Hi Dave

I wasn't reacting to your post directly - but it doesn't matter. Thanks for the addition.

I beg to differ still. First, the notion that whole board considerations may affect how to play in an adjacent or even diagonally opposed corner (ladders) has been around before AI. We were taught to always be selective and not blindly apply joseki. Second, even now that we have a way to quantify the effect, in the majority of cases we'll be talking about small variations. In the example provided by Xela, the original "collapse" by Ishida has been evaluated to be actually favorable. As we have verified, this evaluation holds with any type of (sparse) distribution in the other corners. The magnitude of the evaluation (5-ish) will surpass any global consideration, I reckon.

This leads to two questions:
1) what is the magnitude of global considerations? In the case of a ladder, it will be big. In other cases?
2) which marginal differences remain valid regardless of global considerations? I.e. the difference is relatively small but "absolute".

OMG, half a year of full-time study with AI in vain;) Seriously, what one must not do is to view josekis in local isolation. The global context always plays a role. Some josekis work and some moves fail in most global contexts but either can be different in certain rare contexts.

Joseki study is local tactics, if necessary prolongued global fights and global choice.



Sure, but many still belong to the minority of huge variations.



What is this? 10,000 to 100,000 playouts? Insufficient! I am very sure that it is insufficient as I have gained more than enough related experience, see above.



What kind of magnitude do you mean now?

Your comment was very insightful. Especially the first sentence I thought.

It was also insightful to go a few moves backward and find a position where it isn't, so to speak, too late to start thinking. From that position, it turns out, it is necessary that both sides ignore their best instincts and follow a script or they won't reach the position that was proposed for discussion. I found your explanations to be good. I especially like the following



Great fighting spirit!




What about the hypothesis that making a group in the narrowest part of the board is so unnatural that it shouldn't be unexpected that KataGo abandons it? I'd say that the particular enclosure makes the position feel narrower than some other enclosures would. Maybe this is the plot and the exact moves are the plot twists?



In general when someone is better at something, say better able to "interpret the sequences AI comes up with", it is that they are able to do it differently than you would yourself. In general I'd expect pros to be good at human Go and teaching humans how to play Go. When someone is good at interpreting computer variations it is likely that it is how they use those skills, not only that they have the skill.

It is interesting that you mentioned Mingjiu's lectures. I watched his lectures about the first joseki on more than one occasion. It was long time ago but I think I can confirm that it isn't the ability to "interpret the sequences AI comes up with" that counts when it comes to such rigorous analysis. What counts is the ability to do the same analysis over the board and completely unrehearsed, then there is the ability to do it on the spot when someone asks a tough question. Being able to pick the right pace for a lesson is also important when communicating something complex. It is usually slowing down that is needed. Then there is the ability to sense when a class is no longer able to follow. These are all human skills, some of them are also Go skills, but honestly I think a willingness to use whatever skill you have is a precondition for being good at anything.

So, I don't agree that investigating variations is good type of study "in it self". On the contrary, it can only be a good study when using your own Go skills to accomplish the investigation. One way to accomplish that is to offer a real interpretation, as ez4u has done, another would be to play games using this joseki, which is what I did, I'd like to say not so long ago but time flies. Pondering over them could be a third way.

I don't use Katago much. I agree with the comments above regarding being careful making local interpretations of Katago move choices. It seems to me that moving one stone one space in the diagonally opposite corner can affect evaluation locally. I imagine, given a corner sequence, it would be possible to construct whole board positions, one making the given sequence "bad" and another making it "good". I am thinking that local position evaluation is not well defined. Remember the old Japanese proverb to the effect that there are no joseki for a meijin. Could we say there are no joseki for Katago?

Katago plays its josekis, and even the usual mistakes when exceptionally correct globally, consistently. Katago also respects all stones on the board so a tiny change elsewhere can lead a totally different local continuation indeed. Katago "knows" (or each time reconstructs) josekis but never follows them blindly. The impact of tiny changes elsewhere is limited though and so most decisions can be interpreted by patient (for enough playouts), careful, meticulous, very well pre-educated humans.

