The overlap properties of ∼850 snapshots of convective cloud fields generated by a cloud‐resolving model are studied and compared with previously published results based on cloud radar observations. Total cloud cover is overestimated by the random overlap assumption but underestimated by the maximum overlap assumption and two standard implementations of the combined maximum/random overlap assumption. When the overlap of two layers is examined as a function of vertical separation distance, the value of the parameter α measuring the relative weight of maximum (α = 1) and random (α = 0) overlap decreases in such a way that only layers less than 1 km apart can be considered maximally overlapped, while layers more than 5 km apart are essentially randomly overlapped. The decrease of α with separation distance Δz is best expressed by a power law, which may not, however, be suitable for parameterization purposes. The more physically appropriate exponential function has slightly smaller goodness of fit overall but still gives very good fits for Δz between 0 and 5 km, which is the range of separation distances that would be of most importance in any overlap parameterization for radiative transfer purposes.