Knotwilg's practice

Create a study plan, track your progress and hold yourself accountable.
John Fairbairn
Oza
Posts: 3724
Joined: Wed Apr 21, 2010 3:09 am
Has thanked: 20 times
Been thanked: 4672 times

Re: Knotwilg's practice

Post by John Fairbairn »

So far I tried to play a couple of games with 7 "nexuses" in mind. It's too soon to call any result but both of them ended in bad fights for me.
Now that's interesting because...
I also find that by thinking of concepts I'm getting even lazier in reading ahead.
Precisely. I'm afraid it never occurred to me to mention that, in my view, studying concepts via nexuses, or in any other way, should be done away from actual play. It's just too complex. And you need to give your inner brain time to absorb new information (e.g. sleep on it).

If we look at this in terms of hermaneutics, I'd say that the study of go concepts is a way of studying the context of a game.

In the study of texts, the more contexts you can learn to recognise, the more fluent and accurate your reading becomes. You learn contexts by study both inside and outside the text. So, if your text has the phrase, "she baked a cake with a gefurtel", on the supposition you did not know what a gefurtel was (and so didn't know whether the lady used a gefurtel to make the cake or whether she the made a cake flavoured with a nice gefurtel inside it), how would you go about finding out what the sentence meant? Of course, you would start by looking at the context immediately around the word in the text itself. If that didn't reveal an answer, you might turn to a dictionary. If that didn't enlighten you, you might ask around the internet, and if that got nowhere, you might have to go the extreme of contacting the author's family. Or, if he had died 300 years ago, you might have to give up. But in all that research you did, you probably learnt an awful lot of extra, if apparently useless, information. But that useless information may be useful one day.

Go is like that. You can study so as to learn to recognise as many contexts (positions) as possible, yet it may turn out that you come across a position in an actual game where you have no idea what to play next. And even a pro might find himself in that situation. That's the fun of go. But, more often, and even more often the more you study contexts/positions/concepts (another nexus for you :)), you will find that your previous study provides a guide that enables you to read fluently and accurately because the position is now familiar.

In a way, the new nexus I am describing here is really not much more than the old pro advice to study by playing over lots of pro games. The aspect of that advice that is often overlooked, and what turns it into a study nexus, is that you have to engage your brain while you are doing it. I find that too hard.

The reason I cast this post in terms of hermaneutics is that I came across such a hermaneutical problem myself a couple of days ago. It was a sentence (in Chinese) in which Xu Xingyou criticised a player of an older generation (I think it was Sheng Dayou) for having the very bad habit of "dividing influence." That leapt out at me as a major concept, essentially unknown to me. But there was no context. As with the gefurtel example, I didn't know whether he meant the fault was dividing one's own influence (=overconcentration or the like) or dividing the opponent's influence (=becoming a target for attack) - or both. So I have to find the text in which he said this, and hope it was in the context of an actual game commentary. The problem there is that Xu wrote three long books of long commentaries, so I am now poring over his Jianshantang Yipu. If that doesn't come up trumps, I'll just go onto another book, and so on. Eventually, I expect to come face to face with a whole new concept. Or not. But the search is exciting.
dust
Lives with ko
Posts: 161
Joined: Sun Feb 28, 2016 4:01 am
GD Posts: 0
Has thanked: 138 times
Been thanked: 23 times

Re: Knotwilg's practice

Post by dust »

Knotwilg wrote:So far I tried to play a couple of games with 7 "nexuses" in mind. It's too soon to call any result but both of them ended in bad fights for me.

I do find that the "sharing" comes to mind most and perhaps I'm taking that too much as "taking my share" in an area where the opponent is strong, which is why I get under pressure so much.

I also find that by thinking of concepts I'm getting even lazier in reading ahead.

But hey, it's training for a reason!

'Sharing' is an interesting concept - but i personally tend to reframe it as 'don't lose points' in tactical exchanges and taking large moves (possibly reflecting too much time spent with KataGo and the point-based UI).

'Bad fights' and 'under pressure' suggests perhaps that your fights are 'going all-in' or very heavily invested.
John Fairbairn
Oza
Posts: 3724
Joined: Wed Apr 21, 2010 3:09 am
Has thanked: 20 times
Been thanked: 4672 times

Re: Knotwilg's practice

Post by John Fairbairn »

I do find that the "sharing" comes to mind most and perhaps I'm taking that too much as "taking my share" in an area where the opponent is strong, which is why I get under pressure so much.
"Taking MY share" instead of "taking A share"? It sounds like you may have Kobayashi Satoru's Five Envies:

1. Invading too narrow a space
2. Invading or reducing too deeply
3. Invading without due regard to the status of your own group(s)
4. Going too close to thickness
5. Being too concerned with small groups.

