Quantum spin chains can be represented in several ways, and eigenstates in these dual models have different entanglement structures. By characterizing the dual theories, the optimal tensor network algorithm for simulating the models has been found.