Analysis of Chess Opening Patterns Using Data Mining Methods to Determine Players’ Strategic Tendencies

Femasta Sembiring (Teknik Informatika, STMIK Kaputama, Binjai, 20714, Indonesia)
Rahmadani Rahmadani (Universitas Pembangunan Panca Budi, Medan, 20122, Indonesia)
Muhammad Fauzi Muhammad Fauzi (Teknik Informatika, STMIK Kaputama, Binjai, 20714, Indonesia)

Journal Of Artificial Intelligence And Software Engineering
Vol 6, No 2 (2026): Juni (OnProgress)

Online chess games generate large-scale match data that can be used to objectively analyze opening patterns. This study aims to cluster chess openings based on game statistical characteristics using the K-Means Clustering algorithm. The dataset was obtained from Lichess in PGN format, focusing on Rapid games played by players with ratings above 1200. The data were selected, preprocessed, transformed based on ECO codes, and reduced to the 300 most frequent ECO codes. The variables used were Win Rate, Draw Rate, and Average Moves, which were standardized using Z-Score. Testing was conducted using 3-, 4-, and 5-cluster scenarios and evaluated using the Davies-Bouldin Index. The DBI values for each scenario were 0.9538, 0.8715, and 0.7312. The best result was obtained by the 5-cluster scenario because it produced the smallest DBI value. The clustering results show that chess openings can be grouped into aggressive, solid, balanced, and special-characteristic tendencies.