Graph illustrating the “hard cut-off” approaches of Isler et al. (1998) and Tobias et al. (2010). The y-axis shows the score attributed under the relevant system, weighted for 1 = diagnosability. The x-axis shows effect sizes. In addition to division by three, the Tobias et al. (2010) scores are treated more conservatively by assuming that bare pooled effect sizes are equivalent to controlled unpooled effect sizes. The scores for effect sizes at the lower end of the graph are somewhat artificial with a starting score of 0.34. This is based on the lowest recorded controlled unpooled effect size which passed a statistical significance test for biometrics (see Table 20). In reality, some lower differentiation with larger samples will be scored and some higher variation with lower samples will not be scored at all: see Tables 20–21.