Wacky centre move training from bots.

For lessons, as well as threads about specific moves, and anything else worth studying.
Post Reply
Gomoto
Gosei
Posts: 1733
Joined: Sun Nov 06, 2016 6:56 am
GD Posts: 0
Location: Earth
Has thanked: 621 times
Been thanked: 310 times

Re: Wacky centre move training from bots.

Post by Gomoto »

When I review games were a "wacky centre move" was the best option, I encounter this problem:

Many times I am aware of the correct fuzzy target area, but not of the exact best play in this target area.

When I review the variations that show the move suggested by AI is indeed the best option in the local area, these tactical variations are most of the time above my reading depth.

I find it difficult to learn from these complicated variations that are simply above my abilities :)

I start to perceive some tactical combinations, that I was not aware of in the past. But in my games it is most of the time sheer luck if the combinations I try are adequate to the position.

Back to tsume go, were I know that I dont know.
Top
Uberdude
Judan
Posts: 6727
Joined: Thu Nov 24, 2011 11:35 am
Rank: UK 4 dan
GD Posts: 0
KGS: Uberdude 4d
OGS: Uberdude 7d
Location: Cambridge, UK
Has thanked: 436 times
Been thanked: 3718 times

Re: Wacky centre move training from bots.

Post by Uberdude »

Yes, I don't claim studying these examples is the most efficient way to improve, but I find them more interesting than drilling tsumego :) . Just staring at the position with the suggested move should train my neural network a bit and broaden my horizons. Also this one is not so wacky, on reflection it's quite a traditional "about middle of a sector line from around f17 to h8".
Top
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Re: Wacky centre move training from bots.

Post by Bill Spight »

Gomoto wrote:When I review games were a "wacky centre move" was the best option, I encounter this problem:

Many times I am aware of the correct fuzzy target area, but not of the exact best play in this target area.
As a rule, when things are fuzzy, playing close to the best point loses little.
When I review the variations that show the move suggested by AI is indeed the best option in the local area, these tactical variations are most of the time above my reading depth.

I find it difficult to learn from these complicated variations that are simply above my abilities :)
Poco a poco. :)
I start to perceive some tactical combinations, that I was not aware of in the past. But in my games it is most of the time sheer luck if the combinations I try are adequate to the position.
As against that, when your opponent is at the same level as you, that holds for him as well. :blackeye:

Edit: In this pro game White missed the target area altogether, and Black missed the best play. :)
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
Top
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Re: Wacky centre move training from bots.

Post by Bill Spight »

Uberdude wrote:Another example from following along the live Ahan-Tongshan cup final between Gu Zihao and Fan Tingyu (yike english commentary at http://portal.yikeweiqi.com/online/goli ... om=8175754).
Thanks. :) A droll commentary. ;)
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
Top
Uberdude
Judan
Posts: 6727
Joined: Thu Nov 24, 2011 11:35 am
Rank: UK 4 dan
GD Posts: 0
KGS: Uberdude 4d
OGS: Uberdude 7d
Location: Cambridge, UK
Has thanked: 436 times
Been thanked: 3718 times

Re: Wacky centre move training from bots.

Post by Uberdude »

