BillSpightsElfPositionalJudgmentExercises/Game 20 ELf's Solution

Assuming area scoring with 7.5 pt. komi, after B27 Elf estimates a winrate of 90% (49.1k playouts) for one player. Which player is it?

It's White. White won the game by resignation.

The game is GoGoD 1969-01-29a, a Honinbo League game between Takagawa Kaku (W) vs. Fujisawa Hideyuki. Komi was 4½ pts.

[Diagram]
White is ahead  

Both players were considered to be masters of the opening. According to Elf neither player has blundered so far. How did White (Takagawa) get so far ahead? Fujisawa simply made more, and larger, detectable errors (greater than 4%) than he did. 9 of Black's first 14 moves were errors, only 2 of White's first 13 moves were.

I think that B13 is a teachable moment. :)

[Diagram]
A textbook play  

B13 is a textbook play, and AFAIK, has appeared in some textbooks. According to the theory of the time, it was an ideal extension from the Black enclosure in the bottom right corner, threatening the jump to a or the checking extension to b. In addition it prevents White from making a double wing formation, a doubly ideal formation where White extends on both sides from the enclosure in the bottom left corner. B13 was the obvious play for both Takagawa and Fujisawa.

Elf thinks that it loses 8½% to par. ;) It is not even on Elf's radar. What play does Elf recommend? The Go Seigen/AI shoulder blow at c? Perhaps in a different whole board position. That's actually Elf's fourth choice. In this position Elf's top choice is the pincer at d.

[Diagram]
Elf's main variation for B13  

After B13 White plays the shoulder blow against black+circle. The continuation is joseki. As a student of Takagawa, I do not see the improvement for Black, but this just shows how much the bots devalue the sides.

Another teachable moment occurred a few moves later.

[Diagram]
The solid connection loses 11½%  

The natural looking solid connection, B19, loses 11½% to the AI kosumi at a. Once you see it, the kosumi looks like the obvious shape play, protecting against the push and cut, and exerting influence towards the side. It also prevents a push and cut if Black later plays the keima at b. A human blind spot. This was Fujisawa's largest mistake, BTW.

Elf's mainline after the kosumi

[Diagram]
Elf's mainline after black+circle  

W20 is the usual extension when the keima at a is threatened. Black invades at B21 and takes territory with gote. Then White plays the shoulder blow at W28. Black attacks with the boshi at B29.

[Diagram]
Variation  

If White extends a bit further with W20, Black doees indeed reply with the keima, B21. Then if W22 secures the top side, B23 builds an impressive moyo.

A string of errors

[Diagram]
Five mistakes in a row  

This game sequence shows both of Takagawa's errors, and three of Fujisawa's. We have already examined B19. W20 is the double keima. The saying is, Don't let the double keima get away. Well, Elf thinks that it is a minor error, losing only 5½%. Elf prefers the third line extension at a, after which invades at b, with the same plan as above, after the kosumi.

B21 seems normal, for Black to develop the right side moyo, but Elf thinks that it loses 9½%, making it Fujisawa's second largest error. Elf prefers to invade the top side, which is closer to becoming White territory. It seems to me that the bots prefer to make more compact moyos than the large scale moyos of the 20th century.

Elf's mainline for B21

[Diagram]
Pincer cum invasion  

B21 pincers the white+circle stone while also spoiling the chance for White to make a territorial moyo on the top side. Pushing through a small opening is usually strong, but if Black plays B23 at 25, White will top the tree at 23. White manages to make some territory with W24, after which Black does push through with B25.

W22 is Takagawa's second error, according to Elf. It loses only 4½%, which is right at the the borderline, I think. Here is Elf's mainline variation for W22

[Diagram]
More territorial play  

After W22 makes territory, B23 plays the double keima on the border of two moyos. Then White reduces the right side with W24 and W26. B27 invades the top side before White can play at a. Again note how Elf cares more about almost territory than about the large scale White moyo on the left side.

B23 in the game looks good, as a move on the border of moyos. However, after White plays W22 on the fourth line, Elf regards the top side as more urgent.

Elf's mainline for B23

[Diagram]
Invasion, running fight  

B23 invades the top side on the fourth line, after which both sides start to build a base on the third line. B29 attacks the two White stones, leading to a running fight.

[Diagram]
Invasion, running fight, continued  

B33 - B37 lean on White's left side, and White pushes out into the center. Elf stops here, probably because of a low number of playouts. It seems to me that Black may push into the center at a.

Back to the game

[Diagram]
Small error  

Elf regards B25 as a small error, losing 6% to simply replying at 26. Elf's mainline variation is of some interest. :)

[Diagram]
Simple response  

W26 is a standard tesuji is this corner situation. White follows with a normal shoulder blow at W28. Black pushes up a couple of times, then slides underneath with B33. W34 starts to solidify the White group and creates some eye potential.

Finally, how does Elf see the near future of the game after B27 ?

[Diagram]
Elf's continuation  

In the game Takagawa played W28 at 35, but Elf prefers the keima by 7%. (A third small error by Takagawa.) After B29 threatens the push and cut, White switches to the top side to secure some territory. Black, OC, pushes and cuts and secures a good bit of territory on the right side, while White gets some floating stones.

These last two diagrams illustrate a difficulty for me with bot recommendations, especially if the errors are small. In Elf's mainline Black's error secures a good bit of fourth line territory in exchange for a move on the top side and some floating stones, while the correct play allows White to make a nice group within Black's sphere of influence. It's almost as if you can tell the mistake because it leads to a better looking result. ;)


BillSpightsElfPositionalJudgmentExercises/Game 20 ELf's Solution last edited by 107.210.159.110 on January 4, 2020 - 18:13
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library