Miai in the fuseki statistical analysis

    Keywords: Opening

Moved from Miai in the fuseki

The position under discussion  

Dave Sigaty:

Update - When I wrote the comments on miai in the fuseki I was basing it on materials in Otake Hideos Fuseki Theory which was published at the end of 1986. The final sentence turns out to be rather fateful since it seems the pros have meanwhile demonstrated that the left and right side here are not miai!

Analyzing this variation using the program Kombilo and the game collection from the Go Games on Disk CD (January 2002 update) the following history of the variation can be charted:

  • First appearance November 1983 in the Honinbo League game between Takemiya (Black) and Takagi Shoichi.
  • Appeared a total of 65 times between 1983 and 1995.
  • In those games White played in the middle of the right side 42 times and won exactly half of the games.
  • White also played on the left side in 16 games but won only 25% of such games.
  • Various other moves were played 8 times with no special trend in results.

The variation does not appear in the collection at all in the period 1996-2000. It was played once in May 2001 but this appears to have been an isolated incident.

It appears from this that the white play on the right side was superior to the alternative choice on the left. However, once White played on the right, Black could not do better than equal results. Since it is Black that initiates this line, the disappearance from top level professional Go after 1995 seems to reflect Black's dissatisfaction with his results.

One last interesting point. The left and right were officially viewed as miai from the first (hence Otake's inclusion of it as an example of miai in a book for amateurs written in 1986!). However, the choices made between the left and right were not balanced in actual play. The professionals first concentrated on splitting the right side. The play on the left came later (because the right side failed to yield an advantage?). Of the 16 games where White played on the left, 14 were played in the 90's and 2 were played in 1995 shortly before the variation fell out of favor.

Andre Engels: Using the database from MasterGo, I interestingly get to somewhat different results:

The 'first game' is not in that database, the first game in the database with the position is from 1985.

From 106 games, the right is played 61 times, the left 32 times. There is a third somewhat important move, namely a with 6 occurences. Unlike the GoGoD database, MasterGo shows the left (white wins 41%) not to be worse than the right (43%) in winning percentage, although in number of times played the datasets are similar. The other moves score lower (a scores 17%, the other moves combined 14%).

Most important difference is that the sequence DOES occur after 1995 - there is 1 game from 1996, w from 1997, 2 from 1998, 1 from 2000 and 1 from 2001.

Also, there is much less of shift from right to left than in GoGoD. For the period 1985-1989 I get right 25, left 5 (actually 26 and 6, since 2 other moves were played once, one on the right and one on the left). For the period 1990-1992, the left takes the majority - left 21, right 15, others 1. In 1993-1995 it loses terrain, mostly to a - right 16, left 5, a 5, others 4 (of which 1 should be counted among the right). From 1996, the right seems to get the preference again - right 5, left 1, a 1. Thus, after the shift from right to left, there seems to have been a shift back to the right, as well as the emergence of the immediate invasion as a third alternative.

Charles I don't have Master Go to check; but GoGod has quite a high proportion of Korean games in the hits. Geography often matters, in opening fashions.

Miai in the fuseki statistical analysis last edited by MrTenuki on July 9, 2006 - 02:06
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library