Earticle

초록

한국어

Starting from an empty board, the number of legal positions in a Go game grows exponentially as the number of moves increase. During the early moves, a very large subset of the possible positions have balanced win rates. This suggests that the choice of the opening is not crucial to decide a Go game. Nevertheless it is important for players to know what kind of openings they will most frequently face. To answer this question, we collected half a million games played on Fox Go server from 2013 to 2019 amongst players ranked between 10 kyu and 9 dan. A data-informed methodology was developed to classify and compute the frequency of the different openings before and after the introduction of AI. The most popular first moves are nirensei and hoshi+34. Deeper openings popularity depends on rank. Sanrensei fre-quency decreases with rank, Chinese opening is most popular below 5 dan and Kobayashi between 3 dan and 7 dan (Fox Go Server ranks). The results also show a change in opening playstyle for players ranked above 4 dan after the rise of Go AI. Up to 2019, a similar change is not observed amongst lower ranked players.

목차

Abstract
Ⅰ. Introduction
Ⅱ. Methods
1. Data
2. Expert-informed pattern search
3. Data-driven pattern search
4. Statistical analysis and uncertainty quantification
Ⅲ. Results
1. Corner is gold (4 moves openings)
2. Side is silver (5+ moves openings)
3. Influence of AI on popular openings
Ⅳ. Discussion
Ⅴ. Conclusions
References

키워드

저자

• Quentin Rendu [ Hamburg University of Technology (TUHH) ]
• Carlos G Urzúa-Traslaviña [ Independent Researcher ]
• Stephen Hu [ Qilue Weiqi ]

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  바둑학연구 [Journal of Go Studies]

• 간기

  반년간

• pISSN

  1738-3730

• 수록기간

  2004~2026

• 십진분류

  KDC 691 DDC 794