It is well-known that the mean absolute error of National Hurricane Center intensity forecasts for Atlantic basin tropical cyclones calculated with respect to the best-track analyses shows little or no decrease since 1990. It is often inferred from this result that the quality of these intensity forecasts has not improved over the past two decades. However, this conclusion implicitly assumes that the statistical characteristics of the best-track intensity analyses have not changed over the years. Here, we demonstrate that the relative frequency distribution of best-track intensity change (e.g. the change in best-track intensity over a 24 h window) evolved substantially during the 1990-2010 period. In particular, the relative frequency of large-magnitude intensity changes increased significantly and the relative frequency of small-magnitude intensity changes decreased significantly, resulting in a significant increase in the average absolute intensity change (AAIC). The AAIC is equivalent to the mean absolute error of persistence intensity forecasts, and as such it can be used to define a baseline for an intensity forecast skill score. Such a skill score accounts for the changing difficulty of intensity predictions, as represented by the AAIC. For the National Hurricane Center Atlantic basin intensity forecasts during the 1990-2010 period, the AAIC-based skill score has a statistically-significant increasing trend, suggesting there has been real improvement in intensity forecasts on a decadal time scale.