Planning an endgame move set for the game RISK: a comparison of search algorithms

Seven algorithms used to search for solutions in dynamic planning and execution problems are compared. The specific problem is endgame moves for the board game RISK. This paper concentrates on comparison of search methods for the best plan using a fixed evaluation function, fixed time to plan, and randomly generated situations that correspond to endgames in RISK with eight remaining players. The search strategies compared are depth-first, breadth-first, best-first, random walk, gradient ascent, simulated annealing, and evolutionary computation. The approaches are compared for each example based on the number of opponents eliminated, plan completion probability, and value of ending position (if the moves do not complete the game). Simulation results indicate that the evolutionary approach is superior to the other methods in 85% of the cases considered. Among the other algorithms, simulated annealing is the most suitable for this problem.