This 'n' that

Bill Spight · Post by **Bill Spight** » Sun Dec 15, 2019 9:44 pm

lightvector wrote:
John Fairbairn wrote:Bill

A bit of clarification, please. You say the nobi W12 loses 11.5% under Dobby the Elf. Princess Leela, however, says the win rate between that and atari only falls 3.5% (16k playouts).

I think you've said before that Dobby's numbers are consistently on the high side.

So do we assume all Dobby's numbers are three to four times higher than Leela's? Or do we have to say "apples and pears" - Dobby is so much stronger than Leela that a comparison is a bit futile?
(I'm not quite sure of the connotation you're trying to communicate by calling them weird names so maybe I'm misunderstanding something but... )

JF and I both like word play, so no problem as far as I am concerned. IIUC, Dobby is an elf in the Harry Potter books.

Leela Zero (with latest 40-block networks) is quite a bit stronger than ELF. Definitely stronger at equal visits, probably stronger at equal time, unless you're using very small numbers of playouts or something is unusual about your hardware. And this is even while being less opinionated about opening moves, in the sense of giving less extreme winrates for them.

Thanks for the info. The Elf commentaries were mostly created in October of 2018, ancient history in bot time.

But they use a large number of playouts, so there is that.

JF's reference to a fall in winrates suggests to me that he is using game analysis mode, or some such, which focuses on winrate changes between consecutive plays, which may be the drop he is referring to. And when comparing options for the same play, the less preferred option will often have many, many fewer playouts than the top choice, which means that its winrate estimate is much less reliable that that of the top choice. That's why I recommend yoyoma's suggestion to enter the less preferred play by hand.

ez4u · Post by **ez4u** » Sun Dec 22, 2019 6:59 am

Let's look at this in a little more detail: some historical and some bot. My sources are GoGoD Winter 2019 edition and LZ #254.

This fuseki with the immediate approach answered by the 2-space low pincer has a long history but it is not correct that it was popular at the time of this game. If we search GoGoD we find a total of 85 games stretching from 1647 to 1933. It was, however, mainly a tool of the Yasui family and the main source of 17th century games are biased toward the more famous Honinbos. In the 85 games, 50 featured a player as Black with "Yasui" in their name and 53 featured a "Honinbo" as White. The "Page from Go History" article in Go World #11 is about Dosaku and the superiority of his fuseki theory over that of his rivals of the time. The featured game starts with this 1-2-3. All of these games occurred in the 16th and early 17th centuries.

The last game with a Yasui player as Black appeared in 1722. GoGoD has only 8 games later than 1722. Interestingly, four of those games featured the Yamamoto Genkichi that we see here, two as Black in 1801 (including this one) and two as White in 1809 and 1813.

During the battles of the 17th century, the 2-space low pincer was replaced by the 3-space low pincer where Black responds immediately to White 2, probably due to Dosaku's influence. This is interesting because LZ calculates a nearly identical winrate for Black (at lower visits the 2-space is preferred but the calculated difference falls to only 0.1% better beyond ~100K visits).

Also interesting is what happens when Black responds to White 2 by playing elsewhere. The human preferred play was the 5-3 point at P3 in the lower right, facing the lower left corner. LZ calculates Blue as the upper left star point (D16) instead. However, there is a problem with this analysis. P3 in the lower right does not appear in LZ's policy net. As a result, even with 200K visits LZ had only 9 visits at P3. So should we conclude that D16 is "better"?

Of course the analysis here is more about the joseki rather than the fuseki and indeed the joseki appears far more often than the specific fuseki in this game. If we search GoGoD on an otherwise empty 10x10 corner, we have 167 games with the 1-2-3-"tenuki"-5 of this analysis. White more often replied to the attachment than not, but playing elsewhere was a significant choice at least temporarily. In terms of local responses the two main variations are what is discussed in this analysis: the wedge at D4 played in this game and the outside hane at F4 that Elf "preferred". Both plays first appear in the 1660's in GoGoD.

The outside hane goes back to 1660. The intent when it was played by humans was to trade the outside for the corner. It anticipates the cut by Black and then a series of ataris that ends with White capturing Black's original 3-4 stone.

