These topics are based upon Elf's commentaries on GoGoD games. As always, assume area scoring with 7½ pts. komi.
After
Elf estimates one player's winrate as 90½% with 56.6k playouts. Which player is it?Enjoy!
| Life In 19x19 http://prod.lifein19x19.com/ |
|
| How is your positional judgement? Game 17 http://prod.lifein19x19.com/viewtopic.php?f=15&t=17010 |
Page 1 of 1 |
| Author: | Bill Spight [ Sun Oct 27, 2019 7:36 am ] |
| Post subject: | How is your positional judgement? Game 17 |
These topics are based upon Elf's commentaries on GoGoD games. As always, assume area scoring with 7½ pts. komi. After Elf estimates one player's winrate as 90½% with 56.6k playouts. Which player is it?
Enjoy!
|
|
| Author: | bayu [ Sun Oct 27, 2019 8:50 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
3 corners, neither of which is sealed in. And black gets a free ladder breaker. I take black anytime. But because Bill posted this as something interesting, I bet that white is ahead. No clue why though.
My instinct is to continue like this to provoke black to start a weak group. It's probably a sure fire method to lose the 90% advantage:) |
|
| Author: | TelegraphGo [ Sun Oct 27, 2019 9:05 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
Could this be it? The first one I get right on record? Probably not, since I don't even know the joseki they played  I have a faint recognition of the bottom left and top right results, but not enough to be able to reproduce how they came onto the board. I'll start by looking at the top right. It's pretty clear that W had one stone captured here - the R17 stone only ever gets played when B is afraid of capturing immediately and allowing W a ko by playing R17 themselves like in the position below:
So that means it's a 7 black stones v. 6 white stones commitment. Since B has committed one more stone, we can compare to a similar result where black has committed one more stone than white.
Compared to this result, in game black has the corner much more solidly, but with a lower position. White in game has a significantly stronger position, but with ladder aji. If the ladder is easy for white to handle, then the game position is much better for white than the simple one above - white is just way harder to attack severely. So let's look at the bottom left corner to see how the ladder aji is. It's white's turn right now, and W is about to sacrifice the two stone for influence. The first move that comes to my mind is:
But to all those who've seen this sort of sequence before it's obvious that W is supposed to take advantage of B's one stone before doing this. The question is whether W should employ a driving tesuji by Atari underneath to prevent the open skirt, or simply Atari from the top to prevent the cutting point. Since W has no help nor development on the left side yet, I would prefer leaving an open skirt to a cutting point, so my suggested variation is:
This bottom left position looks absurdly good for W. B is surrounded with only a handful of points (and bad endgame aji on top of that). On top of that, this is the ideal kind position to have on the opposite side of the board of a ladder. If B tries to play a ladder breaker, it will be inefficient, because it will be too close to thickness. All of W's stones work beautifully together - W's bottom right stones aren't really attackable due to W's big lead in influence, so the game looks hideously difficult for B. I've been confident with all the other ones, but this time really - if AI favors black here then I'm gonna start 2-2 invading san-san openings for territory. |
|
| Author: | jlt [ Sun Oct 27, 2019 9:13 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
White has no territory, and Black has three corners and no weak groups. On the other hand, Black has no influence: no stones on the fourth line or higher, and White can get a wall on the lower left and lock Black in. Maybe I prefer White because I don't like being excessively territorial, but it's hard to choose.
|
|
| Author: | Bill Spight [ Sun Oct 27, 2019 10:06 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
Just a note about how I choose the games to show. First, OC, is that I look for games with Elf winrates around 90% or higher around move 30. These may be a bit too easy for those who reply, but they are not always so easy. And that's a good place in the game to evaluate the position. Second, I look for games by Takagawa, Go Seigen, Lee Changho, and Fujisawa Hideyuki because they are human masters of the fuseki. Also, Takemiya, because who understands Cosmic Go? (Nobody, according to Elf.  ) Also, Cho Chikun, because if we still had a Meijin, wouldn't he be it? I have rejected a few games because some of Elf's variations were too far above my head. OC, I do not pretend to understand all of this, but I do want to say something sensible. 
