Probability density of Peirce skill scores from the Monte Carlo experiment (dashed curve) with 1σ error estimate (dotted curves) and from the variables considered for active region emergence (solid curve), using a bootstrap estimate. The left panel shows the results for sample sizes of 85; the right panel shows the results for sample sizes of 50. There is a clear tail of the distribution of the AR emergence variables to larger skill scores not present for the random variables, indicating that it is very unlikely that chance alone accounts for the performance of the best variables in distinguishing PE from NE regions. For the sample size of 85, the variables with skill scores above 0.27 almost certainly have a real ability to discriminate between PE and NE regions; for the sample size of 50, it is difficult to determine if any specific helioseismology variable has a real ability to discriminate between PE and NE regions, but the number of helioseismology variables with large skill scores ( 0.2) suggests that some can discriminate between the two.