Not with outliers, no, I think it handles them particularly well (conservatively.) But yes to small consistent differences (e.g. uniform 1%) that affect the rel

Not with outliers, no, I think it handles them particularly well (conservatively.) But yes to small consistent differences (e.g. uniform 1%) that affect the relative order of results but not by an very important amount. So have to consider effect size even when the test statistic is strong.