|
|
| Author: | Kirby [ Sun Oct 27, 2019 10:37 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
I like white. White seems to have a lot more potential than black. |
|
| Author: | Tryss [ Sun Oct 27, 2019 10:58 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
What was the komi of the original game? Katago with 0 komi think the game is even (52.5% wr for black with 40k playouts)  |
|
| Author: | Bill Spight [ Sun Oct 27, 2019 11:51 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
Tryss wrote: What was the komi of the original game? Katago with 0 komi think the game is even (52.5% wr for black with 40k playouts) 5½ komi. I am avoiding 0 komi games, and even 4½ komi games, because the question of whether the players have made mistakes or not in a 0 komi game is too difficult. In no komi game Elf usually says that White gives up many percentage points in the opening. |
|
| Author: | xela [ Sun Oct 27, 2019 2:16 pm ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
Echoing bayu (post #2 above): I feel as though I'm supposed to choose white. But at my skill level, I wouldn't trust myself to use all that influence well, so I'd probably do better playing black (against someone else of my skill level). And that leads us to an interesting question: evaluation assuming both players are strong, versus evaluation assuming both players are amateurs? |
|
| Author: | jlt [ Mon Oct 28, 2019 12:32 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
@xela: I looked at positional judgements by early versions of LeelaZero, and by Leela 0.11. LZ 30, 20k playouts: 59% LZ 40, 20k playouts: 65% LZ 91, 10k playouts: 65% LZ 116, 4k playouts: 75% LZ 157, 2k playouts: 77% Leela 0.11, 100k playouts: 68%, W+10 So even kyu level versions of LZ favor White. In my case, I wouldn't be confident to play as White, as I am not sure I would be able to use my influence. On the other hand, as Black I would be afraid that White uses his influence to attack me. Tough choice... |
|
| Author: | Knotwilg [ Mon Oct 28, 2019 4:14 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
no-LZ analysis
Although White can enclose Black in the lower left corner like this, it's a 15 point corner and half of her influence (after tiger mouthing at A) is facing a low Black stone. Black doesn't seem to have any inefficiencies, except for being low in the upper right as well. White's ladder represents some aji too. Overall, this seems to be a case where Black has safe territory in three places, while White's influence is far apart. Black has plenty of room to maneuver in that grand sphere of influence and nearly every group of White can potentially be turned into a liability. So I choose Black. Edit: dwyrin's heuristic favors White. And there IS an inefficiency in Black's position: the marked stone in the upper right should be at B. So I'm less confident now of my choice. LZ |
|
| Author: | Bill Spight [ Sun Nov 10, 2019 8:32 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
My discussion of Elf's commentary. Honinbo final, game 2, 1995-05-24a. Kato Masao (W) vs. Cho Chikun After Elf estimates White's winrate as 90½% with 56.6k playouts.White won by resignation.
Elf's mainline continuation
Kato played the two step hane. The rest is Elf's variation.  plays a hanging connection to secure the Black group. Then  takes kikashi against the bottom right corner, then  makes a trumpet connection. Black takes kikashi against that and then invades the top left corner.
 eliminates the aji of the  stone.  prevents the extension on the top side with sente, then  secures the corner and creates an open skirt for White.  makes a three space high extension from White's wall. No point in a low extension because of the open skirt.Black has taken all four corners, but Elf judges White's outside strength to be commanding. Earlier in the game.
Through  was a popular fuseki, with the probe, , regarded as an improvement. However, Elf thinks that is a minor error, losing 7% to par. Elf prefers Wa - Bb, Wc. The bots do like the solid connection at a. 