Click Here To Show Diagram Code: [go]$$W Joseki part 1 $$ $$| . . . , . . . . . , $$| . . . . . . . . . . $$| . . . . . . . . . . $$| . . . . . . . . . . $$| . . . . 7 . . . . . $$| . . . 5 4 8 . . . . $$| . . X 3 X O . . . , $$| . . 9 6 O X 1 X . . $$| . . . . . 2 . . . . $$| . . . . . . . . . . $$+ - - - - - - - - - -[/go]

Play continues to

below and then Black plays elsewhere

Click Here To Show Diagram Code: [go]$$B Joseki part 2 $$ $$| . . . , . . . . . , $$| . . . . . . . . . . $$| . . . . . . . . . . $$| . 6 . 5 . . . . . . $$| . . . 3 O . . . . . $$| . . 2 O X X . . . . $$| . 4 X O X O . . . , $$| . . O X O X O X . . $$| . . . . 1 X . . . . $$| . . . . . . . . . . $$+ - - - - - - - - - -[/go]

Elf chooses a very different variation. The stand that it uses after the first atari does not appear in GoGoD. When we try in LZ, it calculates Blue as the same stand at F5 as Elf. OK let's write that up - the humans are wrong again. Alternatively let's just... wait a while. Two choices dominate LZ's policy net, the stand at F5 and the atari at D4. Its initial calculations do not cover anything else. However, at around 50K visits it has a first peek at the strange-looking alternative of hane at H4. Then at around 130K LZ calculates H4 in more detail. If you have a really fast machine (or in my case if I go and have dinner and then watch a little TV), LZ calculates Blue as shown below given enough visits.

With a million visits LZ calculates a main line as shown below. There are two interesting things here IMHO. First, this variation uses the same corner maneuver by White as the human joseki to take the lower left corner. The difference is that interpolating H4 and the other moves near it leaves more potential on the bottom side. Second, watching the calculated variations rolling out in Lizzie, the bottom left corner (up to White 15 shown in the picture) is stable relatively quickly (by ~160K visits). Thereafter, out to 800K is almost entirely about the rest of the board.

Bill Spight · Post by **Bill Spight** » Tue Feb 18, 2020 8:32 am

Just a brief note.

It's been a busy new year for me so far.

I have finally received my 2080 Ti graphics card, along with the rest of the parts for building a computer. I can get it built locally at a reasonable price, and they will stress test it, as well. I should have it up and running by March.

As those who have been reading along will know, I have had a bee in my bonnet about the margin of error of the bots' winrate estimates. Surely, I thought, the developers must know what it is, but nobody is telling. Well, I just stumbled across a guy who talks about measures of uncertainty in deep learning, of which margin of error is one.

He is Yarin Gal, whose doctoral dissertation was entitled, Uncertainty in Deep Learning (2016). He is now an associate professor at Oxford and leads the Oxford Applied and Theoretical Machine Learning (OATML) group. I may not learn enough to apply his methods to my own research, but I am certainly going to study his work.

If you are interested, here is a link to one of his blog posts, on the science of deep learning. http://www.cs.ox.ac.uk/people/yarin.gal ... _5058.html

lightvector · Post by **lightvector** » Tue Feb 18, 2020 3:49 pm

Bill Spight wrote:As those who have been reading along will know, I have had a bee in my bonnet about the margin of error of the bots' winrate estimates. Surely, I thought, the developers must know what it is, but nobody is telling.

I hope you've realized by now that the developers don't know either.

And part of the reason is the issue that you've earlier claimed is "not your problem" - that there's a mathematical issue defining what "margin of error" means in a quantitative way that also corresponds to being useful for Go players. Regardless of whose problem it is, it's not surprising that if you find it tricky yourself to define what you want, then the bot developers - who are on average much, much worse at Go than you, and therefore are much less familiar with what would constitute a success - would fare no better.

And the curse of the machine learning age is that the machine learns the same anyways - the developers don't need to dig into such issues, so they haven't.

Bill Spight · Post by **Bill Spight** » Tue Feb 18, 2020 4:23 pm

lightvector wrote:
Bill Spight wrote:As those who have been reading along will know, I have had a bee in my bonnet about the margin of error of the bots' winrate estimates. Surely, I thought, the developers must know what it is, but nobody is telling.
I hope you've realized by now that the developers don't know either.

