A common problem in forensic anthropology is the pair-matching of left and right bone antimeres. Several osteometric sorting models have been proposed in the last 20 years, with a recent acceleration in new methodological articles bringing sometimes contradictory results or recommendations. These debates demonstrate the need for a statistical tool both accurate and easily applicable by the final user. We present here an approach based on quadratic discriminant analysis (QDA). This approach is evaluated on antimeric pairs of humeri and femora from the openly available Goldman Data Set, and compared to two classical and previously published methods for osteometric pair-matching, based respectively on linear regressions and t-tests. It is shown that QDA globally outperforms existing solutions for reassociating those long bones, in particular by rejecting fewer true bone pairs at the classical α level of 0.10. The accuracy of all three methods is analyzed through receiver operating characteristic (ROC) curves to assess the influence of the choice of a decision threshold. The application on archaeological commingled remains of pair-matching models learned on a modern reference multipopulation sample is discussed. Finally, an R package containing the functions used for this study, bonepairs, is publicly available online. This ensures the full replicability of results and an easy use of the new method introduced here.