Elf thinks Black made three mistakes in a row.  lost 7% to par,  lost 11½%, and  lost 15½%. All that to capture one stone! After  Elf gives Black a 20% chance to win the game (60.1k playouts).
extends on the second line, avoiding White's kikashi there.  and  play kikashi against the corner, then  returns to the bottom right corner with the solid connection.
makes the solid connection, again preventing the White atari there. -  penetrate the Black corner.  makes a kind of two space extension, and the White wall is still insecure.  pushes along the fourth line,  blocks, and then  hanes.
 at a,  at b extends on the fourth line,  pushes, and  block again.  protects against the cut, and  stands.  cuts, and the fight is on.
captures the  stone, also preventing the atari at a. After the probe and crosscut, preventing the White atari at a becomes a prime objective for Black. The Black shape with five stones surrounding the eye formed by capturing the single White stone appears in textbooks as a good way for White to sacrifice that stone. Elf really doesn't seem to like that shape. penetrates the corner, cutting off the stone. gets the cut in the bottom right. That is a rare treat, as the bots love the solid connection so much. and push along the fourth line, building a wall, then extends on the top side. takes the double keima, and then protects White's incipient territorial moyo.Back to the game.
The solid connection, , is big. So is capturing the stone in a ladder. , although joseki, is a mistake, according to Elf, losing 10% to par. Black should save the stone.
Elf recommends , not only saving the stone but threatening the stones as well. - play the tenuki joseki in the bottom right.Back to the game. After Elf gives Black a 10½% chance of winning (24.3k playouts). White makes a few minor errors in the bottom left corner, but maintains the 90% lead after Black blunders.
is a ladder breaker. parries it, but Elf thinks that 27 is 7½% better. extends on the fifth line, offering the corner for outside strength. Elf regards the solid connection at 29 to be 5% better.
Elf regards  as a minor error, losing 7% to the a. But is a blunder, giving up 18% to par. Even though Black can threaten to break out with sente, what he gains in the corner is not worth the outside strength that White gets in exchange, according to Elf. Elf says that should take gote and make the ladder with b. The two space extension in the actual game loses 13% to par, a blunder when White's winrate is aroundd 90%.Elf's judgement of the play in the top right corner is particularly instructive, as White's sacrifice comes right out of the textbooks. |
|
| Author: | dhu163 [ Mon Dec 26, 2022 10:53 am ] |
| Post subject: | Re: How is your positional judgement? Game 17 |
seems easy All is joseki except for upper right and lower left both of which are better for W. Next move is double hane on lower side. On lower left, W has F5 misplaced and overconcentrated, but can make up for it with the tight local shape of the double hane. On the right, W's outside is put to good use saving a move not needing to extend with R6 and instead directly developing. On the lower left, W has several different aji, probably B6 is best if you have to play something, but W can tenuki and also has several ko threats. Upper right: B has one extra stone. Just around 10 pts with some 3rd line potential that is limited by a cutting point. The shape is way too confined to a small area. W has some points with the ladder as B has no good ladder breaker. Locally perhaps 2pts with 6 or 7 lines of potential with another move. That is simply more than B. Lower left: This was W's first corner and W needs to add another sente move to continue. B has 13 pts. W has a wall of height 13 if W adds another move, with one weakness on the left. This is normally estimated at around 26 pts, which seems reasonable. But W is also supporting the life of the Q6 stones. Basically B expects a centre move as in return for a ladder breaker. If we count the right side as cancelling B's lower right, then W at the moment has 13 more in the lower left and the upper left corner which should be counted as 13 normally also. komi should compensate for B's initiative. Upon discounting the ladder breaker, it seems as though W is leading by around 15 pts. I find that a bit hard to believe. Probably I haven't sufficiently counted the W's weaknesses (perhaps there are 1.5 cutting points, subtracting 3pts?) and the fact that all B's groups are alive. Also W probably has less in the lower left than this estimate because B has potential on both sides. This still doesn't seem consistent to me. If the lead really were that much the winrate should be 99%. Perhaps B hane on the lower side isn't optimal. It is sometimes better to pull back for better aji on the left as the lower side isn't valuable and even after adding a move, problems remain even on the lower side. |
|
| Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|