But more important than that (though probably less useful) is that you have to remember I mentioned the concept of sharing as a characteristic of old Chinese play. This means that success is not just measured by obvious points of territory gained/lost but also by how many groups result. In old Chinese josekis where sharing/dividing (same character) is stressed, it may, for example, be sufficient or better to ENCROACH while maintaining connection to your outside group (because of group tax).

One big benefit of thinking in terms of encroachment instead of invading (= taking MY share) is that invasions tend to put you on the back foot, so you are always on the defensive. Encroaching, in contrast, tends to take place with you having sente all the time, so not only do you get A share, you end up with the initiative and get the next big move elsewhere (that's the share you really want). Sharing in the Chinese sense also tends to take place in sente, and so is a close cousin of encroachment.

Deictic words such as 'the' and 'me' and 'my' play a big part in European languages, and so influence our thinking. Japanese, Korean and Chinese do not have words for "the" and "a". Hypothesis: they are therefore more attuned to fuzzy thinking?
User avatar
Knotwilg
Oza
Posts: 2432
Joined: Fri Jan 14, 2011 6:53 am
Rank: KGS 2d OGS 1d Fox 4d
GD Posts: 0
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Location: Ghent, Belgium
Has thanked: 360 times
Been thanked: 1021 times
Contact:

Re: Knotwilg's practice

Post by Knotwilg »

I have played 2 5H games against AI Sensei with 5 stones and lost them both. I'm trying to play active but that usually results in a fight and we all know how fighting AI ends.

Next time I'm going to try something different. Instead of attacking, I'll focus on defending my weakest group while developing it too.

Edit: I'm pretty amazed by AI Sensei's interface. This is exactly the kind of offering I expect so I'm inclined to go for a subscription.
John Fairbairn
Oza
Posts: 3724
Joined: Wed Apr 21, 2010 3:09 am
Has thanked: 20 times
Been thanked: 4672 times

Re: Knotwilg's practice

Post by John Fairbairn »

I'm trying to play active but that usually results in a fight and we all know how fighting AI ends.
I don't really know what you mean by playing actively, but your next paragraph suggests ("attacking") it means "aggressively" but even this is open to a wide range of interpretations. If it usually leads to a sharp fight, I suppose we can assume it's at the extremely provocative end of the spectrum, no?

But the word 'attack' is quite rare in old Chinese commentaries. They prefer the word "jin" which I like to translate along the lines of "apply pressure" while pointing out that the root meaning is "tight". In other words, the recommended style of play is at the other end of that portion of the spectrum. The police tactic "kettling" may convey the idea better.

Furthermore, once you are attuned to play in this forceful but restrained way, it is much easier to shift to ideas about defence. Yin and yang are kept together instead of it all being yang.

That, in turn, brings up another quibble about claiming to "playing active(ly)". That suggests you are playing to a style instead of playing the position on the board.

There is a good example of the right approach in game 1640HLS057 for those who have GoGoD. It is one of the Games of Tears of Blood in which Huang Longshi is giving three stones to Xu Xingyou. Xu is praised for a "jin" on move 22 but on move 64 he is chided for being too active :o .

The old commentary says, "64 should be the glide at 68. Although it surrenders the initiative, at a later stage it will be like the effect of being able to enjoy something put away for old age."

In other words, Xu did not play the position. The recommended move would have defended a group that doesn't look weak to the typical amateur - it even has a big base! But the weakness that Huang spotted led to bullying and other, collateral damage of Black.

Although this relates to old go, it's perfectly relevant today. I've written before about the Japanese use of semeru (attack) and semary (apply pressure = jin). They share the same etymological root (also the same as semai = narrow) but the difference is significant in practice. New proverb: For a comfortable old age, learn to semaru instead of to semeru.
Elom0
Lives in sente
Posts: 732
Joined: Sun Feb 20, 2022 9:03 pm
Rank: BGA 3 kyu
GD Posts: 0
KGS: Elom, Windnwater
OGS: Elom, Elom0
Online playing schedule: The OGS data looks pretty so I'll pause for now before I change it.
Has thanked: 1028 times
Been thanked: 32 times

Re: Knotwilg's practice

Post by Elom0 »

John Fairbairn wrote:
I'm trying to play active but that usually results in a fight and we all know how fighting AI ends.
I don't really know what you mean by playing actively, but your next paragraph suggests ("attacking") it means "aggressively" but even this is open to a wide range of interpretations. If it usually leads to a sharp fight, I suppose we can assume it's at the extremely provocative end of the spectrum, no?

