The puzzle forecast: Tutorial analytics predict trial and error

Puzzle tutorials are designed to teach puzzle-solving skills. For game designers, the difficulty is predicting if puzzle challenges will present players with opportunities for learning with trial and error. We aim to empower designers with tools and techniques for making those predictions by analyzing the goal chains inherent to good designs. We study PuzzleScript, an online game engine that has made the source code of high-quality puzzle tutorials available. Research on puzzles has yielded algorithms that can generate playtraces of solutions. However, until now the importance of failure traces has been mostly overlooked. As a result, there is a lack of tools with analytics that can help assess challenge. To deliver them, we propose a novel approach that enriches playtraces with verbs. We introduce TutoScript, a language for expressing goal chains in terms of verbs. By combining TutoScript with well-known search algorithms, and by mapping rules to verbs, TutoMate can enrich, analyze and visualize generated playtraces of solutions, failures and dead ends. Two case studies on Lime Rick and Block Faker demonstrate how it helps to analyze simple goal chains, and can also detect broken tutorials. Our solution takes a promising step towards generic techniques for analyzing and generating tutorials.

• View at Publisher

Free Full Text ( Final Version , 754kb )