Clustering Chess Playing Styles: A Machine Learning Approach
| dc.contributor.author | Priohastono, Hudaifah | |
| dc.date.accessioned | 2026-11-13T08:47:35Z | |
| dc.date.issued | 2026 | |
| dc.description.abstract | Mapping chess playing styles is crucial for tactical evaluation and strategic preparation. However, conventional methodologies rely heavily on subjective expert-based assessments, lacking measurable quantitative standardization. This study addresses these limitations through a data-driven approach using the K-Means clustering algorithm within an unsupervised learning framework. The dataset comprises 135 classical chess games from nine top-ranked global players during late 2025, with 15 games per player. Move data from Portable Game Notation (PGN) files were computationally extracted into nine quantitative statistical features representing aggressiveness, mobility, and spatial control. The dataset underwent preprocessing using StandardScaler for feature standardization. The optimal number of clusters was determined through Elbow Method and Silhouette Score analysis. Experimental results indicate that K=6 achieved the highest Silhouette Score of 0.2103, successfully mapping playing styles into six distinct behavioral profiles: very defensive, defensive, balanced, positionally active, aggressive, and very aggressive. The findings demonstrate that K-Means clustering effectively identifies latent patterns in top-tier chess game statistics, providing a valid, transparent, and quantifiable framework for player classification, even with a highly selective sample size | |
| dc.identifier.uri | https://repository.uii.ac.id/handle/123456789/67658 | |
| dc.language.iso | en | |
| dc.publisher | Universitas Islam Indonesia | |
| dc.subject | Chess | |
| dc.subject | Top Global Players | |
| dc.subject | K-Means Clustering | |
| dc.subject | Unsupervised Learning | |
| dc.subject | Playing Style | |
| dc.title | Clustering Chess Playing Styles: A Machine Learning Approach | |
| dc.type | Thesis |
Files
Original bundle
1 - 1 of 1