In this paper we employ network analysis to re-assess competition policy within a macroprudential framework. Such an exercise seems to be relevant as it explicitly addresses a question posed forcefully by Haldane (2009), that is whether policy interventions can alter the topological network structure with the declared aim of improving network robustness. Here we concentrate on the idea that central banks and antitrust authorities have the opportunity to design the structure of the industry by choosing how banks are allowed to merge. Merges change the topology of the system for three reasons: 1) larger banks are formed as the summation of smaller ones; 2) the total number of active banks decrease; 3) large banks generally possess more connections than small banks. One can imagine that different competition policies (e.g., let just one very big bank to form by allowing it to acquire a large number of smaller banks; limit the size of each merger to just two small units, etc.) lead to different network topologies, which could in principle be characterized by different degrees of resilience to shocks. If this is the case, competition policy can be interpreted as an additional tool for macro-prudential regulation aimed at preventing systemic crises. We build an agent-based computational laboratory of an interbank network and employ three different types of M&A strategies as network-changing devices, in order to evaluate their effect on the resilience of the system. Our results suggest that topologies are not all alike: more specifically, it appears that a concentrated and yet asymmetric system is better geared to cope with an external shock. By contrast, concentrated and symmetric markets turns out to be in fact more fragile than a competitive one. The extent of the damage to the system depends on the exposure to interbank claims, the degree of connectivity, the structure of the network and capital requirements. In addition, we put forward the idea that capital requirements should be network-varying. Different shock-amplifying dynamics are observed because flat capital requirements force an inefficient allocation of net worth within the system. For example, it turns out that large banks are forced to hold too much capital whereas small institutions have less than what it is necessary and this misallocation renders the system less resilient. Once we introduce network-varying capital requirements, the robustness of the system improves and this aligns the performance of different topologies. The clear policy implication is that the regulator shall closely monitor the structure of the network and its evolution over time because policy on capital requirements is sensitive to it and one size does not fit all. We need to improve our effort towards the production of reliable and up-to-date data that allows us to map banking networks as precisely as possible.