I have seen references to the "tilting method" in the theory of large deviations. Is there a simple explanation of what this is, exactly?