Accuracy-centered epistemology (ACE) enshrines credal accuracy as the central epistemic good and interprets norms of epistemic rationality as policies that govern its rational pursuit. Recently, Graham Oddie and others have argued that accuracy-centered views are untenable because they measure accuracy using proper scoring rules, which make accuracy entirely a matter of the divergences between credences and truth-values. This, critics say, makes it impossible for accuracy-centered views to accommodate the idea that some credal states are more accurate than others in virtue of assigning higher credences to false propositions with high levels of “truth-likeness” or “verisimilitude.” Specifically, Oddie argues that ACE violates the following reasonable principle: all else equal, shifting a positive increment of credence from a less truth-like proposition to a more truth-like proposition increases overall accuracy. While I grant this principle’s intuitive plausibility, I nevertheless think it is false. Its appeal, I maintain, can be explained in a way that does not depend on the concept of verisimilitude, but requires only notions grounded in accuracy as understood by ACE. For various ways of interpreting verisimilitude — such as error counting, approximate truth, or closeness of estimation — I will offer accuracy-based ways to explain the intuitive plausibility of Oddie’s principle. In the end, it will be seen that our judgments about verisimilitude (insofar as they make sense) derive from considerations of credal accuracy, and not the other way around. Indeed, a complete account of accuracy for credences has no need for the concept of verisimilitude at all.
Key words: strictly proper scoring rule, verisimilitude, credences, estimation, accuracy