Early Estimation of Level Difficulty in Mobile-Games

Abstract

In this report, we investigate several approaches to estimate quantitatively the level of difficulty of mobile games by using data analysis and learning methods. We base our study on data for the game Bubble Witch 3 Saga from the company KING. By defining key parameters and variables of the problem and how they depend on the given data sets, we propose a clear strategy for modelling the game difficulty. We use a probabilistic approach that suggests routes to improve “know-how” on the effect of specific features on the game difficulty, and this can be useful in the design process. In addition, the data shows that the probability of passing a level depends on the number of attempts performed by the player, which implies the existence of a learning process. We use this feature to analyze the probability of passing a level.