You imagine, but those among us who are using KataGo have tried it out. Apart from the obvious ladder breakers, so far I haven't come across diagonally opposed corners to change a joseki more than the precision (margin of error) after X playouts. The biggest influence other corners have on a local position is that, if there are no stones or single stone there, the value of a tenuki in the examined position goes up.

Adjacent corners don't usually offer ladder breakers, but they do have an influence on joseki. In particular pincers are "better" when backed up by same color influence.

Now all these arguments count just as much for the joseki crafted by humans before AI, the difference being that we can now quantify -statistically - the impact of the whole board on local outcomes, whether these local positions are AI or human crafted.

Click Here To Show Diagram Code

[go]$$B connects
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . a . . . . . . . . . . . . . . . . |
$$ | . . X O . O . . . . . . . . . . . . . |
$$ | . . X O O O . . . . . . . . . . . . . |
$$ | . . X O X X O b . . . . . . . . . . . |
$$ | . . O X X X X . . , . . . . . X . . . |
$$ | . . O O O X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

Very often I am blind and I forget to take into account an obvious follow-up that will change my first evaluation of a position.
I am pretty sure that even top professionnals might also (but rarely) be blind by forgetting to take into account a (not really obvious?) follow-up.

Here I suspect it is the case because I do not imagine Ishida saying that black has collapsed in the following diagram:

Click Here To Show Diagram Code

[go]$$B
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . X . , . . . . . , . . . . . , . . . |
$$ | X . X X . . . . . . . . . . . . . . . |
$$ | . X O O . . . . . . . . . . . . . . . |
$$ | . O X O . O . . . . . . . . . . . . . |
$$ | . . X O O O . . . . . . . . . . . . . |
$$ | . O X O X X O X . . . . . . . . . . . |
$$ | . . O X X X X . . , . . . . . X . . . |
$$ | . . O O O X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

The first time I saw this position I confess I was impressed by the 20 white points of territory but here comes the help of katago which considers black position is quite good. After that I reconsidered the position and I said to me : you are a stupid boy, black position is OC far better: let's have a local evaluation of the joseki (without taking into account the whole which I know is a terrible mistake). The white territory has not to be counted 20 points but only 7 points because white has involved one more stone. In addition there is no way to extend this territory in the future. The two black groups are neither weak nor strong but they surrely offer some potential to take points in the future. The white stones in the center are strong but it is certainly not a large wall on which you can push an opponent group. As a conclusion the joseki is locally favorable for black. After this local analysis I agree you have OC to take into account the whole board in order to correct your first evaluation.

Is it possible that this is simply an error in Ishida's joseki dictionary? The most common type of error that I encounter in Go material is swapping black and white.

To me it does look good for black. I'd even go so far as to say black's two groups are strong. It is black to play right?

One model in Go is that it is good to make two strong groups around a singe group in the corner. Flying swallows?

My guess is that - to a pro of Ishida's generation -being squeezed as black is in that joseki sequence seems so unbearable and 'obviously wrong' that it affected the evaluation.

I think one line I learned was like this...

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . .
$$ | . . . , . . . . . , .
$$ | . . . 6 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 8 3 . 7 . . . . .
$$ | . . 0 9 . . . . . . .
$$ | . . 2 . 4 . . . . . .
$$ | . . . , . 5 . . . , .
$$ | . . . 1 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ +----------------------[/go]

Click Here To Show Diagram Code

[go]$$Wm11
$$ | . . . . . . . . . . .
$$ | . . . , . . . . . , .
$$ | . . . X . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . X O . O . . . . .
$$ | . . X O . . . . . . .
$$ | . . X 1 X 6 7 . . . .
$$ | . 9 3 2 4 O . . . , .
$$ | . . 8 O 5 0 . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ +----------------------[/go]

Click Here To Show Diagram Code

[go]$$Wm21
$$ | . . . . . . . . . . .
$$ | . . . , . . . . . , .
$$ | . . . X . . . . . . .
$$ | . . . 6 . . . . . . .
$$ | . . X O . O . . . . .
$$ | . . X O 3 1 . . . . .
$$ | . . X O X X O . . . .
$$ | . O O X X 4 2 7 . , .
$$ | . . X O O X . . . . .
$$ | . . 5 . . . . . . . .
$$ | . . . . . . . . . . .
$$ +----------------------[/go]

...and this doesn't have that many pro games up to . The correct idea in my experience is to tenuki early.

