The paper "How Many Bootstrap Replicates Are Necessary?" by Pattengale et al. addresses the issue of determining the optimal number of bootstrap replicates required for accurate phylogenetic tree inference using Maximum Likelihood (ML). The authors propose two stopping criteria, the Frequency Criterion (FC) and the Weighted Robinson-Foulds distance-based Criterion (WC), to determine when enough replicates have been generated. These criteria are designed to be computed at runtime and do not rely on external reference trees. The study uses 17 diverse real-world DNA datasets, ranging from 125 to 2,554 sequences, to assess the impact of the number of replicates on the quality of support values. The results show that the stopping criteria typically stop computations after 100-500 replicates, producing support values that correlate better than 99.5% with reference values on the best ML trees. The study also highlights that the stopping criteria can recommend different numbers of replicates for different datasets of comparable sizes, emphasizing the dataset-specific nature of the convergence properties of bootstrapping. The authors conclude that their proposed criteria provide a practical solution for determining the necessary number of replicates, making robust bootstrapping under ML inference computationally feasible for most datasets.The paper "How Many Bootstrap Replicates Are Necessary?" by Pattengale et al. addresses the issue of determining the optimal number of bootstrap replicates required for accurate phylogenetic tree inference using Maximum Likelihood (ML). The authors propose two stopping criteria, the Frequency Criterion (FC) and the Weighted Robinson-Foulds distance-based Criterion (WC), to determine when enough replicates have been generated. These criteria are designed to be computed at runtime and do not rely on external reference trees. The study uses 17 diverse real-world DNA datasets, ranging from 125 to 2,554 sequences, to assess the impact of the number of replicates on the quality of support values. The results show that the stopping criteria typically stop computations after 100-500 replicates, producing support values that correlate better than 99.5% with reference values on the best ML trees. The study also highlights that the stopping criteria can recommend different numbers of replicates for different datasets of comparable sizes, emphasizing the dataset-specific nature of the convergence properties of bootstrapping. The authors conclude that their proposed criteria provide a practical solution for determining the necessary number of replicates, making robust bootstrapping under ML inference computationally feasible for most datasets.