We're testing whether narrow, well-picked lists really do beat wide ones. So far the answer looks like yes, but the sample is small.