A new tight inequality for ππ scattering