1. The document proposes a method called "consequence driven decomposition" to identify dangerous contingencies in large power systems when there are too many potential contingencies to evaluate individually. 2. The method works by randomly selecting an initial set of contingencies to evaluate, identifying the most severe, and using this information to select a new set with a higher likelihood of uncovering new dangerous contingencies. 3. It is shown to outperform traditional Monte Carlo sampling by having a higher probability of finding a set number of dangerous contingencies within the available computational resources.