Recent research has suggested that people make physical predictions based on extrapolation from a noisy representation of the world, which gives rise to a probabilistic distribution over possible future worlds. But can people use the uncertainty of their predictions to inform their decisions, or can people access only a single possible future? Here we demonstrate that confidence-sensitive decisions about the future track the amount of uncertainty expected from probabilistic forward extrapolations. Participants were asked to make predictions about where a ball would go and indicate an expected range around that prediction. This range was well correlated with two measures of uncertainty: variability in predictions across participants and the amount of uncertainty expected by a model of physical prediction. This suggests that people form a probabilistic distribution over possible futures in the course of physical prediction and base their decisions about the future on this range.