Rating Logic Puzzle Difficulty Automatically in a Human Perspective

Abstract

Logic puzzle games like Sudoku are getting popular for they are flexible in playing time and space and are useful in education. For puzzles, difficulty is arguably one of the most important factors in problem design. A problem too easy is boring, yet a problem too hard is frustrating. Providing problems with adequate difficulty to avoid boredom or anxiety is thus an important issue. In this paper we rate difficulty level of Sudoku problems with human oriented, general difficulty criteria so that the method can be used to evaluate problems of most logic puzzles. Only few previous Sudoku difficulty research are based on real playing data and the rating methods are limited to Sudoku or at most, constraint satisfaction problems (CSP). We found that the proposed method, despite of its simplicity and generality, can sort Sudoku problems in an order similar to average player solving time, the player perceived difficulty.