It wouldn't surprise me if this was evaluated incorrectly, it is not immediately clear what happens next. I think I already checked that this isn't equal according to computer evaluation, just not recently. There are only a handful of games in online databases and a good deal of them have close by stones. However, the OP position is different, the shape is fixed. The comment by Ishida was reported to be rather definite though.

I have to add that my approach to such josekis is more of "why is at good enough?" and "why is at good enough?". Not "We have to see if we can play 50 moves in this corner!" but the premise of this topic is territorial vs non-territorial play, but I have noticed AI has a tendency for fill-the-board josekis, despite the myth that AI doesn't like big joseki

To be fair, Shibano's book is from 2021, and much of the research as probably done earlier. It's a long time ago in AI-years. I already noted in 2019 that different AIs can give different recommendations in the same position. And KataGo is not the only AI out there.

On the other hand, the book has an awful lot of "if white follows up this mistake with five more mistakes in a row, then black gets the advantage..." And apart from not much work with real programs, there's not much evidence of work with databases either (references to "noone plays this any more" or "AI invented this sequence" often don't line up with what I can see in GoGoD + Go4Go).

It's certainly interesting to get Shibano's opinions on the new trends, as someone on the front line of playing with or against these new patterns at a high level. But I would have preferred a smaller number of themes and some more in depth discussion, including some middlegame positions to show how the fuseki choices lead to different outcomes.

There's a huge middle ground between "Just blindly put down the moves, you'll be OK, it's a joseki" versus "Everything depends on context and we understand nothing so forget trying to think about joseki". It doesn't need to be all-or-nothing here.

The fact is that if you look at a bunch of AI self-play games, you'll see that certain paterns pop up more than you'd expect by chance. And many of those patterns, if they don't work in every full board context, at least work well more often than not. You might easily conclude that these patterns are significant and should be given a special name... Since AlphaGo Zero, Leela Zero and KataGo have come to similar conclusions here, I think it's proven that joseki are an intrinsic part of go. Yes, they need to be handled with caution. We already knew that.

No, it's not a simple misprint in Ishida. There are multiple diagrams and descriptive explanatory text, so it's very clear that to him, the side being squeezed is the side that comes off worse. It's not a simple case of swapping two words or getting the colours mixed up.

I'm reminded of the many anecdotes about "expert judgement" in Thinking, slow and patronising. Although the book is overconfident on some points, there's a lot of good cautionary tales about how a group of experts can all move together in the wrong direction. The insight of "here comes the help of <new point of view> ... I said to me : you are a stupid boy ..." if you want to make it sound profound instead of comical, you can call it a paradigm shift. In primary school mathematics, we learn a lot of "obvious" things that took centuries to see for the first time. (But I did enjoy Gérard's description!)

See also: https://lifein19x19.com/viewtopic.php?f=12&t=19358

AI invents (or is used to invent) very much faster than professional players adopt so databases of their games report! Roughly 80% of what I find with AI in josekis I have never seen before but, on closer study with AI and my own subsequent study, is very convincing. Therefore, WRT josekis, databases of pro games are often hopelessly outdated for the sake of good joseki choice. These databases remain very useful but, in particular, not for thorough, general joseki study.

