Measuring Difficulty of Novel Clockwork Puzzle Instances Using Evolutionary Algorithms

This study debuts the clockwork problem a novel type of combinatorial puzzle; the rules and general variations of the puzzle are explained in order to showcase the puzzle's interesting properties and motivate interest in the puzzle. Following this, the “Do What's Possible” representation for an evolutionary algorithm is used to solve instances of the clockwork puzzle. The results of the evolutionary algorithm across many instances are analyzed to create a difficulty measure for said instances. This type of computational intelligence is preferred as the evolutionary algorithm adequately mimics the approach of a human player who lacks prior knowledge about the puzzle.