Yes, thanks to the kind remarks of yourself, yoyoma, dfan, jlt, jann, and others, I realized that the learning programs do not produce such error estimates.

And part of the reason is the issue that you've earlier claimed is "not your problem" - that there's a mathematical issue defining what "margin of error" means in a quantitative way that also corresponds to being useful for Go players. Regardless of whose problem it is, it's not surprising that if you find it tricky yourself to define what you want, then the bot developers - who are on average much, much worse at Go than you, and therefore are much less familiar with what would constitute a success - would fare no better.

And the curse of the machine learning age is that the machine learns the same anyways - the developers don't need to dig into such issues, so they haven't.

Well, I was glad to find someone else who was bothered by that, and who seems to be pretty sharp. Associate prof at Oxford within 4 years of getting his Ph.D., which was on this topic. He has found a way to measure uncertainty in recurrent neural networks by approximating a Bayesian approach, and even to distinguish between noise and what he calls epistemic uncertainty. I'm a bit skeptical about the last, and this may be over my head, anyway, but it looks interesting.

xela · Post by **xela** » Tue Feb 18, 2020 7:25 pm

OK, lightvector got to it first :-) I too don't believe there's a developer's Conspiracy to Hide the Truth from Bill.

On the one hand, I could write a little essay about the spectrum from classical statistical models to purely empirical machine learning, and Hastie and Tibshirani's attempt to define a reasonable middle ground under the name "statistical learning". And I could remind you of the difference between frequentist and Bayesian statistics, and the difficulty of defining what it is we're trying to measure when we say "winrate". But you already know all that. (We've had this conversation before: the margin of error is already there! If LZ says 93%, then the margin of error is surely 7%. Or maybe 93% on a bad day.)

On the other hand, if what you're asking for is "how confident is the machine in this prediction", I think we already have nearly as much information as we could wish for. Fundamentally, it's trying to predict whether black or white will win. If it says 93%, it's saying this is a very confident prediction. If it says 55%, it's saying maybe one side has a slight edge, but with low confidence. When you ask Lizzie to show you the best move, you quickly see whether the machine can identify a clear-cut "correct move" or whether it's struggling to decide between alternatives. With Lizzie+KataGo, the KataEstimate button will colour in points black or white for who is likely to control which areas, and shades of grey for where it's not sure.

And if that's not enough, we can inspect the search tree in considerable detail. (I haven't forgotten that I intend to post some ladder and ko examples there. Just need to find a free evening or two to write them up. Darn real life, intruding on go again...)

So much amazing progress these last five years! Bill, what more does it take to satisfy you? :-)

xela · Post by **xela** » Tue Feb 18, 2020 7:32 pm

PS thanks for the link to Yarin Gal's blog. I've heard of dropout in neural networks before, but not looked closely at it. In image recognition, if you mask random parts of an image and then still try and guess whether it's a cat or dog, that's a pretty reasonable question. In, go if you mask random stones on a go board and then try to guess who's winning or what the best move is... Well, sometimes you're wrapped up in a big local fight and it really doesn't make much difference what the rest of the board looks like. Or sometimes you're making a joseki choice based on the adjacent corners, and the diagonally opposite corner wouldn't influence the choice so much. So masking a few random stones wouldn't change your assessment. Certainly humans don't consciously consider the position of literally every stone on the board when thinking about every move. But other times, taking even one key stone off the board will make a radical difference. Perhaps averaged over a million training examples, the noise cancels out and we're left with good predictions plus a better appreciation of uncertainty? Or maybe this approach is just too chaotic for go?

Bill Spight · Post by **Bill Spight** » Tue Feb 18, 2020 9:09 pm

Well, there is a post that I have had in the back of my head for some time, about just noticeable differences, chaos on the go board, and training. I have actually addressed the question here same years ago in regard to human training, without using the jargon.

I didn't want to say anything about training bots, because it could be I would be talking through my hat. But it's an idea I have wanted to play around with and get some data on, with my new axe.

I had not thought of masking an intersection as a kind of just noticeable difference, but it is.