Here's a nice example of a "kakoi" surrounding move (as JF likes to mention) a pro* (black) should have made playing BenseonDarr on Fox, but didn't so got reduced and fell behind. White has just cut at 1, and black sacrificed the left 3 stones to build the top centre.
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X . . . . . . . . . X O . . |
$$ | . O X X X 1 2 . . , . . . . . X O . . |
$$ | . . . . . 3 4 . . . . . . . . X O . . |
$$ | . . . . . 5 . . . . . . . . . X O . . |
$$ | . . . . . . . 6 . . . . . . . X . . . |
$$ | . . . . . 7 . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . X O . . |
$$ | . X O . . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . . . . . O . . . . . X O X . |
$$ | . . . . . . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Seeing as black lost a lot on the left, he should actually consolidate his gains on the top with this kakoi move, as often the case a large knight move from a boundary stone. This would give black 42%. Even fancier black could attach at k4 first to make some kikashi and then kakoi.
Click Here To Show Diagram Code
[go]$$Bcm8
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X . . . . . . . . . X O . . |
$$ | . O X X X O X . . , . . . . . X O . . |
$$ | . . . . . O X . . . . . . . . X O . . |
$$ | . . . . . O . . . . . . . . . X O . . |
$$ | . . . . . . . X . . . . . . . X . . . |
$$ | . . . . . O . . . . 1 . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . X O . . |
$$ | . X O . . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . . . . . O . . . . . X O X . |
$$ | . . . . . . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Instead black tenukid, and BensonDarr made some exchanges then reduced at the k14 sweet spot. LZ 203 would have just played k14 directly (because s8 was a slack answer to s7, thinks black can resist with q8 to avoid giving the sente moves from centre and can deal with s8 bad aji). Black was down to 36% now.
Click Here To Show Diagram Code
[go]$$Bcm8
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X 7 6 . . . . . . . X O . . |
$$ | . O X X X O X . . , . . . . . X O . . |
$$ | . . . . . O X . . . . . . . . X O . . |
$$ | . . . . . O . . . 0 . . . . . X O . . |
$$ | . . . . . . . X . . . . . . . X . . . |
$$ | . . . . . O . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X 9 . |
$$ | . . 1 . . . . . . . . . . . . X O 8 . |
$$ | . X O 2 . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . . . . . O . . . . . X O X . |
$$ | . . 5 3 4 . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Black obediently (-1.5%) answered the reducing move (resist with k13 attachment a little better, then a fight and sabaki), white jumped back and black attacked the shape at f3 (making sense of d2 choice over c2). However, white ignored (how it then settled the wall after black f5 jump was nice) and snipped off the h13 stone and black was down to 31%. What had started out with potential to be a large territory to the 8th line plus outward influence has been cut down to the 4th line, and white got strong (not sure I would say thick) in the centre to boot. All because of failing to play kakoi.
Click Here To Show Diagram Code
[go]$$Bcm18
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X X O . . . . . . . X O . . |
$$ | . O X X X O X . 5 1 . . . . . X O . . |
$$ | . . . . . O X . 4 . . . . . . X O . . |
$$ | . . . . . O 6 . . O . . . . . X O . . |
$$ | . . . . . . . X . . . . . . . X . . . |
$$ | . . . . . O . . . 2 . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X X . |
$$ | . . X . . . . . . . . . . . . X O O . |
$$ | . X O O . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . 3 . . . O . . . . . X O X . |
$$ | . . X X O . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
* Ji Xiang 3p https://www.goratings.org/zh/players/1868.html
Top
Uberdude
Judan
Posts: 6727
Joined: Thu Nov 24, 2011 11:35 am
Rank: UK 4 dan
GD Posts: 0
KGS: Uberdude 4d
OGS: Uberdude 7d
Location: Cambridge, UK
Has thanked: 436 times
Been thanked: 3718 times

Re: Wacky centre move training from bots.

Post by Uberdude »

Can you find the perfect move for White to reinforce his moyo?
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . X . . X O X . . . |
$$ | . . . , . . . . . , . . . O X , X . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . X X X . O . . . . . . . . . . . . . |
$$ | . . O , . . O X . , . . . . . O . . . |
$$ | . . . . X O . O O . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Shin Jinseo did,
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . X . . X O X . . . |
$$ | . . . , . . . . . , . . . O X , X . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . 1 . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . X X X . O . . . . . . . . . . . . . |
$$ | . . O , . . O X . , . . . . . O . . . |
$$ | . . . . X O . O O . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Yeah, this one is not so wacky, but I was impressed Shin hit the sweet spot. Reminds me of a move of Yoda from 3-4 shimari we discussed a while back.
Top
gennan
Lives in gote
Posts: 497
Joined: Fri Sep 22, 2017 2:08 am
Rank: EGF 3d
GD Posts: 0
Universal go server handle: gennan
Location: Netherlands
Has thanked: 273 times
Been thanked: 147 times

Re: Wacky centre move training from bots.

Post by gennan »

I would strongly consider N4
Top
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Re: Wacky centre move training from bots.

Post by Bill Spight »

Uberdude wrote:Can you find the perfect move for White to reinforce his moyo?
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . X . . X O X . . . |
$$ | . . . , . . . . . , . . . O X , X . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . X X X . O . . . . . . . . . . . . . |
$$ | . . O , . . O X . , . . . . . O . . . |
$$ | . . . . X O . O O . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Shin Jinseo did,
I confidently answered no. ;)

But then I did, too. :D Old dogs, new tricks, folks. It's not too late for the human race. :)
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
Top
Uberdude
Judan
Posts: 6727
Joined: Thu Nov 24, 2011 11:35 am
Rank: UK 4 dan
GD Posts: 0
KGS: Uberdude 4d
OGS: Uberdude 7d
Location: Cambridge, UK
Has thanked: 436 times
Been thanked: 3718 times

Re: Wacky centre move training from bots.

Post by Uberdude »

gennan:
I don't like n4 much because:
1) Too close to thick white bottom left group
2) Doesn't actually make the lower right corner white, e.g. still 3-3 or 3-4 attach, so if playing a move on lower side I'd prefer p3 or o3 for real cash
3) Doesn't help mitigate black's q7 shoulder hit aim.
Top
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Re: Wacky centre move training from bots.

