It is currently Fri Mar 29, 2024 1:19 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 25 posts ]  Go to page Previous  1, 2
Author Message
Offline
 Post subject: Re: Applying endgame principles in other phases of the game
Post #21 Posted: Wed Oct 21, 2020 5:47 am 
Judan

Posts: 6087
Liked others: 0
Was liked: 786
Strong Ai play can be far from perfect play unless AI uses proof play or mathematically proven expertise of perfect play.

Top
 Profile  
 
Offline
 Post subject: Re: Applying endgame principles in other phases of the game
Post #22 Posted: Wed Oct 21, 2020 6:38 am 
Honinbo

Posts: 9545
Liked others: 1600
Was liked: 1711
KGS: Kirby
Tygem: 커비라고해
Knotwilg wrote:
John Fairbairn wrote:
Quote:
For me the alphago move was an easy guess thanks to Uberdude's opening gospel based on LeelaZero.


I think it may be time for St Dude to take up his hammer and chisel again, if he is willing to engrave more tablets.


I'm not 100% sure but I think Dude used the word gospel tongue in cheek, or at least as a catch phrase, which is the way I have adopted it. Unfortunately I don't have the time and energy to follow the pro evolution post-AlphaGo and it's very comforting to see that human experts keep trying to be creative and keep studying the opening for themselves, to try and have an edge over those simply obeying any "gospel".

One thing that has definitely not been satisfactorily explained or understood, is why a 3-3 invasion would be so urgent that it's played almost without hesitation while a 3-3 opening would be so slow that it's hardly ever played. Might be a perversion (self confirming loopholes) by the algorithms. If the pros now move to more 3-3 openings, that seems to indicate AI is indeed not fully onto the truth there.


I'm a fan of Uberdude's Opening Gospel, but even early on when I was trying it out, I noticed that it wasn't always perfectly aligned with the top recommendations I'd see on KataGo. But in a lot of cases, you'll still be somewhat on track, even if you've lost a small amount.

A common example I'll see is the shape in the bottom right here:
Click Here To Show Diagram Code
[go]$$Bc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X O . . . . . . . . . . . . a . . |
$$ | . . X O . . . . . , . . . . . X . . . |
$$ | . . X O . . . . . . . . . . . . . . . |
$$ | . X O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . W . . |
$$ | . . . O . . . . . , . . . . X , . . . |
$$ | . . . . . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Usually when I analyze with KataGo with that kind of shape, it's somewhat big to help out that white approaching stone. Strictly following the gospel, I don't see a reason to do it, though. According to the gospel, I should probably 3-3 invade in the top right.

But my version of KataGo really seems to like playing here:
Click Here To Show Diagram Code
[go]$$Bc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X O . . . . . . . . . . . . . . . |
$$ | . . X O . . . . . , . . . . . X . . . |
$$ | . . X O . . . . . . . . . . . . . . . |
$$ | . X O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . W . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . O . . . . . , . . . . X , . . . |
$$ | . . . . . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


Implicitly, there's the idea of an "urgent move" in the gospel, so it could be argued that helping this stone is urgent. But without defining "urgent", the other rules of the gospel can be broken at any time for the sake of playing something that seems urgent.

That being said, the 3-3 invasion in the top right is still an option KataGo is considering early on here, and I don't think it'd be that bad to play it.

_________________
be immersed

Top
 Profile  
 
Offline
 Post subject: Re: Applying endgame principles in other phases of the game
Post #23 Posted: Wed Oct 21, 2020 8:45 am 
Judan

Posts: 6087
Liked others: 0
Was liked: 786
Kirby wrote:
But my version of KataGo really seems to like playing here:
Click Here To Show Diagram Code
[go]$$Bc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X O . . . . . . . . . . . . . . . |
$$ | . . X O . . . . . , . . . . . X . . . |
$$ | . . X O . . . . . . . . . . . . . . . |
$$ | . X O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . W . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . a b . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . O . . . . . , . . . . X , . . . |
$$ | . . . . . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]



Either KataGo is wrong or an old book by a Japanese pro saying that the white move was in a bad position because the exchange Black a - White b favoured Black.

