Understanding why some species persist while others vanish remains a central challenge in ecology. Traditional approaches either rely on mechanistic models that are conceptually clear but hard to apply to real-world scenarios, or on statistical tools that lack generality and interpretability. Here, we introduce a framework rooted in the statistical physics of complex systems but designed for ecological applicability. At its core is a single measurable quantity - the competitive balance - that synthesizes the processes shaping persistence: dispersal diversity, interspecific competition, and abundance. Without requiring direct measurement of species traits or detailed dispersal parameters, the metric captures how the spatial structure and heterogeneity of communities translate into species-level vulnerability. Our analyses reveal a fundamental trade-off: for a given rarity, greater heterogeneity in dispersal strategies reduces vulnerability. We validate the framework with tropical forest data, but its scope extends across ecosystems. By linking statistical physics with functional perspectives on biodiversity, this approach offers an interpretable and scalable tool to assess species vulnerability in the Anthropocene.