Effects of material advantage and space advantage on the Komodo and Stockfish chess engines

(1) Menlo-Atherton High School, (2) Department of Computer Science and Engineering, University of Michigan

https://doi.org/10.59720/23-131
Cover photo for Effects of material advantage and space advantage on the Komodo and Stockfish chess engines
Image credit: The authors

Chess engines are computer programs built to play chess. Currently, chess engines are significantly better at chess than the highest ranked human chess players. The utilization of chess engines is the most widely accessible, reliable, and efficient way to study chess. Because of their overwhelming chess capabilities, chess engines are able to aid chess players ranging from amateurs to Grandmasters. A key feature of chess engines is their ability to analyze a chess position and provide a numerical evaluation that describes the degree to which a side is winning. However, different chess engines employ different methods of evaluating a position. Some engines highly value “material advantage” and will greatly favor the side that has captured more pieces. Others put great importance in “space advantage” and will generally try to move their pieces from starting position to positions where they control more squares as early as possible. This study examines how two top-rated chess engines, Stockfish and Komodo, differ in their methodologies for evaluating a chess position. We hypothesized that Komodo’s evaluation would change more than Stockfish’s when material was lost, and that Stockfish’s evaluation would favor the side with more available advancing moves than Komodo. We discovered that Komodo indeed valued material advantage significantly more than Stockfish. However, the two engines did not differ significantly in their evaluations of space advantage. These findings have important implications for chess players, as they allow for accurate imitations of an engine’s play style, which could serve to more effectively increase performance.

Download Full Article as PDF

This article has been tagged with: