Please use this identifier to cite or link to this item:
Title: Classification of Chess Games and Players by Styles Using Game Data
Authors: Jayasekara, M.G.P.B.
Issue Date: 26-Jul-2021
Abstract: In a strategic game like chess, players have perfect information and there is no involvement of chances or physical skills. Even though the game is played on a board, the entire game runs on the two players’ minds. Players have to think smart and make plans to win the game. But, the way two players think about the same situation could be completely different. Therefore their approach to playing could be different from each other. Due to this very reason, there are recognized well-known playing styles which are bound to different persons. Current well-known styles such as Aggressive, Positional, Solid, Defensive etc. are introduced by chess experts with the help of their experience and expertise. However, these are not of an outcome of any scientific research. Therefore, in this research, I’m exploring if these well-known styles can be scientifically and logically separable and if those can be predicted looking at game data. Since the raw data (i.e. moves) of a chess game does not provide any direct information about the game’s nature, but only contains notations of a sequence of moves, to extract features from a game, those notations should be understood and certain preprocessing steps are required. Chess game engines can be used for this. Raw game data can be fed to chess game engines and information regarding states of the game can be obtained at each move. Such data can be collected for all or selected steps of a game and then that information can be used to construct features of the game. This process was followed in this project and a set of features were extracted from thousands of games. Those extracted data were then clustered and identified the natural clusters in those. In this research, I prove that some of those well known styles are separable from others and some are overlapping. I also prove that prediction models can be created to predict such styles looking at the game data.
Appears in Collections:2018

Files in This Item:
File Description SizeFormat 
2014MCS033.pdf2.34 MBAdobe PDFView/Open

Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.