Algorithms that rank items using paired comparison data are becoming increasingly widely applicable. New data is always being gathered to try to improve their ability to make predictions. I demonstrate that it is possible to seek data in such a way that the predictive ability of algorithms will improve without attention to the particular content of the data, present two methods of doing so, and discuss the kinds of algorithms to which they are applicable. In addition, I introduce an alteration to the Colley Matrix [6] rating algorithm that allows it to include equally-matched comparisons (ties) in a meaningful way.