But the word 'attack' is quite rare in old Chinese commentaries. They prefer the word "jin" which I like to translate along the lines of "apply pressure" while pointing out that the root meaning is "tight". In other words, the recommended style of play is at the other end of that portion of the spectrum. The police tactic "kettling" may convey the idea better.

Furthermore, once you are attuned to play in this forceful but restrained way, it is much easier to shift to ideas about defence. Yin and yang are kept together instead of it all being yang.

That, in turn, brings up another quibble about claiming to "playing active(ly)". That suggests you are playing to a style instead of playing the position on the board.

There is a good example of the right approach in game 1640HLS057 for those who have GoGoD. It is one of the Games of Tears of Blood in which Huang Longshi is giving three stones to Xu Xingyou. Xu is praised for a "jin" on move 22 but on move 64 he is chided for being too active :o .

The old commentary says, "64 should be the glide at 68. Although it surrenders the initiative, at a later stage it will be like the effect of being able to enjoy something put away for old age."

In other words, Xu did not play the position. The recommended move would have defended a group that doesn't look weak to the typical amateur - it even has a big base! But the weakness that Huang spotted led to bullying and other, collateral damage of Black.

Although this relates to old go, it's perfectly relevant today. I've written before about the Japanese use of semeru (attack) and semary (apply pressure = jin). They share the same etymological root (also the same as semai = narrow) but the difference is significant in practice. New proverb: For a comfortable old age, learn to semaru instead of to semeru.
Wait, so 'style' should only come into effect where you cannot differentiate between moves when trying to play the position? I haven't really been thinking of it that way . . . Moreso as something you try to impose from the outset, buut comparing the two philosophies one instantly sees that the latter is erronous . . . Maybe that's the main problem when amatuers attempt to implement a new style of play for them compared to when professionals do it.
User avatar
Knotwilg
Oza
Posts: 2432
Joined: Fri Jan 14, 2011 6:53 am
Rank: KGS 2d OGS 1d Fox 4d
GD Posts: 0
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Location: Ghent, Belgium
Has thanked: 360 times
Been thanked: 1021 times
Contact:

Re: Knotwilg's practice

Post by Knotwilg »

This 4th game with 5H was very different. I made not a single "mistake" or "blunder", only "inaccuracies" according to AI Sensei. The threshold for mistake is at 4.7 points, which is the 1 kyu study level. Still, the accumulation of inaccuracies plus some uncertainty on behalf of the free version with 50 playouts, accounted for a 4 point loss.

kvasir
Lives in sente
Posts: 1040
Joined: Sat Jul 28, 2012 12:29 am
Rank: panda 5 dan
GD Posts: 0
IGS: kvasir
Has thanked: 25 times
Been thanked: 187 times

Re: Knotwilg's practice

Post by kvasir »

Knotwilg wrote:Still, the accumulation of inaccuracies plus some uncertainty on behalf of the free version with 50 playouts, accounted for a 4 point loss.
Still those seven inaccuracies and the 8th losing move do only total -16.7 points when analyzed deeply with the strongest network. That is only 27% of the total point loss!
User avatar
Knotwilg
Oza
Posts: 2432
Joined: Fri Jan 14, 2011 6:53 am
Rank: KGS 2d OGS 1d Fox 4d
GD Posts: 0
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Location: Ghent, Belgium
Has thanked: 360 times
Been thanked: 1021 times
Contact:

Re: Knotwilg's practice

Post by Knotwilg »

kvasir wrote:
Knotwilg wrote:Still, the accumulation of inaccuracies plus some uncertainty on behalf of the free version with 50 playouts, accounted for a 4 point loss.
Still those seven inaccuracies and the 8th losing move do only total -16.7 points when analyzed deeply with the strongest network. That is only 27% of the total point loss!
Correct, but that's not too discouraging. It means that if I become a considerably better player, removing even inaccuracies from my game and only playing "good moves", I can compete with AI at 4 stones. That's what a 4-5d does, I think. That is under the assumption I play this defensive kind of style. When I go into fighting mode, "blunders" will remain in the analysis because of the higher stakes. That might be a one sided approach to improvement but it's one I'm considering, since going into unpredictable fights is probably one of my biggest problems.
User avatar
Knotwilg
Oza
Posts: 2432
Joined: Fri Jan 14, 2011 6:53 am
Rank: KGS 2d OGS 1d Fox 4d
GD Posts: 0
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Location: Ghent, Belgium
Has thanked: 360 times
Been thanked: 1021 times
Contact:

Re: Knotwilg's practice

Post by Knotwilg »

Same approach, a 2 point loss this time. More mistakes, although AI dropped the threshold to 4 point mistakes it seems.

