Tracking accounts of knowledge formulated in terms of counterfactuals suffer from well known problems. Examples are provided, and it is shown that moving to a dispositional tracking theory of knowledge avoids three of these problems.