Identifying key players in a set of coupled individual systems is a fundamental problem in network theory [1][2][3]. Its origin can be traced back to social sciences and the problem led to ranking algorithms based on graph theoretic centralities [4]. Coupled dynamical systems differ from social networks in that, first, they are characterized by degrees of freedom with a deterministic dynamics and second the coupling between individual systems is a well-defined function of those degrees of freedom. One therefore expects the resulting coupled dynamics, and not only the network topology, to also determine the key players. Here, we investigate synchronizable network-coupled dynamical systems such as high voltage electric power grids and coupled oscillators on complex networks. We search for network nodes which, once perturbed by a local noisy disturbance, generate the largest overall transient excursion away from synchrony. A spectral decomposition of the network coupling matrix leads to an elegant, concise, yet accurate solution to this identification problem. We show that, when the internodal coupling matrix is Laplacian, these key players are peripheral in the sense of a centrality measure defined from effective resistance distances. For linearly coupled dynamical systems such as weakly loaded electric power grids or consensus algorithms, the nodal ranking is efficiently obtained through a single Laplacian matrix inversion, regardless of the operational synchronous state. We call the resulting ranking index LRank. For heavily loaded electric power grids or coupled oscillators systems closer to the transition to synchrony, nonlinearities render the nodal ranking dependent on the operational synchronous state. In this case a weighted Laplacian matrix inversion gives another ranking index, which we call WLRank. Quite surprisingly, we find that LRank provides a faithful ranking even for well developed coupling nonlinearities, corresponding to oscillator angle differences up to ∆θ 40 o approximately.