A comparative study of the interplanetary magnetic field (IMF) polarities inferred from ground-based measurements of the cosmic ray north-south (N-S) anisotropy (CR polarities) and of the geomagnetic field at high latitudes (GEOM polarities) is presented. The reliability of the two methods is tested for different solar, interplanetary, and geomagnetic scenarios using a large set of plasma and magnetic field data taken by the Earth's satellites in the period from 22 May 1965 to 31 December 1999. In general, the GEOM method gives better predictions than the CR method (the overall success rates of the two methods being 92% and 76%, respectively, for the investigated period). Nevertheless, we single out scenarios in which the reliabilities of the two methods become comparable. In fact, while the GEOM polarities are largely unaffected by the solar, interplanetary, or geomagnetic conditions, the CR polarities strongly depend on several parameters (e.g., the solar wind bulk velocity, the IMF intensity, along with the solar activity phase). Thus our analysis points to the criteria which will allow us to select the most reliable IMF polarity at any time, to be used in any study concerning solar, interplanetary, or terrestrial phenomena.