Procedural content generation and evolving game content based on player preferences, performance, and behavior has been an ever-growing research topic in the field of computer science and game design. In this research, we used genetic algorithms to emphasize the distinct unforgiving nature of roguelike games in real time, while the player is in-game, without designing the levels by hand. This will lead to ever changing interactions and inputs from the player and the game environment, having only the game's mechanics and genetic algorithm, besides the player's input, dictate the pace of the game thus influencing the game's dynamics. To demonstrate our proposition, we designed an experimental turn-based 2D roguelike game called Fireflies. The core concept of game is to have the probability of the generation of hostile enemies with offensive elements being stronger against that of the player's defense increases as the player progresses through the game.