For your non-admissible heuristic demo, it might also be interesting to look at squaring the euclidian distance and the affect of biasing the algorithm in this way.