation errors in model parameters - including by ignoring their dynamics - often tempt an optimization of the return/risk ratio based on them to extreme portfolio weights, which turn out to be relevant risk drivers in practical implementation. The use of robust optimization procedures and statistical methods to correct estimation errors are the subject of the research project, which is carried out in cooperation with Spängler IQAM invest GmbH.