Methods of global optimization are widely used to solve various applied problems. At the same time, for metaheuristic methods there is no guaranteed convergence of generated test points to the exact solution. To monitor the quality of the optimization process and the related exploitation and exploration, the diversity value of the population is used. Appropriate method settings are selected based on this value. However, in the absence of guaranteed convergence, this value only gives an intuitive performance and allows us to draw some plausible conclusions. The purpose of this study is to develop a new approach to assessing diversity in the framework of numerical optimization methods, directly related to the convergence of the process. The new approach is based on the understanding of diversity in the information theory. This approach allows us to estimate the required number of method iterations to obtain a given approximation of the solution. The effectiveness of the proposed approach is demonstrated using the example of the survival of the fittest algorithm (SoFA).