This paper presents a formulation for a method for the adaptive selection of angular flux expansion orders for use in COarse MEsh radiation Transport (COMET) method solutions to whole-core reactor problems. An important aspect of the COMET method is an assumed angular flux expansion on mesh interfaces. Previously, this expansion was held constant throughout a problem. However, the adaptive method described in this paper chooses the angular flux expansions automatically and allows them to vary between meshes. To demonstrate the method, a pressurized water reactor benchmark problem with UO2 and mixed oxide fuel assemblies is solved. Three different configurations for different insertions of control rods were considered. For all configurations, the agreement between the standard and adaptive COMET solutions was excellent, with eigenvalue agreement being 2 pcm or less and average pin fission errors never exceeding 0.1%. Increases in computational efficiency by factors of 2 to 2.6 were observed over standard COMET solutions employing the full flux expansion considered in the problem. In addition, a lower flux expansion suggested by literature as well as the results of the adaptive calculation was used in the standard COMET method to solve the problem. The adaptive COMET solution has a run time similar to this lower expansion, which is to be expected since many of the flux expansions chosen with the adaptive method match this lower flux expansion. The results of this study are encouraging and imply that adaptive COMET solutions improve upon the standard method by increasing computational efficiency when a flux expansion is used that is higher than required for desired accuracy. The method also limits the need for intuition and numerical experimentation in achieving flux expansions that result in COMET calculations that achieve satisfactory accuracy.