Fig. 6

Advising accuracy using oracle sets. This figure plots average advising accuracy using oracle advisor sets with different estimators, at varying set cardinalities. The horizontal axis is the cardinality of the advisor set: the number of parameter choices from which the advisor selects. The vertical axis is average true accuracy of the parameter advisor, where the average accuracy within each difficulty bin is then averaged across bins. The curves plot advisors using: the Facet estimator augmented by predicted coreness, the original Facet estimator with no augmentation, Facet augmented by TCS column quality scores, Facet augmented by ZORRO quality scores, and using TCS as the estimator. The dashed black curve is Facet augmented by true coreness: the limit attained with a perfect coreness predictor. As baselines for comparison, the dashed grey lines are the average accuracies of the standard aligners MAFFT, MUSCLE, and Clustal Omega, under their default parameter settings