This paper advances a new epistemic transcendental argument (ETA) against determinism. I argue that determinism entails a higher degree of epistemic risk than libertarianism. To make this claim precise, I formalize risk through a metric space W and represent two theories of epistemic risk as alternative metrics, D and N. By exploring the concepts of modal closeness and normalcy, I show that these metrics are superior to traditional conceptions in accounting for intuitions about risk and ensuring objective assessment. Given some additional assumptions about knowledge and the past, I construct an ETA grounded in epistemic risk aversion: under determinism, all possible worlds count as modally close and normal, making many beliefs epistemically risky. This proposal departs from the Epicurean tradition and opens a new route against determinism.