How do you think about this joseki?

oren · Post by **oren** » Sun Oct 25, 2015 10:02 am

I was taught it depends on the stones in the bottom right. In general it's not joseki but can be played depending on surrounding situations.

RobertJasiek · Post by **RobertJasiek** » Sun Oct 25, 2015 10:45 am

I do not have numbers for them. Probably I studied more non-joseki variations than joseki variations. I do not recall any (non-fighting, middle game like variation) not fitting my theory. Professional judgements hardly offer a systematic analysis applicable to several (even many) joseki positions. So all I could do was comparing to pro dictionary statements "this is joseki", study GoGod database frequencies etc. Apart from outdated josekis, everything agrees to my theory on the rough this-is-joseki-statement-level, where information is available.

Charles Matthews · Post by **Charles Matthews** » Sun Oct 25, 2015 11:36 am

RobertJasiek wrote:Apart from outdated josekis, everything agrees to my theory on the rough this-is-joseki-statement-level, where information is available.

RobertJasiek wrote:
MinjaeKim wrote:Are there some known methods for such evaluation?
Yes: my joseki evaluation method.

<snip>

Stone difference = 6 - 6 = 0. // No adjustment is needed.
Territory count = 17 - 6 = 11.
Influence stone difference = 1 - 6 = -5.
Ratio = |11 / (-5)| = 2.2.

The ratio is within the valid range [1.5..3.5] for ordinary josekis without special advantages or disadvantages for a player so the result is equivalent to joseki.

Hmmm ... an empirical formula, based on one algorithmic number, one semi-algorithmic which charitably speaking wouldn't be too debatable, and one "fudge factor". I wasn't aware that this sort of data-mining had gone on.

I don't object to the three dimensions: from a teaching point of view including the "local tally" pleases me.

I tend to object on logical grounds to any confusion of necessary and sufficient conditions. This sort of formula can hardly be yielding a sufficient condition for joseki. My reasoning: pros would not play a line that clearly lost 1 point. But that would move the number only 0.2 in this example? So the calculated result cannot be robust enough to make those discriminations, if it is near the central value 2.0.

So, mainly a test for the plausibility of joseki-like exchanges, that can filter out the "funny looking kids"? But there is perhaps more, if there are discrete "niches" to be understood here.

RobertJasiek · Post by **RobertJasiek** » Sun Oct 25, 2015 11:55 am

The ratio type of josekis is part of the theory. There are a few other types, which are necessary. E.g., the influence stone difference 0 falls into other types also to avoid division by zero.

There are more than three dimensions, but three of them are numerical and involved in the calculations. Also therefore the numerical part is not sufficient.

Since the evaluation is an approximation, very similar local results cannot be distinguished in this manner. If everything else is equal (same stone difference, same outside shapes, similar inside aji) but one sequence is, say, 2 points of territory better than the other, it can happen that the method does not detect this. However, such is scarce in josekis or joseki-like sequences. Almost always, the outdated variation creates a different shape on the outside than the modern variation. Nevertheless, it can happen that the outdated variation is better in some global positional environments. Being 2 points better locally does not say "always better in the global context". My method is more tolerant, and this is a good thing because knowledge of a 2 points better variation must not inhibit strategic flexibility. What my method does not do is to find the 2 points better variation automatically. It is not a variations generator.

Bill Spight · Post by **Bill Spight** » Sun Oct 25, 2015 12:13 pm

Charles Matthews wrote:This sort of formula can hardly be yielding a sufficient condition for joseki. My reasoning: pros would not play a line that clearly lost 1 point.

Excellent point!

Kirby · Post by **Kirby** » Sun Oct 25, 2015 7:31 pm

It's playable for black if black has stones in the bottom right. Otherwise, white has too much influence.

I've used this a lot in games. When white's influence/thickness isn't too much and white takes gote, I'm happy. At my level, white sometimes tenuki, and you have to punish it:

Click Here To Show Diagram Code: [go]$$c (White tenukis) $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . X . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . X 5 . | $$ | . . . . . . . . . . . . . . . O O 3 . | $$ | . . . . . . . . . . . . . . . . . 4 . | $$ | . . . . . . . . . . . . . . . 2 1 . . | $$ | . . . . . . . . . . . . . . . 7 . . . | $$ | . . . , . . . . . , . . . . . , O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Anyway, it doesn't really matter what other people think about the position regarding if it's joseki or not.

How do you feel about it? Just keep it in your pocket as a possible option during your games, and when you think the resulting situation will be good for you, pull it out.

ez4u · Post by **ez4u** » Sun Oct 25, 2015 8:03 pm

Kirby wrote:It's playable for black if black has stones in the bottom right. Otherwise, white has too much influence.

