Abstract : This study focused on enhancing the Glicko-2 algorithm in matchmaking systems to improve the responsiveness of rating deviation. The study addressed the algorithm's failure to recognize consecutive draws as an opportunity to adjust the user's rating deviation, overlooking important insights into performance progression. These consecutive draws were identified, and a Hidden Markov Model (HMM) was applied to classify them as certain or uncertain. The results were then used to refine the rating deviation with varying depths based on the HMM, providing a more accurate reflection of their impact on performance. Testing on 10,000 sequences of 15-20 matches, with an initial rating of 1500 and deviation of 150, showed that the enhanced algorithm outperformed the original algorithm. The average rating deviation (sigma) was reduced by 20.30% (59.746±0.409 vs. 74.969±0.053), indicating more stable performance assessments, which resulted in a 24.51% decrease in the average rating (mu) change per match (1.657±0.050 vs. 2.195±0.048). The accuracy of the enhanced algorithm's Hidden Markov Model classification was 93.81%, reflecting a marked improvement in performance evaluation. These results demonstrate that the enhanced algorithm's improved rating deviation reduced the need for large adjustments, leading to more confident and accurate rating updates. The findings suggest the enhanced algorithm provides a more reliable assessment of users' performance, enabling better matchmaking.

Keyword : Algorithm, Chess, Draw, Enhancement, Glicko-2, Hidden Markov Model, Matchmaking