Analizador de posiciones del tablero del Go

Thumbnail Image
Publication date
Defense date
Journal Title
Journal ISSN
Volume Title
Google Scholar
Research Projects
Organizational Units
Journal Issue
Throughout humankind's history games have been a defining part of it. For humans, games are a unique phenomenon. They are a challenge established within a defined set of abstract rules, an example of what humans desire: overcoming obstacles and going forward. These challenges are not required for survival, solving and mastering them are its own rewards. A great amount of games have been created during our history. Games are extremely varied, difficult, sophisticated, simple. The world nowadays has a lot of games to offer and one of its main categories is tabletop games. Again, most of them have been created and popularized throughout history. One of the better known games is Go. Go's main feature is being fairly simple in terms of rules which can be explained within a short time. However, it may take a whole life for a normal person to master these rules to their maximum extent. This makes it very interesting and peculiar because it is one of the most difficult games ever created. It is a very complex game given its age, and after 2000 years it is still being studied. In chess, another tabletop game, automatic players are able to play at a grandmaster level by using techniques based on the exhaustive exploration of possible moves, like min-max search with alpha-beta pruning, with the help of carefully designed evaluation functions. This approach is significantly less useful with Go. Go has too many possibilities in terms of movements so the approach taken in Chess gives far less advantage with Go. Go is a complex game for humans to play, but it is even harder to play for computers. Despite its age there are ongoing investigations about Go's computer analysis. No optimal strategy has been found yet for Go. It is one of the games that has not been solved yet and it is considered one of the most difficult to handle. Any effort made in this area is important because of its complexity. Any valuable addition will push further investigation forward. The motivation of this project is to improve our knowledge of the computational analysis of Go. Following the reasoning made before, a Go tool will be developed. This tool will be able to analyze any given Go board using an algorithm known as Monte-Carlo Tree Search. Some playing agents used in Go competitions use this algorithm. However these agents use MCTS only to play, so this project will take a different take on this algorithm. MCTS will be used to retrieve information about influence and other different features of a Go board. Furthermore, the tool developed will be used in order to further analyze information retrieved with machine learning techniques.
Algoritmos, Juegos de ordenador, Método de Monte Carlo
Bibliographic citation