The power of natural selection through survival of the fittest is nature's ultimate tool for the improvement and advancement of species. To apply this concept in catalyst development is attractive and may lead to more rapid discoveries of new catalysts for the synthesis of relevant targets, such as pharmaceuticals. Recent advances in ligand synthesis using combinatorial methods have allowed the generation of a great diversity of catalysts. However, selection methods are few in number. We introduce a new selection method that focuses on the stability of catalytic intermediates measured by mass spectrometry. The stability of the intermediate relates inversely to the reactivity of the catalyst, which forms the basis of a catalyst-screening protocol in which less-abundant species represent the most-active catalysts, 'the survival of the weakest'. We demonstrate this concept in the palladium-catalysed allylic alkylation reaction using diphosphine and IndolPhos ligands and support our results with high-level density functional theory calculations.