Quantitative information-flow models typically assume that secret information is static. But real-world secrets, such as mobile device locations and account passwords, evolve over time. So it's not just the current value of a secret that matters, but also how the secret changes. If information leaks about how secrets change, adversaries might be able to predict future (or past) secrets. Hiding the correlation between time and the value of a secret can even be more important than hiding the secret itself. This paper formalizes information flow in the context of dynamic systems with adaptive adversaries, and shows how to quantify leakage of information about secrets and about how secrets change. Careful modeling of the adversary's resources, and his decision about when to attack a system, turns out to be essential. We give operational interpretations of our metrics, relative to well-defined attack scenarios and attacker capabilities.