Optimality of central and projection algorithms for bounded uncertainty