BTW, the original dropout method dropped 50% (

) of intermediate layers and only 20% of the input layer. That seems extreme for go,

And the original method did not work for recurrent networks. Now Gal makes it work. Interesting stuff.

xela · Post by **xela** » Tue Feb 18, 2020 10:27 pm

A passing thought: is one colour go a form of drop-out training for humans?

Bill Spight · Post by **Bill Spight** » Wed Feb 19, 2020 3:03 am

xela wrote:A passing thought: is one colour go a form of drop-out training for humans?

Well, switching the color of a point could be considered a just noticeable difference.

Bill Spight · Post by **Bill Spight** » Tue Apr 07, 2020 10:21 am

Unplanned obsolescence

In the AI era a number of joseki are gone or on the way out. That makes sense, as the bots have advanced the top level of go by decades if not centuries. Uberdude recently started a topic ( viewtopic.php?t=17352 ) about what he appropriately dubbed the 3-3 double hane switcheroo ponnuki joseki. An example, from Kim Mili, 3 dan, (White) vs. Cho Hyeyeon, 9 dan, on May 16, 2018, is below.

Click Here To Show Diagram Code: [go]$$Wcm50 Switcheroo shocker $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . X O . . . . | $$ | . . O . X O . . . . . . X O O O O . . | $$ | . . O X . X . . X , . . X X O X X . . | $$ | . . . . . . . . . . . . . O X . . . . | $$ | . . . O . . . . . . . . . O . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . 6 . . . . . . . . . . . . . O X . . | $$ | 8 3 4 . . . . . . . . . . O . O . . . | $$ | . 2 O O 5 . . . . , . . X O X O O . . | $$ | . 7 X X O . . . . . . . . X X X O X . | $$ | . . 9 . X 1 . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . X . | $$ ---------------------------------------[/go]

Instead of capturing

in the usual joseki, Cho opted to hane on the left side and capture

. According to the Elf commentary, by comparison this decision cost Cho a whopping 37% in winrate!

Surprisingly, it seems that the switcheroo is typically a mistake, if not a blunder.

Still, Uberdude found an example where the switcheroo was correct (

), according to KataGo, from Pak Yeonghun, 9 dan, (White) vs. Paek Hongseok, 7 dan, on June 27,2009. Elf agrees, judging that the switcheroo gains 4½% in winrate by comparison.

Pros may not have completely abandoned the switcheroo, as it seems to be appropriate in some positions. However, it apparently can be a game losing blunder, one which even 9 dans have misjudged.

Bill Spight · Post by **Bill Spight** » Tue Apr 07, 2020 12:45 pm

Unplanned obsolescence, II

Mistakes

Can a mistake be a joseki? Obviously, it can. A joseki is simply a standard sequence of play, and over time the community of go players learns better plays in many instances, and joseki become obsolete. OC, sometimes an obsolete move or joseki makes a comeback. For instance, the mini-Chinese made a comeback after 200 years, and the initial 4-4 move made a comeback in Japan after a much longer period. Now it is usually the preferred first move in the corner.

It was known in the mid 20th century that there was something wrong with the sequence for White in the example below, but it was still considered to be joseki.

Click Here To Show Diagram Code: [go]$$Wc $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , 7 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 5 3 . . | $$ | . . . . . . . . . . . . . . . 1 2 . . | $$ | . . . , . . . . . , . . . . . , 4 . . | $$ | . . . . . . . . . . . . . 6 . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

One idea was not to connect at 5, but to occupy the top right corner instead. It took AlphaGo to teach us that the solid connection is fine, but White can tenuki instead of playing the extension.

While not exactly heresy, this was a surprise, because it had been believed that because

,

, and

lacked eye potential, they needed an extension to make a base.

Click Here To Show Diagram Code: [go]$$Wcm6 Tang Yi, 3 dan (W) - Wang Yubo, 2 dan, 2018-07-13 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . X . , X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . a . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . c 7 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . b . . . . . . . . . . . . 5 3 . . | $$ | . . . . . . . . . . . . . . . 1 2 . . | $$ | . . . , . . . . . , . . . . . , 4 . . | $$ | . . . O . . . . . . . . . 6 . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

So is

a mistake? Maybe White should make an enclosure, for instance at a or b. Actually, Elf, in its commentary, prefers the high extension at c. But only by 1%.