I've used this a lot in games. When white's influence/thickness isn't too much and white takes gote, I'm happy. At my level, white sometimes tenuki, and you have to punish it:

$$c (White tenukis)
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . X . . . . . |
$$ | . . . , . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . X 5 . |
$$ | . . . . . . . . . . . . . . . O O 3 . |
$$ | . . . . . . . . . . . . . . . . . 4 . |
$$ | . . . . . . . . . . . . . . . 2 1 . . |
$$ | . . . . . . . . . . . . . . . 7 . . . |
$$ | . . . , . . . . . , . . . . . , O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------
Click Here To Show Diagram Code
[go]$$c (White tenukis) $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . X . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . X 5 . | $$ | . . . . . . . . . . . . . . . O O 3 . | $$ | . . . . . . . . . . . . . . . . . 4 . | $$ | . . . . . . . . . . . . . . . 2 1 . . | $$ | . . . . . . . . . . . . . . . 7 . . . | $$ | . . . , . . . . . , . . . . . , O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]
Anyway, it doesn't really matter what other people think about the position regarding if it's joseki or not.

How do you feel about it? Just keep it in your pocket as a possible option during your games, and when you think the resulting situation will be good for you, pull it out.

Half the pro games in my database where this appears, White tenukis after 5 with no special advantage to either side. There is no example of 7. After White cuts, how does Black 'punish' White?

Kirby · Post by **Kirby** » Sun Oct 25, 2015 9:25 pm

You can play S12 then, and white is split.

You don't have to do it immediately, but since this exists, white didn't get much of a position locally compared to if he takes gote in the "joseki". Of course, it depends on the rest of the board.

In some board positions, it may well be better for white to tenuki. And in some board positions, it may not be good for black to select this line of play in the first place. But that's the same with anything.

Locally, though, I think black 7 above is often a good followup for black to split white up.

MinjaeKim · Post by **MinjaeKim** » Sun Oct 25, 2015 10:07 pm

RobertJasiek wrote:MinjaeKim, my justification is having studied a representative selection of 400 josekis (that do not necessarily become middle game fights immediately) of which 399 comply to my theory and 1 does not comply because its judgement depends on a possible middle game ko, for which ko evaluation theory is required.

My method presumes that both players have strived to create good shapes. In this example, you can verify the existence of good shapes by the created thick shapes. WRT to numbers of influence stones, my method works with approximate counts, which tolerate counting 1 per stone despite different shapes. I have worked out the theory to distinguish significant influence stones to be counted from not counted stones.

You cannot "presume that both players have strived to create good shapes". Some variations get avoided because of shape issues. One of the simplified variations of the taisha joseki is usually not played because it gives black better shape. See below.

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

Click Here To Show Diagram Code: [go]$$c joseki? $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . 1 . . . . | $$ | . . . , . . . . . , . . . . . , 2 . . | $$ | . . . . . . . . . . . . . . 5 . 4 . . | $$ | . . . . . . . . . . . . . . . 3 . . . | $$ | . . . . . . . . . . . . . . . . 6 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

But as I understand, your method will give the same number to both.

Charles Matthews wrote:So, mainly a test for the plausibility of joseki-like exchanges, that can filter out the "funny looking kids"? But there is perhaps more, if there are discrete "niches" to be understood here.

RobertJasiek wrote:Since the evaluation is an approximation, very similar local results cannot be distinguished in this manner. If everything else is equal (same stone difference, same outside shapes, similar inside aji) but one sequence is, say, 2 points of territory better than the other, it can happen that the method does not detect this. However, such is scarce in josekis or joseki-like sequences. Almost always, the outdated variation creates a different shape on the outside than the modern variation. Nevertheless, it can happen that the outdated variation is better in some global positional environments. Being 2 points better locally does not say "always better in the global context". My method is more tolerant, and this is a good thing because knowledge of a 2 points better variation must not inhibit strategic flexibility. What my method does not do is to find the 2 points better variation automatically. It is not a variations generator.

Okay now I understand what you are doing. Still, I believe you should have more justification than simply that 399 of 400 joseki positions suggested by pros fit to your formula. There's also some chicken and egg problem here since you're assuming the pros have made a correct decision to file out locally optimal exchanges as 'josekis'. You haven't yet explained what makes up a joseki, what comes before a sequence of locally optimal plays is completed.

MinjaeKim · Post by **MinjaeKim** » Sun Oct 25, 2015 10:18 pm

Kirby wrote:How do you feel about it? Just keep it in your pocket as a possible option during your games, and when you think the resulting situation will be good for you, pull it out.

I very often think about playing this variation but have played it out few times. The diagonal attachment starting this joseki is often called a bad move. But the end result do seem playable for both. I found 36 pro games starting with the invasion in the middle, so professional players seem to think at least this is playable. I found this variation in a book which is a dictionary of new josekis, and I was very surprised when the author Kim Seongrae 8p named this as '정석 (joseki)'. I have had many thoughts about this position, so I happened to ask for people's opinion.

