The aim of the project is to investigate the possibilities of computer analysis of databases of Go game (Baduk, Weichi) records using computer data-mining algorithms.

What is here for players?

GoStyle Web App is a web application which analyzes your Go game records, determines your playing style, strength, and gives various study hints, including a list of relevant professional players to review and book recommendations.

Research publications and online materials

• Paper about machine learning part of our methodology:
  Evolving Non-linear Stacking Ensembles for Prediction of Go Player Attributes, will be published in Proceedings of IEEE SSCI, 2015.
  Pre-print is available from arXiv:1512.09254. Or have a look on the SSCI presentation.
• Paper about feature extraction part of our methodology. If you want to cite us, please cite this paper!
  Evaluating Go game records for prediction of player attributes, Proceedings of IEEE CIG, 2015.
  Pre-print can also be found on arXiv:1512.08969.
• Style Questionnaire – If you are a strong player, you can help us by assessing the playing style of selected professionals.
• Results of the Style Questionnaire (pdf), compiled 27. May 2013 — we asked some experts (strong amateur players or professionals) to determine style of selected top ranked professionals (see the questionnaire right above). We use this data to feed the algorithms in the GoStyle Web App.
• Evaluation of the Go Game Records – Presentation given at the European Go Congress 2013 Science Conference, you can have a look at the slides.
• My Master thesis Meta-learning methods for analyzing Go playing trends (pdf) gives a thorough view of the whole subject.
• You can also have a look on our ancient slideshow summary, and the original paper which laid the basis of the methodology we use today. Or just browse the (old) source code.

What is new here?

• 4.2.2016 – Upgraded the server, this should fix occasional "Server Error" problems.
• 15.1.2016 – Added a blogbost about CNN for predicting strength. This is very exciting as it might improve the webapp accuracies a lot while potentially requiring smaller game sample. Stay tuned!
• 11.1.2016 – Updated the Research Section.
• 7.1.2014 – Web App: Recommendation of professional player now links directly to the database of games of the pro, using Waltheri's go pattern search.
• 6.1.2014 – I've done a facelift to make the stuff here nicer and smartphone friendly. I have also improved clarity of the texts.
• 4.1.2014 – Added online (html) versions of our papers.
• 20.12.2013 – Major extension, added study material (Go books) recommendations to the Web App. I also improved KGS support.
• 27.5.2013 – I published the styles of the selected professionals we feed to the algorithms as a standalone document. See the Research Section.
• 13.4.2013 – Public launch of the project web, the Web App and also the questionnaire.