Surely a preference that slight does not indicate that the low extension is a mistake.

Actually, the date is a big clue that

is OK. By 2018 pros were checking their opening plays with AI. It is rare to find an opening error by them nowadays.

How about the next example?

Click Here To Show Diagram Code: [go]$$Wcm8 Pak Cheongsang, 9 dan (W) - Yun Chunsang, 7 dan, 2008-07-13 $$ ------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . X . . . . . . . . a . . . . | $$ | . . . O . . . . . , . . . . b , X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , 7 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . c . . . . . . . . . . . . 5 3 . . | $$ | . . . . . . . . . . . . . . . 1 2 . . | $$ | . . . , . . . . . , . . . . . , 4 . . | $$ | . . . O . . . . . . . . . 6 . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Is

a mistake? Should White approach the top right corner at a or b, or enclose the bottom left corner at c? Elf prefers all three to the extension, and prefers the high approach at b by 5%. This game was played before the AI era, OC.

So is

a mistake? Well, maybe. IMHO, an educated guess, is that it is, according to Elf, but just barely. I have enough experience with the Elf commentaries that I was pretty sure that Elf would prefer an approach the the top right corner by enough to say that the extension was a mistake. Now, Elf is only one bot, and the commentary is a year and a half old. Bots are getting better all the time. Maybe another bot would prefer the extension.

Still, as a player I could afford to avoid the extension, even if it is not a mistake. How bad is it to emulate a superhuman player?

But as an analyst I worry about whether the extension is a mistake or not. Pak, a 9 dan pro, preferred the extension. Was he wrong?

While I'm at it, what about the idea of leaving the connection unplayed? Since AlphaGo played it, maybe not playing it is a mistake.

Click Here To Show Diagram Code: [go]$$Wcm10 Ch'oi Ch'eolhan, 9 dan (W) - Kang Tongyun, 9 dan, 2018-02-13 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . 7 . . . . . | $$ | . . . . . . . . . . . . . 4 5 3 1 . . | $$ | . . . O . . . . . , . . . . 6 X 2 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , a . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 9 . . | $$ | . . . . . . . . . . . . . . 0 8 O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

at a

Instead of making the solid connection White jumped into the top right corner. Was that a mistake?

Elf thinks so, to the tune of 9½%. That winrate difference is more substantial than 5%, enough for me to guess that it is a mistake, and that other top bots will agree that it is, if not with the exact estimate of the winrate difference. Besides, it wasn't just the fact that AlphaGo played the connection, it is that I have seen Elf's opinion of other omissions of the connection. In fact, a winrate difference of 9½% is on the low side.

Did a pro 9 dan make a mistake here in 2018? My educated guess is yes. But nobody really knows.

lightvector · Post by **lightvector** » Tue Apr 07, 2020 2:51 pm

Bill Spight wrote: While I'm at it, what about the idea of leaving the connection unplayed? Since AlphaGo played it, maybe not playing it is a mistake.

$$Wcm10 Ch'oi Ch'eolhan, 9 dan (W) - Kang Tongyun, 9 dan, 2018-02-13
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . 7 . . . . . |
$$ | . . . . . . . . . . . . . 4 5 3 1 . . |
$$ | . . . O . . . . . , . . . . 6 X 2 . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , a . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . 9 . . |
$$ | . . . . . . . . . . . . . . 0 8 O . . |
$$ | . . X . . . . . . . . . . . . O X . . |
$$ | . . . , . . . . . , . . . . . , X . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------
Click Here To Show Diagram Code
[go]$$Wcm10 Ch'oi Ch'eolhan, 9 dan (W) - Kang Tongyun, 9 dan, 2018-02-13 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . 7 . . . . . | $$ | . . . . . . . . . . . . . 4 5 3 1 . . | $$ | . . . O . . . . . , . . . . 6 X 2 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , a . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 9 . . | $$ | . . . . . . . . . . . . . . 0 8 O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]
at a

Instead of making the solid connection White jumped into the top right corner. Was that a mistake?

