Francisco, Rodrigo Alexandre Nunes2026-01-152026-01-152025http://hdl.handle.net/10400.5/116624Tese de Mestrado, Engenharia Informática, 2025, Universidade de Lisboa, Faculdade de CiênciasProviding an appropriate challenge to a player is an extremely important component when designing a game since it may directly influences a game’s reception and longevity. In order to properly balance some game for a big demographic of player’s with different playstyles, various developers employ various techniques capable of adjusting the difficulty of the game dynamically, with some iteration proving more or less successful depending on the games genre. With this in mind our project presents a systematic approach to a Tower Defense game that iteratively adjusts specific attributes in enemy creeps over the course of a level depending on performance metrics taken from the player and characteristics of a level. By implementing a feedback-driven balancing loop using a multi-objective genetic algorithm, enemies can be dynamically tweaked inside each game-play session in order to promote new experiences that challenge the player while maintaining an appropriate game-play pace. Additional game elements such as DDA events are implemented in the game in a manner to fix moments of stagnant game-play and add to the replayability of the game. This method was then tested over multiple sessions with a diverse portfolio of participants, showcasing how this approach reacts to players with different skill levels and strategies. The goal study is to highlight how certain design elements inherent to Tower Defense games can be analyzed to properly rate a player’s performance and subsequently how said performance may be used to customize one’s experience, leading to a better overall product.application/pdfengStrategy GamesArtificial IntelligenceDynamic Difficulty AdjustmentGenetic AlgorithmsPlayer Performance Analysismaster thesis