In moving block systems for railway transportation a central controller periodically communicates to the train how far it can safely advance. On-board automatic protection mechanisms stop the train if no message is received during a given time window. In this report we consider as reference a typical implementation of moving-block control for metro and quantify the rate of spurious Emergency Brakes (EBs), i.e.~of train stops due to communication losses and not to an actual risk of collision. Such unexpected EBs can happen at any point on the track and are a major service disturbance. Our general formula for the EB rate requires a probabilistic characterization of losses and delays. We derive an exact formula for the case of homogeneous and independent packet losses and we use the results of this analysis to design an efficient Monte Carlo method that takes into account correlated losses due to handovers. We validate our approach via discrete-event simulations, including simulations with ns-3 for which we have developed additional modules for train systems. Our approach is computationally efficient even when emergency brakes are very rare (as they should be) and can no longer be estimated via discrete-event simulations.