A model experiment is presented for thequantitative selection of relative scientometric impact indicators used in evaluating the scientific impact of papers. The Relative Subfield Citedness (Rw) indicator proved to be the most appropriate according to the criteria chosen. RW increases with the number of citations to the papers and, in contrast to other relative impact indicators, does not decrease if an author chooses to publish most of his papers in journals with large impact factors or if most of the citations to his papers are to the ones in journals with the largest impact factors.