With decreasing speed the film thickness in an EHL contact decreases. Below a certain speed asperity contact will take place and gradually the contact enters the mixed lubrication regime. Vice-versa, with increasing speed beyond a certain speed the film thickness has reached a level where asperity contacts have become so rare that the contact will be in the full film regime. It seems only logical to expect that the speed at which the first (or last) significant asperity interactions start to take place is influenced by the micro-geometry of the surfaces. In experiments performed on a two-disk rig under conditions of pure rolling, using one very smooth and one rough disk it was indeed observed that the “lift-off” speed defined as the speed above which full film lubrication prevails, differed significantly for surfaces with a different micro-geometry. The test results can be seen as ranking for the surface micro-geometries in terms of their film generating capability. In this paper the question is addressed if a ranking as observed in the tests can be predicted in advance, using the load conditions and the measured surface micro-geometry as input, without having to resort to full-scale numerical simulations of any sort. Based on the amplitude reduction formula proposed by Venner et al. (2000), for roughness in contacts under pure rolling a model is constructed that, given the load condition and a measured surface micro-geometry, determines the deformed micro-geometry and subsequently a measure of “probability of contact.” For a given contact this measure can be plotted as a function of speed to obtain a theoretical “lift-off” curve. For the different types of surface micro-geometries used in the tests such a curve is compared with the experimental results, showing a promising agreement in ranking.