Statistical significance of differences in rank distributions. Cα RMSD rank distributions for the best-scoring non-native structures for all functions are compared. P-values show the likelihood that better rank distributions for the function on the left are the result of chance. P-values less than 0.05 have been colored in red, showing statistically significant differences in these distributions. These ranks are among decoy structures only. The null hypothesis of this one-tailed Wilcoxon test is that neither distribution is lower than the other. The alternative hypothesis is that functions on the left achieved lower ranks for their best-scoring decoys than functions along the top.