Match each heuristic with the search it produces, assuming costs between nodes are variable. (That is, the cost between any 2 nodes is fixed, but. different nodes have different costs.) Order search by cost from orighin Only select moves that take us doser to goal Follow one path as far as possible: if we get stuck, back up until wu find another choice we haven't tried Keep the best N candidates; generate all children trom each; again select the best N candidates Order candidates by actual cost to that node plus heuristic cost from there to goal