This paper introduces BPR+, a novel loss function for graph self-supervised learning that extends Bayesian Personalized Ranking (BPR) by incorporating even-hop paths to better capture global connectivity and topological structure.