Here's the review as downloaded with a summary of 5 mistakes. In each variation I have added notes WHY I made the mistake.
I find this very useful and very user friendly.

kvasir
Lives in sente
Posts: 1040
Joined: Sat Jul 28, 2012 12:29 am
Rank: panda 5 dan
GD Posts: 0
IGS: kvasir
Has thanked: 25 times
Been thanked: 187 times

Re: Knotwilg's practice

Post by kvasir »

Knotwilg wrote:Correct, but that's not too discouraging.
I might say it is discouraging if you believe in this method for understanding why you lost a particular game. However I'd rather say it is encouraging that it isn't only about what the computer says.

Somehow this idea that it's all about the computer evaluation is popular despite that it doesn't even pinpoint the greater part of the errors. Not that my analysis is necessarily that good but rather to highlight that there are other things that are important, not only single move mistakes, I'd call the main mistakes as follows:
#56 no need to defend this group, I want to play M10 instead to be in the center (not saying it is the best move but respect your handicap stones)
#76 Q15 instead doesn't give up as much in the corner
#84 as mentioned capture the stones white is peeping at
#164 the bad blunder
#100-#120 white gets too much in the center (which might explain how the lead almost vanishes)

I don't think I can get behind that #40, #50 and #54 are mistakes in a 5 stone handicap game - I think they are good :)
User avatar
Knotwilg
Oza
Posts: 2432
Joined: Fri Jan 14, 2011 6:53 am
Rank: KGS 2d OGS 1d Fox 4d
GD Posts: 0
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Location: Ghent, Belgium
Has thanked: 360 times
Been thanked: 1021 times
Contact:

Re: Knotwilg's practice

Post by Knotwilg »

kvasir wrote: I might say it is discouraging if you believe in this method for understanding why you lost a particular game. However I'd rather say it is encouraging that it isn't only about what the computer says.
If we try learning with AI (rather than from AI) we can only combine AI move evaluation with human interpretation, preferably with someone stronger (like you are to me)

1) When AI evaluates a move as a mistake (>4 pts in my case)
a) and the humans agree and can explain it, there's a clear lesson to be drawn
b) the expert doesn't refute it but neither can the humans explain it, then it's hard to learn from
c) the expert thinks it's a good move, then the expert might be wrong OR the expert makes a good case that the move simplifies the game and removes uncertainty, typically when winning or with a large handicap, like here with #40, #50 and #54

2) When AI evaluates a set of moves as small mistakes, the humans can
a) still point out patterns in these smaller mistakes, and turn it into a lesson how to avoid making them
b) or point to the combined effect (like you pointed to #100-120 neglecting my central handicap stone), drawing a conceptual lesson out of it
c) or disregard them as incidental, not fundamental

3) When AI evaluates a move as the "blue move"
a) then usually humans don't blink, since AI confirms something that was intentional
b) but if the expert points it out as a mistake, like you did with my 2nd line hane at #76 in the lower right, I would assume there's something the expert is overlooking (in this case: sente, I think)
kvasir
Lives in sente
Posts: 1040
Joined: Sat Jul 28, 2012 12:29 am
Rank: panda 5 dan
GD Posts: 0
IGS: kvasir
Has thanked: 25 times
Been thanked: 187 times

Re: Knotwilg's practice

Post by kvasir »

Knotwilg wrote:b) but if the expert points it out as a mistake, like you did with my 2nd line hane at #76 in the lower right, I would assume there's something the expert is overlooking (in this case: sente, I think)
kvasir wrote:Still those seven inaccuracies and the 8th losing move do only total -16.7 points when analyzed deeply with the strongest network. That is only 27% of the total point loss!
I already ran analysis with 10k playouts and the strongest network so this is more a case of the 50 playout version being very wrong.