Top
 Profile  
 
Offline
 Post subject: Re: Applying endgame principles in other phases of the game
Post #24 Posted: Wed Oct 21, 2020 8:49 am 
Oza

Posts: 3647
Liked others: 20
Was liked: 4626
Quote:
That being said, the 3-3 invasion in the top right is still an option KataGo is considering early on here, and I don't think it'd be that bad to play it.


Agreed, but with one caveat. This is still just an unformed notion in my mind, but I think the word 'invasion' needs looking at again in the light of AI wisdom.

In military terms. invasion has strong associations with occupation, and I think, we carry that over to go. When we invade we look to live, and that further plays into the idea of living as big as possible. Obviously some invasions end in a running battle, but that's part of go being a two-person game, and equally obviously invading and living can sometimes be too small and so be the wrong strategy. But, on the whole, I'd say the ideal invasion involves occupying the opponent's putative territory with a healthy life. That in turn means we accept being attacked and surrounded, with the result that most such "good" invasions are gote.

But what I notice about AI play is that bots don't seem to make gote invasions, except in extremis. They will happily start sente invasions, but if things go pear-shaped they seem quick to abort, so as to keep the initiative. In my own mind I think of such bot play not as invasions but rather as 'raids'. In guerilla warfare is often a better model for go than regimental warfare.

This is not a new idea, at least for me. If you look at old Chinese games, invasions are relatively rare. The explanation is usually quite simple. Group tax applies. If you invade and live you are making a new, separate group, so there's -2 straightaway, and there's a strong chance the opponent, in enveloping you, will join two of his groups up, so there's another -2 points.

What you normally see instead is what I call 'encroachment' - a sort of mini-invasion or incursion (侵) where you make sure your encroaching forces stay firmly connected to friendly stones. And if possible you also try to drive in the wedge in such a way that you either split the enemy (分), or at least confirm his groups stay separated.

This is turn means that boundary plays have a much bigger profile in OC go, and explains why the Guanzi Pu (often incorrectly thought of as being a manual of endgame play) was such an important book - it was a novel manual of boundary plays. It does not feature many problems of the type you might expect if you have the word 'endgame' in your mind, i.e. those that have to do with making an extra point or so. But it does have a goodly number of relatively large-scale positions where the point is to make a mini-invasion or 'encroachment'. There are even encroachment josekis. Maybe the best known is the 3-6 Encroachment or 三六侵分, often played in the late opening even. 侵分 can be split up to mean 'intrude and separate' but is actually a single go word which was important enough to be taken over from Chinese to Japanese to mean 'yose' (i.e. boundary plays). It was common in Japan pre-war (rather than guanzi 官子), but usually required furigana to show the unexpected reading yose. That does not sit well with the modern temper, so we now get a simpler kanji or even just katakana.

All that is to make a case that 'invasions' can be and indeed have been re-thought in the evolution of go theory. So there is no reason why we should not ponder whether further re-thinking is necessary in the latest stage of evolution. Call them just sente invasions and gote invasions if you prefer, but in botdom I think it's important now to add the attributive.

Top
 Profile  
 
Offline
 Post subject: Re: Applying endgame principles in other phases of the game
Post #25 Posted: Wed Oct 21, 2020 10:18 am 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
John Fairbairn wrote:
But, on the whole, I'd say the ideal invasion involves occupying the opponent's putative territory with a healthy life.


Pour moi, the ideal invasion is one where it is unclear whether to go for the kill or not. (In practice, they go for the kill. ;))

One thing that has impressed me about the bots is how often they prefer very shallow reductions. Something that, like a lot of AI play, I associate with Go Seigen. :)

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 25 posts ]  Go to page Previous  1, 2

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group