In my talk I offer two answers to the above question. The first is semantical: I show that we can conceive of propositions about chance as empirical, and why this helps to make sense of how we determine chances in statistics. The second answer I give is metaphysical: chances can be taken to result from the interplay between levels of description that do not neatly go together. This avoids an appeal to ignorance or explanation as a foundation of chance. Both answers will rely on early work about chances done by von Mises.