Weight interfaces enable users of Virtual Reality (VR) to perceive the weight of virtual objects, significantly enhancing realism and enjoyment. While research on these systems primarily focused on their implementation, little attention has been given to determining the weight to be rendered by them: As the perceived weight of objects is influenced not only by their absolute mass, but also by their weight distribution and prior expectations, it is currently unknown which simulated mass provides the most realistic representation of a given object. We conducted a study, in which 30 participants chose the best fitting weight for a virtual object in 54 experimental trials. Across these trials, we systematically varied the virtual objects’ visual mass (three levels), their weight distribution (six levels), and the position of the physical mass on the grip (three levels). Our Bayesian analysis suggests that the visual weight distribution of objects does not affect which absolute physical mass best represents them, whereas the position of the provided physical mass does. Additionally, participants overweighted virtual objects with lower visual mass while underweighting objects with higher visual mass. We discuss how these findings can be leveraged by designers of weight interfaces and VR experiences to optimize realism.