On convergence of "Divide the Best" global optimization algorithms