Performing A* search on weighted graphs is particularly difficult because most admissible heuristics measure some form of distance to the goal, but fail to take into account the inherent threat that is attributed to an area of the graph. We design a heuristic which overcomes this by replicating the graph on a coarse, low-resolution grid which preserves the threat of the original graph and does not break admissibility. While this replica requires extra computation, the amount of time spent creating it is negligible to the time spent on the actual A* search. Due to this lack of pre-computation, our heuristic is best suited for large maps that have dynamic threat that changes frequently. One such example is maps from real-time strategy game StarCraft, in which enemy units which are constantly moving around the map pose a threat to the player. On these maps, our heuristic, compared to uniform cost search, popped 56% fewer nodes in 33% less time.