Elf thinks so, to the tune of 9½%. That winrate difference is more substantial than 5%, enough for me to guess that it is a mistake, and that other top bots will agree that it is, if not with the exact estimate of the winrate difference. Besides, it wasn't just the fact that AlphaGo played the connection, it is that I have seen Elf's opinion of other omissions of the connection. In fact, a winrate difference of 9½% is on the low side. Did a pro 9 dan make a mistake here in 2018? My educated guess is yes. But nobody really knows.

A recent KataGo 40 block network (b40c256x2-s2619890176-d761667663, so slightly newer than the ones released, but again probably similar) clearly dis-prefers white's play, enough to probably never choose to play this way itself. But taking the evaluation at face value for now, calling it a mistake might be a stretch, it thinks the difference is pretty slight. See followups and evaluations below.

Click Here To Show Diagram Code: [go]$$Wc Not Kata's preferred way for white $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . X O O O . . | $$ | . . . O . . . . . , . . . . X X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , 1 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 2 3 . . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . X X O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . 4 . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

If white plays the two point extension, Kata expects black to force and then build the bottom with a pincer. Seems pretty straightforward. (Black ~54.5%, +0.5 points)

Click Here To Show Diagram Code: [go]$$Wc Kata's different idea for white $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . X O O O a . | $$ | . . . O . . . . . , . . . . X X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . c b . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . 1 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . d . . O . . | $$ | . . . . . . . . . . . . . e X X O . . | $$ | . . X . . . . . . . . . . f . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Kata actually mildly prefers this odd-looking knight's move, which at first sight feels cramped. My amateur dan understanding is Kata wants to avoid black forcing from above as in the above diagram, and the higher knight move activates some aji in Q5 better. Kata expects black to perhaps play "a" next. If/when black plays "b", white plans to answer with "c", and depending on what the white group on the right needs or what happens on the bottom in different variations, variously plans to use one of d, e, or f to get value out of Q5 and the higher knight move. (Black ~53.5%, +0.4 points).

Running forward a bit more though, actually it seems like some variations end up with black back at about 54.5%, so maybe this ends up being similar in evaluation to the two-point extension, although with a very different resulting game.

Click Here To Show Diagram Code: [go]$$Wc Earlier for white $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . d . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . c . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 1 O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , 2 4 . . . , . . . a . , X . . | $$ | . . . O 3 . 5 . . . . . . b . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Expected line for the solid connect. Next black plays "a" or "b", and white plays "c" or "d" or maybe other moves. As mentioned, clearly preferred to either of the earlier two results: (Black ~50.5%, +0.0 points). But that's quite small of a difference by human standards.

Edit: All of the above was interactively in Lizzie, waiting for often around 60k-120k visits each move. Since the players seem to be Korean, I took a guess and also went with Japanese/Korean rules, 6.5 komi.

Bill Spight · Post by **Bill Spight** » Tue Apr 07, 2020 11:16 pm

lightvector wrote:
Bill Spight wrote: While I'm at it, what about the idea of leaving the connection unplayed? Since AlphaGo played it, maybe not playing it is a mistake.

$$Wcm10 Ch'oi Ch'eolhan, 9 dan (W) - Kang Tongyun, 9 dan, 2018-02-13
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . 7 . . . . . |
$$ | . . . . . . . . . . . . . 4 5 3 1 . . |
$$ | . . . O . . . . . , . . . . 6 X 2 . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , a . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . 9 . . |
$$ | . . . . . . . . . . . . . . 0 8 O . . |
$$ | . . X . . . . . . . . . . . . O X . . |
$$ | . . . , . . . . . , . . . . . , X . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------
Click Here To Show Diagram Code
[go]$$Wcm10 Ch'oi Ch'eolhan, 9 dan (W) - Kang Tongyun, 9 dan, 2018-02-13 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . 7 . . . . . | $$ | . . . . . . . . . . . . . 4 5 3 1 . . | $$ | . . . O . . . . . , . . . . 6 X 2 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , a . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 9 . . | $$ | . . . . . . . . . . . . . . 0 8 O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]
at a

Instead of making the solid connection White jumped into the top right corner. Was that a mistake?

