This study presents an approach to developing a test environment and constructing decision models for the puzzle game Wordle using reinforcement learning (RL) techniques. The research aims to create a neural network (NN) capable of solving Wordle, leveraging RL to predict words based on prior guesses. The study focuses on the process of puzzle solving as it pertains to RL and explores a NN designed for word prediction, utilizing information theory (IT) principles such as entropy. Additionally, an algorithm based on these concepts is proposed. The novelty of this work lies in the development and comparative analysis of NN models against mathematical models constructed with IT, offering insights into their effectiveness in puzzle-solving tasks.