I couldn't resist this one, although I suspect that a lot of people have seen Black's mistake. Elf reckons that it loses 16½% to par.
Perhaps the bots will agree about Black's best play, I dunno.
Black to play.
Enjoy!
| Life In 19x19 http://www.lifein19x19.com/ |
|
| Opening problems for AI: Problem 27 http://www.lifein19x19.com/viewtopic.php?f=15&t=17631 |
Page 1 of 1 |
| Author: | Bill Spight [ Fri Jul 10, 2020 11:22 pm ] |
| Post subject: | Opening problems for AI: Problem 27 |
I couldn't resist this one, although I suspect that a lot of people have seen Black's mistake. Elf reckons that it loses 16½% to par. Perhaps the bots will agree about Black's best play, I dunno. Black to play.
Enjoy!
|
|
| Author: | Harleqin [ Sun Jul 12, 2020 4:18 am ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
These points stand out for me:
'a' builds thickness in sente against the white group. 'b' presses White low on the right. 'c' might build shape afterwards, somewhat loosely connecting 'a' and 'b'. 'd', 'e', or 'f' could take the second move in that corner, but I can't say what's best. Maybe this calls for probing something else first, but I'd be reluctant to take gote with something like 'c' instead of playing in this big area. Playing in the lower left with 'g', 'h', or 'i' is also big, but again, I can't say what's best. Finally, 'j' is almost sente, which either means that Black could build something on the wide left side, or that the left side is not as wide as it looks. I think that I would play 'a' in any case. White then probably wouldn't answer 'b' and take the lower right, after which it feels slow to me to follow up on 'b' (White would go on with the press or a counter pincer in the lower left side), so I guess 'b' is too early at that point. I'd be very much inclined to play the lower right now, but it also feels big to play on the left side first, so White will probably then ignore almost anything there and play in the left. On the other hand, if Black plays in the lower left, White's best move in the right seems to be the “classic” tight shimari, which we nowadays regard as a bit slow (with the possibility of the shoulder hit and all that). So finally, I'd feel like playing 'g' now. It seems like taking a stance (in Karate, I'd call it “kamae”; is that the right word here, too, John?). My prediction (wishful reading?) then goes something like this:
Or:
|
|
| Author: | Bill Spight [ Sun Jul 12, 2020 11:41 am ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
The game is GoGoD 1674-08-30a, Honinbo Dosaku (W) vs. Yasui Chitetsu, in which Dosaku induced Chitetsu to play  and become overconcentrated. Only Dosaku understood the concept at that time.  lost 16½% to par by Elf's reckoning.
In the game Dosaku continued with  , but Elf prefers to play the press in the bottom left corner first.
After   bolsters his group with sente and then  invades the bottom right corner.Elf plays in the bottom left corner.
Black plays the shoulder hit in the bottom left corner. White answers with  , but a, b, and c are also playable. What follows is an AI joseki.
plays to confine White rather than taking the  stone.  pushes through once and takes bad shape in order to create two cutting points in Black's position. encloses the bottom right corner. a is also playable, OC. Then  plays the footsweep in the top right corner instead of accepting overconcentration.Edit: I just noticed. Dosaku played the swastika 5-3 (5-3 in all four corners, with rotational symmetry). |
|
| Author: | Knotwilg [ Sun Jul 12, 2020 1:18 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
Can't really imagine anything else than:
Black makes profit while attacking. |
|
| Author: | Uberdude [ Sun Jul 12, 2020 1:34 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
@Knotwilg And if white doesn't agree the 2 stones are biggest, will you insist they are?
So maybe black should apply the 1-2-3 principle and play q18 directly and not hope white is obediant |
|
| Author: | sorin [ Sun Jul 12, 2020 3:03 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
KataGo with komi set to zero has some unexpected suggestion. The favorite is black "a" in the lower-left, which seems unusual by human standards. In fact I don't remember ever seeing this move in human pro play! This way black is winning by 93.5%, 9 points ahead. KataGo's 2nd choice is black "b" in the upper-right (92.9%, also 9 points ahead). I also set the komi to 6.5, and got the same top 2 moves (but lower winning percentages, 68% and 65%, and lower point differences, of course).
|
|
| Author: | Knotwilg [ Sun Jul 12, 2020 3:41 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
Agreed. Good application of 1-2-3 and of “play away from strong stones” |
|
| Author: | Bill Spight [ Sun Jul 12, 2020 5:25 pm ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
sorin wrote: KataGo with komi set to zero has some unexpected suggestion. {snip} In fact I don't remember ever seeing this move in human pro play! GoGoD 1931-10-28a, Maeda Nobuaki, 4 dan (W) vs. Go Seigen, 4 dan. Once again Go Seigen anticipates the bots!
|
|
| Author: | dhu163 [ Fri Jun 04, 2021 10:30 am ] |
| Post subject: | Re: Opening problems for AI: Problem 27 |
Speculatively: Current position: 1. Standard shape theory is to connect L16 unless you really want sente. You can also play L18 here as B is fairly solid but W can atari M15 and not capture N18 as it is light. 2. The question is how valuable is W M15? Opp to play 1. In lower left, W cut is a move but not very efficient when already on 2nd line. My instinct is C11/D12 ish Past: 1. B seems overconcentrated at top but is quite thick in upper right while W has some weak stones. W still owes a move in lower right. However, B played a vulnerable pincer on lower side. Seems fairly balanced to me. Check: 1. missed the point |
|
| Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|