Whale Co-Ownership Network
A fund-by-stock network projected to stock-to-stock co-ownership, revealing which names cluster in the same institutional hands.
Category: Whales
What it measures
Each market-quarter is built as a fund-by-stock bipartite graph, then projected onto a stock-to-stock graph where the edge weight between two names is the number of funds that own both.
Formula
The live field shown on the signal board is weighted_degree. The full table also stores degree, eigenvector_centrality, and community_id. - Build bipartite graph B(fund, stock), with an edge when a fund holds a stock - Project to stock graph G(stock_i, stock_j), edge_weight = count of funds holding both stocks - degree = number of stock-to-stock neighbors - weighted_degree = sum(edge_weight) across the stock's projected edges - eigenvector_centrality = centrality score weighted by connected important neighbors - community_id = modularity community from the projected stock graph
Theory and robustness
High centrality means a stock sits at the heart of crowded institutional clusters, which can amplify shared-exit risk. Clustering quality is scored by modularity, and stability is checked by re-running at an edge-weight threshold plus or minus one and requiring a Rand index above 0.7.