Post by Bill Spight »

Uberdude wrote:Here's a nice example of a "kakoi" surrounding move (as JF likes to mention) a pro* (black) should have made playing BenseonDarr on Fox, but didn't so got reduced and fell behind. White has just cut at 1, and black sacrificed the left 3 stones to build the top centre.
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X . . . . . . . . . X O . . |
$$ | . O X X X 1 2 . . , . . . . . X O . . |
$$ | . . . . . 3 4 . . . . . . . . X O . . |
$$ | . . . . . 5 . . . . . . . . . X O . . |
$$ | . . . . . . . 6 . . . . . . . X . . . |
$$ | . . . . . 7 . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . X O . . |
$$ | . X O . . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . . . . . O . . . . . X O X . |
$$ | . . . . . . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Seeing as black lost a lot on the left, he should actually consolidate his gains on the top with this kakoi move, as often the case a large knight move from a boundary stone. This would give black 42%. Even fancier black could attach at k4 first to make some kikashi and then kakoi.
Having answered no to the next question, I spent my time revisiting this one. Lately I have been looking at center play in the Elf commentaries, which has reinforced how little I understand. ;) But neither do the pros, even players like Takemiya with a lot of experience with center play. The main reason, I suspect, is that we just don't evaluate the center and moyos very well. I think Shusaku is undersung in this regard, as he was overshadowed by Go Seigen.
Click Here To Show Diagram Code
[go]$$Bcm8
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X . . . . . . . . . X O . . |
$$ | . O X X X O X . . , . . . . . X O . . |
$$ | . . . . . O X . . . . . . . . X O . . |
$$ | . . . . . O . . . . . . . . . X O . . |
$$ | . . . . . . . X . . . . . . . X . . . |
$$ | . . . . . O . . . . 1 . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . X O . . |
$$ | . X O . . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . . . . . O . . . . . X O X . |
$$ | . . . . . . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
With komi, I think that humans understand that Black is trailing. The question is how to catch up? The question is not just technical, but, for humans, psychological. Since the bots try to increase their chance of winning, against a similar opponent against which they have played millions of games, I think that the psychological metaphor is applicable to their play, too. Against such an opponent, what offers the best chance to catch up? Against a human peer, I suspect that the best chance is in the bottom left corner. :b8: does not press enough.

In a no komi game, however, the kakoi looks a lot more reasonable. It's a play that Dosaku may well have found in the 17th century, despite the unfamiliarity of the position. As well as some players in the 17th century and later. :)
Instead black tenukid, and BensonDarr made some exchanges then reduced at the k14 sweet spot. LZ 203 would have just played k14 directly (because s8 was a slack answer to s7, thinks black can resist with q8 to avoid giving the sente moves from centre and can deal with s8 bad aji). Black was down to 36% now.
Click Here To Show Diagram Code
[go]$$Bcm8
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X 7 6 . . . . . . . X O . . |
$$ | . O X X X O X . . , . . . . . X O . . |
$$ | . . . . . O X . . . . . . . . X O . . |
$$ | . . . . . O . . . 0 . . . . . X O . . |
$$ | . . . . . . . X . . . . . . . X . . . |
$$ | . . . . . O . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X 9 . |
$$ | . . 1 . . . . . . . . . . . . X O 8 . |
$$ | . X O 2 . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . . . . . O . . . . . X O X . |
$$ | . . 5 3 4 . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Black obediently (-1.5%) answered the reducing move (resist with k13 attachment a little better, then a fight and sabaki), white jumped back and black attacked the shape at f3 (making sense of d2 choice over c2). However, white ignored (how it then settled the wall after black f5 jump was nice) and snipped off the h13 stone and black was down to 31%. What had started out with potential to be a large territory to the 8th line plus outward influence has been cut down to the 4th line, and white got strong (not sure I would say thick) in the centre to boot. All because of failing to play kakoi.
FWIW:
Click Here To Show Diagram Code
[go]$$Bcm8
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X . . . . . . . . . X O . . |
$$ | . O X X X O X . . , . . . . . X O . . |
$$ | . . . . . O X . . . . . . . . X O . . |
$$ | . . . . . O . . . . . . . . . X O . . |
$$ | . . . . . . . X . . . . . . . X . . . |
$$ | . . . . . O . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . 3 . . . . . . . . . . . . . . X . . |
$$ | . . 1 . . . . . . . . . . . . X O . . |
$$ | . X O 2 . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . . . . . O . . . . . X O X . |
$$ | . . a . . . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
My feeling is to play :b8: - :b10:. The descent to a is another possibility. In the actual game, Black tried to have it both ways, playing the atari at :b8: to prevent White's extension there, and then going for a bit extra in the corner with the hane-and-connect. That my well be an overplay, but against fellow humans it may offer the best practical chance of catching up. :)
Click Here To Show Diagram Code
[go]$$Bcm18
$$ ---------------------------------------
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . O . O X . X . . . . . . . . . . . . |
$$ | . X O O O X X O . . . . . . . X O . . |
$$ | . O X X X O X . 5 1 . . . . . X O . . |
$$ | . . . . . O X . 4 . . . . . . X O . . |
$$ | . . . . . O 6 . . O . . . . . X O . . |
$$ | . . . . . . . X . . . . . . . X . . . |
$$ | . . . . . O . . . 2 . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X X . |
$$ | . . X . . . . . . . . . . . . X O O . |
$$ | . X O O . . . . . . . . . . O O X . . |
$$ | . . X O . . . . . . . . . . O X X . . |
$$ | . . X O . . . . . , . . . . O O X . . |
$$ | . . X O . 3 . . . O . . . . . X O X . |
$$ | . . X X O . . . . . . . . . . O O X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
I think that the peep at :b20: was Black's aim, accepting mutual destruction. Yes, Black's play seems to have been objectively mistaken, but it still may offer the best chances against a fellow human opponent. :)
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
Top
John Fairbairn
Oza
Posts: 3724
Joined: Wed Apr 21, 2010 3:09 am
Has thanked: 20 times
Been thanked: 4672 times

