Assessment of geological uncertainties via probabilistic modelling has become widespread in recent decades. However, like in any other field of knowledge, similar errors occur repeatedly in such models. The aim of this paper is to accumulate the experience of uncertainty analysis in geology in order to upgrade the approaches to accomplishing this task and improve the quality of the results obtained. A large number of probabilistic models created by specialists from different companies were investigated. Based on this analysis characteristic errors were revealed and divided into groups. The groups include wrong concept of reservoir geology, incorrect assessment of unknown parameters, analysis of only volumetric variables for major produced fluid, lack of correlation in dependent parameters / objects or wrong correlations, small sample of equiprobable model realizations, mismatch of model scale and assigned tasks. Each group of errors is illustrated by examples. Causes and consequences of errors are exposed. Recommendations for their detection and reduction are given. The conducted study showed that it is impossible to provide high-quality probabilistic uncertainties assessment without an integrated multidisciplinary team of specialists that are competent in creation of such models. In addition, the necessary conditions for this are well-established communication within the team, peer involvement from the early stages of the project and the use of a correct evaluation methodology.