Exploring the application of neural networks in complex card games

Identifier:  TFG:6975

Handle: https://hdl.handle.net/20.500.11797/TFG6975

Authors:  Llatser Torres, Jordi

Abstract:
The aim of this work has been to develop a tool based on Artificial Intelligence to learn how to make decisions in a card game as complex as Magic: The Gathering. The project is focused on a part of the game where a card must be chosen from a set of up to 15 cards from the global game set of about 300 cards, considering the selections that have been previously made. Machine learning methods based on neural networks have been used, taking advantage of data from real players making these same selections. A game data representation model has been designed and a neural network has been implemented using the Python programming language. In this project we have managed to train an artificial intelligence in the previously defined task, managing to imitate the card selection made by players using a testing dataset from known plays. The system has also been evaluated by some players, obtaining good qualitative results.