Another ancient game.
Black to play.
As always, assume komi.
Enjoy!
| Life In 19x19 http://www.lifein19x19.com/ |
|
| Opening problems for AI: Problem 42 http://www.lifein19x19.com/viewtopic.php?f=15&t=17710 |
Page 1 of 1 |
| Author: | Bill Spight [ Sat Aug 08, 2020 7:55 am ] |
| Post subject: | Opening problems for AI: Problem 42 |
Another ancient game. Black to play. As always, assume komi.
Enjoy!
|
|
| Author: | pwaldron [ Sat Aug 08, 2020 6:42 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
I find positions like these quite difficult to deal with. Black just has to make a good move, without an obvious weak group to target or reinforce. 'a' through 'd' all look plausible. 'a' looks like a good move, but I am concerned it is slow. 'b' just looks like pulling a single stone out, but I don't see the point. White has a base and 'd' to help. I'm worried about just making a target. My preference is 'c'. It moves first into the largest open space and aims at the weaknesses in the large enclosure. 'd' is interesting, building strength for future activities, but I'm not strong enough to figure if that's better than 'c'. |
|
| Author: | Knotwilg [ Sun Aug 09, 2020 6:11 am ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
The sequences I see. No choice yet.
|
|
| Author: | Joaz Banbeck [ Sun Aug 09, 2020 12:45 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
I'm stealing PWaldron's diagram because I think that it clearly defines the major places of interest.
I differ, however in my evaluation of the four plays. All look good for black, but both 'b'and 'c' can go badly for black if white gets too much strength around 'd'. Therefore, I'd play 'd' first, thereby having the strength to make 'c' and 'b' miai. And black still has 'a' for later if he comes out of a 'b; or 'c' fight with sente. |
|
| Author: | Shaddy [ Sun Aug 09, 2020 4:56 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
To those who would play the top right: consider the tewari
|
|
| Author: | Joaz Banbeck [ Sun Aug 09, 2020 7:20 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
Shaddy wrote: ...consider the tewari... If given the choice between being too weak or too strong, I vote for too strong. I'll still play Q14. |
|
| Author: | Kirby [ Sun Aug 09, 2020 7:31 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
Joaz Banbeck wrote: Shaddy wrote: ...consider the tewari... If given the choice between being too weak or too strong, I vote for too strong. I'll still play Q14. I also like Q14. If we're talking tewari, I'd answer the peep, probably:
|
|
| Author: | Uberdude [ Sun Aug 09, 2020 9:55 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
Kirby wrote: I also like Q14. If we're talking tewari, I'd answer the peep, probably:
A good move if you are playing 5-in-a-row. |
|
| Author: | Uberdude [ Sun Aug 09, 2020 10:10 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
@Shaddy Shaddy wrote: To those who would play the top right: consider the tewari
Sure, that move on its own is slow, but is it worse than this exchange for White which is what you need to compare it to for a valid tewari? One concrete downside of this exchange other than just losing points and making black stronger is white losses the standard continuation of attach and bulge with a-c as black can super easily play d now that White can't counter atari at 29.
Or an alternative tewari, black played the knight answer in response to the approach, very common move these days. Instead of locally normal move at a white attached at 1, black haned outside, for sure a bad exchange for white. Then white tenukid, and black fixed with 4. 4 is probably slow, but is it worse than 1 for 2?
|
|
| Author: | Knotwilg [ Mon Aug 10, 2020 5:03 am ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
I have made up my mind. I must say that John's musings about heuristics that take #stones/area into account, help me make decisions in unclear situations (or rather unclear vision on situations). First of all, I'm ruling out the 1 space jump at the top. It's separating strong groups. There's no point to that. The top right is confusing most of us. It's overconcentrated already but there's bad aji. How to reinforce effeciently? Probably by playing away from it but still affecting it. The lower left gives an opportunity to make territory while attacking but there's a danger of losing sente. The lower right is the area of growth for White. It's where most of my sequences aim for next. John's heuristic also points to the lower right as an area for Black to move into. Of all my sequences, I like this one best:
It fulfills all requirements raised above and also keeps sente. Is it better though than simply playing a 1 space check? Maybe 4 forcing stones is a little too heavy.
 seems to lose the initiative.