OK, it's easy to say it's right - the computer says so - but I can try to explain.
Click Here To Show Diagram Code
[go]$$B game
$$ . . . . . . . . . .|
$$ . . O . X X . . . .|
$$ . . . . . O X . . .|
$$ . . . . X O . X . .|
$$ . . . . . O . O 1 .|
$$ , . . . O . X 2 3 .|
$$ . O . . O X . 4 5 .|
$$ . . . . . . . 6 . .|
$$ . . . . . . . . . .|
$$ --------------------+[/go]
Click Here To Show Diagram Code
[go]$$B better
$$ . . . . . . . . . .|
$$ . . O . X X . . . .|
$$ . . . . . O X 3 . .|
$$ . . . . X O 2 X . .|
$$ . . . . . O 1 O 7 .|
$$ , . . . O . X 5 . .|
$$ . O . . O X 6 4 . .|
$$ . . . . . 8 . . . .|
$$ . . . . . . . . . .|
$$ --------------------+[/go]
Both game and "better" variations are sente. I roughly count that there is a 10+ points difference, there is no need to count exactly, it is too much to give up for sente. Black is also simply going to win if he keeps the potential territory, his is closer to be secure than white's. Finally, when dealing with big endgame/territory moves you often have little choice other than to simply take them, you can think about the alternative but the bigger the move the harder it is to make up what you give up for sente.
John Fairbairn
Oza
Posts: 3724
Joined: Wed Apr 21, 2010 3:09 am
Has thanked: 20 times
Been thanked: 4672 times

Re: Knotwilg's practice

Post by John Fairbairn »

Dieter: This seems to me to be a classic example of why it is bad to talk about sente.

Sente in western usage usually means "I can play a move he has to answer." That in turn leads to making inadequate evaluations in terms of "I kept sente." And that's too testerone-driven.

The concept to get hold of here is the "initiative." It is held by the person who has control of the game. Not necessarily by the person who is to move next. As kvasir correctly points out ("Black is also simply going to win if he keeps the potential territory, his is closer to be secure than white's.") Black can put himself in control here. He has the strategic initiative. Sente in the sense of "he has to answer this move" is a mere tactical weapon for mere tactics.

This control aspect is dominant in Chinese commentaries used the characters that can be rendered as sente, but it is actually a recurring theme in Japanese commentaries. It's just expressed in different ways (e.g. tenuki does not mean just playing elsewhere. It means 'skipping a move' so as to take control elsewhere).

Huang Longshi wrote the first lengthy tract on go theory. It is a marvellous work, still of great value today. One of the tenets he puts forward is that calculation is all-important but is not complete until you know who has the initiative, i.e. who is in control of the situation.

Although he uses the characters for sente it would be a gross misrepresentation to translate his words as "calculation is not complete until you know who has sente." For two reasons. One is that that would give the impression of evaluation variations in terms of "who kept sente" in the above deprecated sense. More interestingly, it fails because of the western-language speakers necessary obsession with singular and plural. This is absent in Chinese and Japanese, so so we could more accurately think of "initiative" as being "initiatives."

HLS doesn't go into this point directly (why would he? - he was a Chinese speaker). But there are strong indications that he thought that way. We can see this at one level in the constant criticism of inferior players not seeing the "big picture." This is especially prominent in the commentaries of Xu Xingyou, who got his go theory from HLS.

But at a more arcane level, we can see it in HLS's theory of the Five Lands, or types of area on the go board. Each has its own attributes. E.g. there is the killing ground where hand-to-hand fighting takes place, or the area of communications where groups can make contact with other groups. Other attributes are things like thickness and aji. HLS gives standard ways of addressing the situations in each area, and these can readily be seen as "control measures" (e.g. the "eight sages and the five attacks" - siege referring to surrounding manoeuvres). We can therefore treat each area in such a way that we can assess who has the initiative/control in each area, and who has control of the game as a whole can be assessed by the sum of the area initiatives (plural). You may need to be an HLS, XXY or a Dosaku to do that well, but every journey starts with the first step. In this case, the first step should be to think in terms of control/initiative, not sente. Just as you would (I hope) go hiking in hiking boots, not flip-flops, even though they can both be called footwear.
hl782
Lives in gote
Posts: 316
Joined: Sat May 23, 2015 4:58 pm
Rank: KGS 3D Tygem 5D
GD Posts: 0
Has thanked: 51 times
Been thanked: 40 times

Re: Knotwilg's practice

Post by hl782 »

Yeah I agree with John's comment above. In Korean we say someone lived/died/invaded/reduced in sente/gote (선수/후수) to indicate who kept the initiative of the game after a sequence has played out.

Playing a move that an opponent has to reply is a forcing move, and they tend to be 'sente' because its in the privilege of the player who plays it, but is not indicative of maintaining control of the game imo. I guess this is where American literature is a bit lacking?
Post Reply