Pre-retrieval query performance prediction (QPP) estimates a query’s effectiveness before retrieval, enabling tasks such as query routing. Prior pre-retrieval methods built on hand-crafted indicators treat queries independently and often scale poorly across collections and metrics; recent learning approaches often optimize relative difficulty without calibrated absolute estimates and still rely on single-query evidence, while neighborhood models assume a query behaves like the average of its nearest neighbors and cannot exploit multi-hop structure or handle heterogeneous neighborhoods. We propose a relational formulation that treats effectiveness as a smooth function over a query–query affinity graph structure, learning how information should propagate across related queries to produce context-aware predictions and extending to unseen queries by situating them within this graph structure. Our proposed method consistently outperforms classical and learning baselines, including higher rank correlations on MS MARCO Dev (Spearman \(\rho =0.409\) ) and DL-Hard (Spearman \(\rho =0.440\) ), and competitive results on TREC DL 2020 (higher Pearson and Spearman), indicating that our approach provides an effective basis for QPP.