The threshold of the filtering heuristic is directly correlated with the probability of finding the true best-fit model (Heuristic accuracy) and inversely related to the number of models for which we avoided the likelihood calculation (Computational savings).