A short literature review

About this post

Hello. It has been a long time since I've posted my chess data analysis. Most of the reason was that I lost ideas trying to come up with blog topics. I have been doing some literature reviews related to chess analysis so I could perform better in my posts. This paper is one of the articles I decided to review.

Short Comment

If you are interested in how to connect different chess openings, I highly recommend this paper. I was thinking of possible blog topics. This paper did a very good job of analyzing chess data to generate a chess opening similarity graph. The paper can be accessed here. The paper proposes a bipartite graph model to interpret chess data which was mind-boggling.

During the matrix multiplication to generate we do M^T M = W* in the article, maybe if you reverse the order of the matrix and do M M^T a similarity between player opening choices can be generated.
The paper chose bicm to remove all the noise related to the W* matrix but they could have removed some statistically insignificant noise beforehand but they didn't choose to do so.

I had some questions regarding some decisions.

• The data was extracted from blitz games. Blitz is the most played time control but maybe Rapid could have provided more insight considering the opening research
• games were filtered over 2000 rating which is not supported by good reasoning

However, they are not the game-changing factor for this research. The research also collected data for over a year, making it trustable.

Overall, this paper is amazing. I highly recommend reading it. It doesn't only end on visualizing chess similarity but goes on and predicts how a player's opening repertoire can change in the future.

Future posts

I will come back to blogging some chess data analysis. If there are any topics you wish me to analyze, please leave it in the comment!