Anthropic activities impact ecosystems worldwide thus contributing to the rapid erosion of biodiversity. The failure of traditional strategies targeting single species highlighted ecosystems as the most suitable scale to plan biodiversity management. Network analysis represents an ideal tool to model interactions in ecosystems and centrality indices have been extensively applied to quantify the structural and functional importance of species in food webs. However, many network studies fail in deciphering the ecological mechanisms that lead some species to occupy the most central positions in food webs. To address this question, we built a high-resolution food web of the Gulf of California and quantified species position using 15 centrality indices and the trophic level. We then modelled the values of each index as a function of traits and other attributes (e.g., habitat). We found that body size and mobility are the best predictors of indices that characterize species importance at local, meso- and global scale, especially in presence of data accounting for energy direction. This result extends previous findings that illustrated how a restricted set of traitaxes can predict whether two species interact in food webs. In particular, we show that traits can also help understanding the way species are affected by and mediate indirect effects. The traits allow focusing on the processes that shape the food web, rather than providing case-specific indications as the taxonomy-based approach. We suggest that future network studies should consider the traits to explicitly identify the causal relationships that link anthropic impacts to role changes of species in food webs.