I totally agree with xela here. Which doesn't mean I disagree with others per se, in some cases we might misunderstand each other. In my own schematic words:

1. pre-AI there were lots of joseki, established corner patterns, which give locally best play or acceptable variations, which need caution to apply in the context of the whole board
2. every decade or so, these joseki would get updates with the latest professional insights, some joseki going out of fashion, others coming back in, still others being new; the caution still applied
3. the past decade, there has been a major revision due to the advent of AI; some patterns have been discarded, others restored, many new have emerged; the caution still applies;
4. the fundamental difference pre- and post-AI is that we can now (statistically) quantify the results and can likewise quantify their value in a whole board context
5. contrary to a widespread misconception about the current state of AI - which I'm not accusing anyone in particular of - the evaluation remains largely statistical and does not portray one narrow path to victory, i.e. AI is not (yet) deterministic; the whole board mattes, but not (much) more than before AI; the difference in evaluation rarely goes beyond the margin of error and when it does, it is often easy to explain (like a ladder breaker); this is why, IMO, joseki still exist today

Points 1 to 3: ok.

Point 4: One must be careful not to misinterpret statistics! Typical mistakes: too few playouts, reversion statistics misunderstood, overlooking contradictions in statistics only dissolved by much deeper search.

Point 5: Far beyond statistics, go theory can be taken into account with related effort.

I'm referring to the opposite: several times Shibano shows a diagram and says "AI invented this move" (or words to that effect), yet I can see multiple examples in human games from before 2015.

Knotwilg says that joseki exists sill today. I agree with that but I fear that we do not have all the same definition of joseki. OK we can simply say it is a standard sequence of moves in the corner (see https://senseis.xmp.net/?Joseki) but what means "standard" and what means "seqeunce". In particular how do you handle tenuki moves.

Let's take an example:

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . .
$$ | . . . , . . . . . , .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . 3 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 2 . 4 . . . . . .
$$ | . . . , . 5 . . . , .
$$ | . . . 1 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ +----------------------[/go]

After this seqeunce of 5 moves I guess black can play tenuki. If it is true then is this five moves sequence a joseki?

Later in the game black may followed by :

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . .
$$ | . . . , . . . . . , .
$$ | . . . 6 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . 3 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 2 . 4 . . . . . .
$$ | . . . , . 5 . . . , .
$$ | . . . 1 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ +----------------------[/go]

Do you consider to being another joseki or do you consider the move alone as joseki in such situation?

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . .
$$ | . . . , . . . . . , .
$$ | . . . 6 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 8 3 . 7 . . . . .
$$ | . . 0 9 . . . . . . .
$$ | . . 2 . 4 . . . . . .
$$ | . . . , . 5 . . . , .
$$ | . . . 1 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ +----------------------[/go]

Seeing the sequence above it may well happen that tenuki moves can be played after 5, after 6, after 7 ...
What do you call joseki?
What about a sequence of 20 or 30 moves?

In addition remenber that "joseki" could also give a local result which is not even but which can be justied by the environment like the joseki beginning by

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 2 1 . . . . . , .
$$ | . . 3 . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ +----------------------[/go]

'Joseki' can be used with different meanings incl.
- standard sequence
- standard sequence that is correct in at least some positions
- standard sequence that is established by professional players' games and consensus of equality (with the implicit interpretation that different numbers of played stones are taken into account)
etc.

No problem. Just every medium should clarify its intended meaning.

What is your problem with plays elsewhere?! Just annotate them as such in the otherwise local sequence!

Yes, one must be careful! The percentages and winrates shown by KataGo are not statistics in either the Bayesian or frequentist sense (*). At this stage of our knowledge, they are not much more than a heuristic. How many playouts are too few? You can not name an exact number! Whatever number of playouts you do, there could always be more. However deep the search is, there may still be a surprise lurking at the next level. There is no certainty to be found in this direction.

I don't wish to discourage you from exploring. I'm just cautioning you that interpretation of the metrics (not statistics) produced by AI is still some mixture of art and science. When I convey (my interpretation of) KataGo's results in approximate terms, this is not due to ignorance on my part. Quite the opposite: it is a deliberate strategy.

Have you heard the saying "Don't let the perfect be the enemy of the good"?

(*) One can in fact calculate a frequentist confidence interval around the winrate, based on the number of playouts. This is after all what guides the choice of nodes for each playout. So there is an underlying statistic -- but one that's not displayed on the screen in most GUIs, and in any case you need to be very careful indeed about what it's measuring. Discussions in this forum about "margin of error" are generally using the term in a colloquial sense.