Life In 19x19 http://www.lifein19x19.com/ |
|
Opening problems for AI: Problem 39 http://www.lifein19x19.com/viewtopic.php?f=15&t=17686 |
Page 1 of 1 |
Author: | Bill Spight [ Fri Jul 31, 2020 8:31 am ] |
Post subject: | Opening problems for AI: Problem 39 |
Black to play Which way?
Which way should Black approach the stone? At a or at b? Enjoy! |
Author: | Knotwilg [ Fri Jul 31, 2020 10:57 am ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
AI have taught us a number of things including the prevalence of sente and the heuristic of efficiency. Efficiency is probably the key argument here: it's the lower black wall that needs to be made efficient. The top right stone is flexible and is more yearning for a reinforcement than for an extension.
The invasion of the lower right at A will probably die, with those black stones blocking off the route at the lower side. An invasion at B can still be met with a flexible answer, cutting or bending around. If Black can jump to C, het gets a nice "barmkin".
The approach from the other side is neither fish nor flesh. It doesn't reinforce the corner. renders Black's wall inefficient. Later White can further scoop out Black's territory at A. An invasion at B will likely give Black a wall from which is an inefficient extension. |
Author: | Farodin [ Fri Jul 31, 2020 11:15 am ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
I'm torn between "Push the enemy towards your thickness" and "thickness needs an extension". Maybe I should stop listening to proverbs and think for myself While I don't like the whole-board situation for black (the H4 stone makes me visibly uncomfortable), I would choose option B, hoping to get a move at H5 in at some point or another, creating a large moyo that white will have to reduce at some point. This puts the black stone in the upper right in a bit of a pinch, but the AI has shown us that 4-4 stones are surprisingly resilient. But honestly, I'm choosing B just because A's prospects seem even more dismal. Black's upper-right corner would not be secure yet, and he can kiss any dreams of a large territory on the upper side goodbye, given the low white stone already being in place at F17. I'm really curious about the answer to this one, and whether ELF decides to do anything about that H4 stone sticking its head out... |
Author: | Uberdude [ Fri Jul 31, 2020 1:30 pm ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
Is this a trick question? Neither? |
Author: | Gomoto [ Fri Jul 31, 2020 2:42 pm ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
It is a half a point question. |
Author: | Gomoto [ Fri Jul 31, 2020 2:50 pm ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
Why does AI prefer the b direction by half a point? If a similar strong black position would be more closed at the bottom, aka around the R3 corner , it wouldn't. |
Author: | pwaldron [ Fri Jul 31, 2020 7:02 pm ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
Bill Spight wrote:
I'm going for 'b', largely because of the extra potential in the area around 'c'. When I started studying pro games more seriously I was struck by how many points showed up in those areas. |
Author: | Bill Spight [ Sat Aug 01, 2020 8:56 am ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
The game is GoGoD 2003-01-20o, between Otake Hideo, 9 dan (W), and Oya Koichi, 9 dan.
Elf reckons that loses 12½% to the approach at a, making it a definite mistake. 9 dans do not blindly follow proverbs, OC, but surely the one about driving the opponent's stones towards thickness was in the back of Oya's mind. And he was surely aware of the threat of to slide to b or otherwise take advantage of the weakness of Blacks' corner by playing at c or d, for instance. It is likely that he planned to reply to with before he decided on . Unfortunately, Elf regards as losing 9% by comparison with the kosumi at c, which threatens the shoulder hit at e. Otake's play does not escape criticism, either. , a two space extension that makes a base, and , a one space jump, are bread and butter plays, but they are minor errors, or at least questionable. Elf reckons that loses 5% to par and loses 6%. IMO the bots are going to lead to a profound reconsideration of what we think we know about go, something that by human progress alone would have taken centuries to accomplish. Let's take another look at , through Elf's eyes.
switches to the bottom left, threatening to cut off three Black stones. replies strongly by crawling into the White corner. bolsters the White group with sente.
returns to the right side and blocks on the 2d line, making some territory. By comparison with , White in this line of play on the right side makes territory and counters the influence of the Black wall. Next, Elf's mainline variation for is instructive.
plays kikashi against the corner before making a base with . (If Black plays at 40 instead of blocking the corner, White can push into the corner at 39.) undercuts the White base, but simply blocks on top. switches to the bottom left. Top choice: Approach from the bottom
extends from the bottom corner. The fact that it might be close to the Black wall is no big deal. In fact, Black is building a territorial moyo. switches to the bottom left, and plays into the center, as well. Takagawa points out that it is by no means necessary to make a base or extend from a side wedge like , even though the ability to do so is part of the rationale for the play. Takagawa argues that preventing Black's ideal extension to or a is good enough. OTOH, the bots don't think much of those extensions early in the game. Anyway, in the mainline variation White tenukis and later plays an enclosure cum 3 space pincer against .
The hane, , is a strong play dividing Black in two on the left side and attacking both ways. bolsters his stones in the bottom left with sente. Then extends Black's territorial moyo into the center, as anticipated by Knotwilg, pwaldron, and, I suppose, John Fairbairn. White now continues the attack on the left side. I would have chosen the approach of in this game until I ran across a Jowa game years ago where Black played an side extension from a wall that seemed to me to be too close and thus a bit inefficient, and then played a one space jump from it, making territory from thickness. I decided that maybe Jowa was right and I wasn't. I think that the bots like the way top players in the 19th century handled play on the side over the way that top players in the 20th century did. This is another example, it seems to me. |
Author: | RobertJasiek [ Sat Aug 01, 2020 10:14 am ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
Bill, what hardware do you use? |
Author: | Bill Spight [ Sat Aug 01, 2020 10:39 am ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
RobertJasiek wrote: Bill, what hardware do you use? All this comes from the Elf GoGoD commentaries. I use them because I can explore dozens of games per day, and they use many rollouts. |
Author: | Gomoto [ Sat Aug 01, 2020 2:02 pm ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
FatKata: Attachment: bill39b.JPG [ 131.68 KiB | Viewed 7060 times ] Attachment: bill39aa.JPG [ 131.39 KiB | Viewed 7060 times ] |
Author: | dhu163 [ Mon Dec 26, 2022 8:11 am ] |
Post subject: | Re: Opening problems for AI: Problem 39 |
Speculatively: Current position: 1. The lower W group isn't 100% alive after F2, but for now H5/H6 is very big as it threatens a cut and B's influence. 2. I think I would choose A because the lower right corner is too open and W can reduce naturally from the centre on both sides. B needs to profit from the wall by getting more territory in the upper right. Note that B O1 has impact on W's life below even if not sente. The main problem is that W has H6 in sente and perhaps H7 or something also as B is thin on the left. Even after A, Black can continue attacking with the influence, with R5 after R7. Perhaps Q17 follows but W is in some danger. Opp to play 1. R13 is first instinct with miai of O17 and R7. Past: 1. B seems to have a good result in the lower right as W spent 2 moves at H2, M2 to live, and the connection to the left is not bad even if somewhat thin because it is miai with the H5 area. B getting a base in W's area is always a bit uncomfortable. So to get it with good influence and the lower right too, without W getting much territory surely means that B is doing ok here? Regardless D7 looks misplaced for territory though. H5 seems clearly bigger but perhaps B feared G5. So perhaps it was worth it in order to get W to defend the lower side. However, I still find myself unconvinced by B's thin shape on the left. On the other hand it seems not bad compared to W's slow moves. So I think B is ahead. Check: 1. I got this wrong. I was too scared of W managing to live or otherwise use the aji. My logic was simply incorrect, not keeping up the pressure. |
Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |