On the convergence of best uniform deviations