Towards a second law for Lovelock theories