Then again ... White's  is so severe that it's always sente and it's always leaving Black with problems he wouldn't have if he played there himself.Still, I don't like a Black move there either. So I guess Black is simply behind, if he needs to play a bad move. |
|
| Author: | Uberdude [ Mon Aug 10, 2020 7:14 am ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
Knotwilg You like that outside attachment of the big shimari. Are you aware that's generally considered a bad move, both pre- and post-AI (because White simple nobi is efficient answer, unlike with small shimari), but think this position is an exception, or you just didn't know that or got confused with the low shimari and like attachments more now? The AI attachments against the big shimari are, most common q4, and sometimes p4. Though I don't like either of them much here as I don't want to make the white corner stronger given black m3 might be a good move to defend that lower side group in the future and then I'd want to have kept the option for shenanigans inside with q3. |
|
| Author: | Knotwilg [ Mon Aug 10, 2020 8:29 am ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
Uberdude: Quote: You like that outside attachment of the big shimari. Are you aware that's generally considered a bad move, both pre- and post-AI (because White simple nobi is efficient answer, unlike with small shimari), but think this position is an exception, or you just didn't know that or got confused with the low shimari and like attachments more now? I was looking for a way to keep sente. A simple check doesn't do that. The attachment may be inspired by seeing it against the small shimari. You're undoubtably right it is not a good choice against a big shimari. By showing the sequences I'm seeing, I'm also showing are sequences I'm not seeing. |
|
| Author: | Bill Spight [ Mon Aug 10, 2020 8:40 am ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
I must say I am pleased to have sparked such an interesting discussion. I was lucky to find this problem. The game is GoGoD 1620-00-00at, Nakamura Doseki (W) vs. Yasui Santetsu. How we got here.
Elf considers  to be a mistake and  to be a blunder. Better to occupy the last open corner at a. With mistakes on both sides we later reach the problem position where Black would be significantly behind if giving komi. Even with no komi, Black is maybe a few points behind, I suppose.
The one space jump is never bad, they say. Well, Elf does not like  and reckons that it loses 13½% to par. It doesn't much like  , either, which loses 11% to par.Here is Elf's mainline variation for White's reply.
takes kikashi. a and 30 are also playable.  and  push down on Black on the bottom side.  reaches out for lebensraum and to threaten the White corner.
 and  continue to press down, building a wall.  prevents a White descent here with sente. Then  plays the double keima in the bottom left, extending Black's moyo into the center.Top choice: Honte, shared blind spot
is a thick honte in the top right, preventing the White kikashi. It has been, and will continue to be, a blind spot for both players. Elf wanted White to take the kikashi immediately after , and has wanted White to do so for all but two moves since then. It has wanted Black to play the honte at move 13, and at all but three moves since then. Black finally played there with .
As people have pointed out, the top side per se is not urgent. switches to the bottom side, and  bolsters the top edge of Black's left side moyo.
continues play on the bottom side, much as in the above variation. After extends towards the bottom right corner,  pushes into the bottom left corner, threatening the nearby Black stones, which Black protects with a 4th line extension. Then  reduces the Black left side.I mainly chose this play because Elf preferred a thick play in the opening, when bots usually prefer light play. Given the human blind spot, I suppose that other bots will prefer it, as well. I think I would have missed it, too. The White kikashi seems equitable, and preventing it at the risk of overconcentration doesn't look as big to me at this point as the block at E-06 or P-05 against the bottom right corner. But Elf has considered the Q-14 point to be urgent for quite a while. |
|
| Author: | Shaddy [ Mon Aug 10, 2020 9:38 am ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
@Uberdude Hum, ok. This one, then:
1 is very nearly a pass. 2 for 3 is bad but loses only the attachment to the isolate stone, which isn't much. edit: just saw the solution. i am doing very badly at these.. i suppose that means my evaluation is close with bill's. i did notice that the temperature on this board is low, but still undervalued the difference between both players' moves in the top right. i would have played something against the bottom right corner. |
|
| Author: | dhu163 [ Mon Dec 26, 2022 8:58 am ] |
| Post subject: | Re: Opening problems for AI: Problem 42 |
Speculatively: Current position: 1. There are many small fights on the board. The upper right looks a bit bizarre. K17 is well behind enemy lines, but there is no eyespace or counterattack potential. So although K15 seems like a move, Q14 seems much more attractive locally. 2. Otherwise, the lower left W group isn't alive and can be attacked naturally by D3. Perhaps E6 is also possible but it seems W can lean at G4 naturally and then F6. 3. B can expand the left but there are some weaknesses and the upper W side is hotter as K17 is weak. Later M3, R6 will be big. Opp to play 1. Q14, D12 look natural. Past: 1. B's shapes seem less efficient, especially the upper right, and K17. The lower left is worse for W but I think it is mitigated by W's centre strength from the moyo. Check: 1. basically correct. Looking at other people's misunderstanding of the lower right, it seems people seem too happy to lean on the weakness from the right side which gives W more space on the centre, corner and lower side. That's locally bad shape, helping W by playing forcing moves close to a weakness but no close enough to even gain profit from it (which requires the 4-4 attachment). It's not worth it, when there was plenty of aji to live in the corner, which is natural when B is alive on both sides. |
|
| Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|