Analyzing the Valuation of Piece-Square States in Chess: A Machine Learning Approach

The Chessable Research Awards for the Fall 2023 cycle have announced their winners, and they are none other than undergraduate student Aditya Gupta and graduate student Denise Trippold. Aditya Gupta, in a guest blog post, delves into the research he conducted with the help of Dr. Nick Polson, Dr. Vadim Sokolov, and FIDE Trainer Shiva Maharaj. Aditya, who was pursuing dual enrollment in mathematics at the University of Illinois when he received the Chessable Research Award, will be starting as a freshman in computer science at Stanford University in the spring of 2024.

Aditya’s research focused on determining the relative value of a chess piece based on the square it is placed on, rather than just the commonly accepted point values. By combining the valuation of both the piece and the square, Aditya aimed to provide chess players with a more nuanced understanding of piece placement and its impact on the game.

Using Machine Learning methods, Aditya and his team analyzed a dataset of 2000 chess games played by grandmasters to empirically determine the valuations of different piece-square states. They then trained a neural network to predict the winning probability of a given piece on a specific square.

The results of their research showed that different squares on the chessboard have varying values for different pieces, reflecting common chess strategies such as the importance of central squares and the disadvantage of pieces on the edges. By formalizing these values, Aditya hopes to provide players with a new tool for analyzing and improving their gameplay.

This innovative approach to evaluating piece-square states in chess showcases the potential of Machine Learning in advancing the understanding of the game. Aditya’s work not only contributes to the field of chess research but also offers valuable insights for players of all levels looking to enhance their strategic thinking on the board.