The Adaptive Difficulty Level in a Hyper-Casual Game Through Facial Expression

One type of game that is popular and growing today is Hyper-casual. Even though it looks easy to play, some players may find it very difficult or frustrating if they are not matched with qualified skills to play this game. Moreover, this genre has a low retention rate. Therefore, there needs to be a way to keep the game going so players don’t find it too difficult or too easy when playing this hyper-casual game. This study aims to create an adaptive difficulty level system at the game level based on the player’s facial expressions. A Deep Learning Algorithm performs facial expression recognition called Deep Convolutional Neural Network (DCNN) model previously trained on the FER Plus dataset. From facial expressions that the system can recognize, it will affect the level of existing game obstacles. Changes in the player’s facial expressions during the game will be able to make dynamic changes at the game level. We use the Thirsty Ghost, similar to the Flappy Bird game, as an object to implement the adaptive difficulty level. Playtesting was done by comparing games with adaptive difficulty levels with identical games without adaptive difficulty levels. The questionnaire results show the Thirsty Ghost Game with the proposed adaptive difficulty level is preferred because it is easier to play. Players also can achieve better scores in the game with adaptive difficulty rather than without this system.