Probabilistic methods are used in solving various engineering problems, where a computational model contains random variables. One of the significant fields where probability methods have been applied is also analysis of structural reliability and estimation of probability of failure. A number of probabilistic methods have been evolved. The proposed method under development: Direct Optimized Probabilistic Calculation (DOProC) is highly efficient in terms of computation time and solution accuracy and is mostly faster than in case of other standard probabilistic methods. The novelty of the DOProC lies in an optimized numerical integration that easily handles both correlated and statistically independent random variables and does not require any simulation or approximation technique. The conference contribution summarizes the principles of this method and demonstrates its basic possibilities with focus on probability of failure estimation. The paper presents unpublished details of probabilistic computations based on this method, including a reliability assessment, which is aimed at an improved way of calculating the probability of failure. Originally, the probability of failure was determined from the resulting safety margin histogram. The improved and presented way of computation determines the probability of failure directly during computational operations within the basic calculation algorithm. Significantly, it refines the estimation of probability of failure and represents a relevant step for better application of the DOProC method in solving of structural reliability problems.