Overconfidence is considered to be one of the most prominent behavioral biases, and has been studied with students, executives, and investors using different variants of a confidence interval task with large intervals of 80% or 90%. On average this research has found that individuals are overprecise, leading this type of overconfidence to be labeled as the "most robust form of overconfidence" (Moore et al., 2015). We elicit both large (80%) and small (20%) confidence intervals regarding the future price of a simulated asset. We find that only 18% of individuals are consistently overprecise, compared to 28% which are consistently underprecise. At the same time, and in line with the literature we replicate overprecision for the large interval. This suggests that aggregate overprecision may be a consequence of using large rather than small confidence intervals.