Rank-based statistics can be improved by considering both ranks and signs, even without assuming symmetry. This new approach allows for more efficient inference in statistical models. The signs contribute to the variance, potentially increasing efficiency in regression and time series analysis.