Elf thinks so, to the tune of 9½%. That winrate difference is more substantial than 5%, enough for me to guess that it is a mistake, and that other top bots will agree that it is, if not with the exact estimate of the winrate difference. Besides, it wasn't just the fact that AlphaGo played the connection, it is that I have seen Elf's opinion of other omissions of the connection. In fact, a winrate difference of 9½% is on the low side. Did a pro 9 dan make a mistake here in 2018? My educated guess is yes. But nobody really knows.
A recent KataGo 40 block network (b40c256x2-s2619890176-d761667663, so slightly newer than the ones released, but again probably similar) clearly dis-prefers white's play, enough to probably never choose to play this way itself. But taking the evaluation at face value for now, calling it a mistake might be a stretch, it thinks the difference is pretty slight. See followups and evaluations below.

Many thanks.

I see I was not clear. The 9½% winrate difference (as we know, Elf calculates larger differences than other top bots) is between the solid connection for

and

in the game. To be precise, for the solid connection Elf estimates White's winrate as 48.2% (94,223 rollouts).

in the game, at R-17, got only 1 rollout, so Elf inherited its winrate estimate from Elf's top choice for

, which was Q-17, not the game play, which was the block at R-16. (It would have been preferable to have Elf actually make the play at R-17 instead, but Elf did not do that. {shrug}). That yields a winrate estimate for White of 38.5% (29,001 rollouts). The winrate difference is 9.7%, which I rounded to 9½% because I don't trust the third digit.

I gave the game continuation to give the reader an picture of what was going on. Through

Black has lost a couple of percent by Elf's estimation, while after

all of White's plays through

have been Elf's top choice. I had intended to give around 10 plays from ELf's main variation for the solid connection to show a comparison, but I forgot to do so. I'll do so below.

$$Wc Not Kata's preferred way for white
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . O . . . . . |
$$ | . . . . . . . . . . . . . X O O O . . |
$$ | . . . O . . . . . , . . . . X X X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , 1 . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . 2 3 . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . X X O . . |
$$ | . . X . . . . . . . . . . . . O X . . |
$$ | . . . , . . . . . , . . . . . , X . . |
$$ | . . . O . . 4 . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------
Click Here To Show Diagram Code
[go]$$Wc Not Kata's preferred way for white $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . X O O O . . | $$ | . . . O . . . . . , . . . . X X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , 1 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 2 3 . . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . X X O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . 4 . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]
If white plays the two point extension, Kata expects black to force and then build the bottom with a pincer. Seems pretty straightforward. (Black ~54.5%, +0.5 points)

Elf agrees that

was not optimal, and gives the following in reply.

Click Here To Show Diagram Code: [go]$$Wcm20 Elf also likes Black 21 here $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . X O O O 4 . | $$ | . . . O . . . . . , . . . . X X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , 1 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 2 3 . . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . X X O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

For

Elf estimates Black's winrate as 63.6% (40,357 rollouts). (As we know, Elf likes Black's chances.

)

$$Wc Kata's different idea for white
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . O . . . . . |
$$ | . . . . . . . . . . . . . X O O O a . |
$$ | . . . O . . . . . , . . . . X X X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . c b . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . 1 . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . d . . O . . |
$$ | . . . . . . . . . . . . . e X X O . . |
$$ | . . X . . . . . . . . . . f . O X . . |
$$ | . . . , . . . . . , . . . . . , X . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------
Click Here To Show Diagram Code
[go]$$Wc Kata's different idea for white $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . X O O O a . | $$ | . . . O . . . . . , . . . . X X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . c b . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . 1 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . d . . O . . | $$ | . . . . . . . . . . . . . e X X O . . | $$ | . . X . . . . . . . . . . f . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]
Kata actually mildly prefers this odd-looking knight's move, which at first sight feels cramped. My amateur dan understanding is Kata wants to avoid black forcing from above as in the above diagram, and the higher knight move activates some aji in Q5 better. Kata expects black to perhaps play "a" next. If/when black plays "b", white plans to answer with "c", and depending on what the white group on the right needs or what happens on the bottom in different variations, variously plans to use one of d, e, or f to get value out of Q5 and the higher knight move. (Black ~53.5%, +0.4 points).

Elf agrees, at least with the keima for

.