RobertJasiek · Post by **RobertJasiek** » Sun Oct 25, 2015 11:22 pm

MinjaeKim, I have provided more justification elsewhere. "Chicken and egg": I have used pro input for the initial calibration of the ratio range; now afterwards, my method stands on its own. The method provides a joseki evaluation, but you are right that this evaluation alone is not a definition of what is a joseki and everybody is right that a local joseki evaluation does not clarify whether a variation fits a global strategic context. Reading and strategic planning remain necessary.

However, for josekis in their model discussion, there is no such thing as "a sequence of locally optimal plays" because "optimum" must be dynamic and allow a set of possible aims to choose from. For different possible strategic aims, different sequences can be locally optimal. What you can do is to assign every joseki-like variation a superset of set of aims for which it can be good in at least some global contexts.

Kirby, it is not only the amount of generated influence you are interested in but the amount of influence that can be used well.

oren, maybe not everybody speaks of a joseki, but maybe some would call such a sequence a joseki-like follow-up sequence:)

Kirby · Post by **Kirby** » Mon Oct 26, 2015 12:04 am

RobertJasiek wrote: Kirby, it is not only the amount of generated influence you are interested in but the amount of influence that can be used well.

Yes, I agree. Maybe it wasn't clear from my post, since I didn't explicitly state it like this.

That's why I think this is okay for black when white doesn't have a position in other parts of the board (e.g. bottom right) where white's resulting influence can be used well.

Charles Matthews · Post by **Charles Matthews** » Mon Oct 26, 2015 6:19 am

RobertJasiek wrote:However, for josekis in their model discussion, there is no such thing as "a sequence of locally optimal plays" because "optimum" must be dynamic and allow a set of possible aims to choose from. For different possible strategic aims, different sequences can be locally optimal. What you can do is to assign every joseki-like variation a superset of set of aims for which it can be good in at least some global contexts.

It would maybe be clearer not to use a term like "joseki", which is anyway ambiguous. If one speaks of say "dictionary joseki" one gets more context. Before the go database, there were not only small (one or two hundred) and large (up to 5000 and more) collections made of corner sequences; and typically standardised judgements attached (for that time, country, and implicitly editorial policy).

I recently bought a modern Japanese joseki dictionary, and opening it at a random page the judgements on three diagrams are "joseki", "former pattern" and (I think) "unclear" as a continuation of that former pattern. A big part of amateur appreciation and improvement is reverse engineering of pro judgements. So, we should want to know at least what judgement is involved in the verdict "joseki".

It is also true that judgements made on corner patterns without context (i.e. sub-board judgements) are of only a limited value. Another part of amateur appreciation and improvement is reverse engineering of "joseki choice". And another part, I have to say, is "looking at sides as a unit", which I made much of on Sensei's Library a dozen years ago, when the go databases started to kick in.

Still, Robert's data mining seems to be much to do with "dictionary joseki", and this is the sort of area that SFKs might well find useful, as a pedagogic point. Go is a global game, and therefore there is no such thing as "a sequence of locally optimal plays" is a sort of knockdown argument for anything at all. But I don't think it should be used that way. It is properly used as a caveat. (As for example if you are discussing the endgame, and have to say "unless ko threats are important".)

What I think would help me (or at least my knowledge of Japanese, and Korean because the big joseki dictionary I have is a Korean edition) is to shade what Robert has done with some other of the stock judgements that resemble shorthand descriptions.

As for the sequence originally posted, I would classify it as "joseki" in the sense of set pattern, and also in the sense of "shape-fixing pattern" (an underestimated concept). But not as "dictionary joseki" for at least a couple of reasons (as has been said, the sample of sequences is sparse, and also it is a follow-up sequence).

RobertJasiek · Post by **RobertJasiek** » Mon Oct 26, 2015 8:47 am

It is more than reverse-engineering: it also involves invention of previously not described theory.

Bantari · Post by **Bantari** » Mon Oct 26, 2015 1:01 pm

RobertJasiek wrote:White 4 on the board or not on the board alters the degree of the white thickness greatly. Therefore, White 4 contributes to the great outside influence as much as the other white stones.

True, and as I understand you, this is wht White 4 is counted as "influence" stone in your calculations.

But cannot the same be said for some of the Black stones, most notably Black Q16? I would say it has a big influence on the Black thickness and on the effectiveness of the other Black stones, not to mention on the number of territory points. So should it not also be counted as "influence stone" for Black?

And if not, where is the line separating such "stones important for thickness greatly" from stones which are important for thickness but not quite that greatly?

Life In 19x19

How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?

Re: How do you think about this joseki?