Click Here To Show Diagram Code: [go]$$Wcm20 Elf agrees on White 20 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . X O O O b . | $$ | . . . O . . . . . , . . . . X X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . a . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . d c . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . 1 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . X X O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . . O . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

In reply it expects Black to play at a, b, or c, in that order of preference. If Black replies at c it also likes the jump attachment at d.

For the keima it estimates Black's winrate as
58.6% (205,470 rollouts). Based on that, maybe

in the game was a minor error, according to Elf. (One reason I still like the Elf commentaries is the humungous number of rollouts.

)

Running forward a bit more though, actually it seems like some variations end up with black back at about 54.5%, so maybe this ends up being similar in evaluation to the two-point extension, although with a very different resulting game.

Not sure what you mean. Are you referring to an existing variation, or are you making each move yourself?

$$Wc Earlier for white
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . d . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . c . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . 1 O . . |
$$ | . . X . . . . . . . . . . . . O X . . |
$$ | . . . , 2 4 . . . , . . . a . , X . . |
$$ | . . . O 3 . 5 . . . . . . b . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------
Click Here To Show Diagram Code
[go]$$Wc Earlier for white $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . d . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . c . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 1 O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , 2 4 . . . , . . . a . , X . . | $$ | . . . O 3 . 5 . . . . . . b . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]
Expected line for the solid connect. Next black plays "a" or "b", and white plays "c" or "d" or maybe other moves. As mentioned, clearly preferred to either of the earlier two results: (Black ~50.5%, +0.0 points). But that's quite small of a difference by human standards.

Edit: All of the above was interactively in Lizzie, waiting for often around 60k-120k visits each move. Since the players seem to be Korean, I took a guess and also went with Japanese/Korean rules, 6.5 komi.

Yes, it was a Korean league game with 6.5 komi. 1 hr. main time, sudden death. Black won by resignation.

For comparison, here is Elf's mainline for the solid connection, up through

.

Click Here To Show Diagram Code: [go]$$Wcm10 Elf's mainline for the solid connection $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 9 . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 8 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . 7 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . 1 O . . | $$ | . . X . . . . . . . . . . . . O X . . | $$ | . . . , 2 4 . . . , . . . 6 . , X . . | $$ | . . . O 3 . 5 . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Elf and KataGo agree through

, maybe through

.

is a nice touch, preventing Black from playing there if White invaded on R-17 instead.

RobertJasiek · Post by **RobertJasiek** » Wed Apr 08, 2020 11:26 pm

Bill Spight wrote:
$$Wcm50 Switcheroo shocker
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . X O . . . . |
$$ | . . O . X O . . . . . . X O O O O . . |
$$ | . . O X . X . . X , . . X X O X X . . |
$$ | . . . . . . . . . . . . . O X . . . . |
$$ | . . . O . . . . . . . . . O . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . 6 . . . . . . . . . . . . . O X . . |
$$ | 8 3 4 . . . . . . . . . . O . O . . . |
$$ | . 2 O O 5 . . . . , . . X O X O O . . |
$$ | . 7 X X O . . . . . . . . X X X O X . |
$$ | . . 9 . X 1 . . . . . . . . . . X O . |
$$ | . . . . . . . . . . . . . . . . . X . |
$$ ---------------------------------------
Click Here To Show Diagram Code
[go]$$Wcm50 Switcheroo shocker $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . X O . . . . | $$ | . . O . X O . . . . . . X O O O O . . | $$ | . . O X . X . . X , . . X X O X X . . | $$ | . . . . . . . . . . . . . O X . . . . | $$ | . . . O . . . . . . . . . O . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . 6 . . . . . . . . . . . . . O X . . | $$ | 8 3 4 . . . . . . . . . . O . O . . . | $$ | . 2 O O 5 . . . . , . . X O X O O . . | $$ | . 7 X X O . . . . . . . . X X X O X . | $$ | . . 9 . X 1 . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . X . | $$ ---------------------------------------[/go]
Instead of capturing :w50: in the usual joseki, Cho opted to hane on the left side and capture :w52:. According to the Elf commentary, by comparison this decision cost Cho a whopping 37% in winrate!

I am unsure whether I understand you. Is the diagram above 37% better than Cho's variation? Please show Cho's variation with a diagram! What was the 37% mistake in it?

Life In 19x19

This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that