Re: Wacky centre move training from bots.

Post by John Fairbairn »

Re kakoi and psychology:

I was looking for a book yesterday and just happened to come across the manga Hoshizora no Karasu. I flipped through the first few pages to refresh my memory, and of course came across my alter ego, the grumpy old grandpa pro who just does want to baby-sit little Waka - 'ttaku washi wa gaki no o-mamori nazo shitake nee ga). But he does so by teaching her how to be play go to pro level. But in the robust but subtle way of times of yore - no snowflakes allowed here! On the one hand he lets her help by finding the black moves (the ones hard to read) on a game record he wants to play over, while on the other he bellows at her at the little brat (gaki): "Koko washi no jinchi, kokorahen mo wasi no" - this is MY territory, and all this around here too!

But what really caught my eye (obviously because the skilled manga artist so designed the page) was Waka telling grandpa she understood one important piece of advice: Always only play moves that have an effect on your opponent.

In a way that's trite advice, so obvious that ... Well, so obvious that we keep ignoring it.

But we also ignore it in two different ways. One is just plain forgetting it. The other is more subtle - we are aware of it but deliberately override it. This comes up a lot with prophylactic moves, of which kakoi is one.

There is a nexus of such moves. It includes, in what may be regarded as an ascending scale, mamori (not baby-sitting here but with interesting associations, and at least the easiest to understand), honte, and kakoi - the hardest to understand.

Honte is somewhere in the middle in that it's not too hard to understand, but at the same time it's not too hard either to make a mess of it. Hane Naoki recently wrote a book about honte in which he castigated the "fake honte". This is a safe, solid move that actually doesn't have any effect on the opponent. It is a showboating, grandstanding, or as the ancient Chinese masters would say, a move that looks full but is empty.

One reason that kakoi is so hard is that it is difficult to play it in such a way that it has an effect on the opponent. And this is actually hard even for top pros. Here's a famous example with some AI insight.

In 2011 (i.e. pre AI times) Iyama played the triangle move below. It caused a bit of a sensation at the time as it was a title match and the analysis team in the press room were astounded by it. It was dubbed an "Ear reddening move of the Modern Era".



In Iyama's game, he played the triangled stone to force A and then he surrounded at B.

This same position was recently put to a group of pros. None of them got Iyama's move, but a couple of stronger pros (e.g. an 8-dan) got a similar idea - a move in the same area but not s effective, e.g. G).

Weaker pros such as little Sumire and Ueno Asami opted for E, and in Sumire's case this was combined with an aim at I. Another young pro suggested F with a similar idea. We are talking about moves here that DO have an effect on the opponent, but maybe we should say "impact" rather than "effect" - the effect is too blatant, in a way. Iyama's way was more subtle. H was another youngster's unsubtle choice.

The strongest player, Yuki Satoshi 9-dan, chose the subtlest way: C, expecting White to defend at the triangled point, and then D - surrounding on a truly large scale in which the future effect on the opponent is actually rather hard to see.

But we can assume there was an impact, because modern AI analysis points that way.

I didn't spend much time on it, but LeelaZero naturally liked the 3-3 point in the lower left. Apart from being a stylistic quirk this can be seen as an extemporising move - at some point attention has to revert to the top of the board. And even at this stage, LZ did rate moves in that area highly.

It liked B and C very much. But it also had a soft spot for Iyama's move. The more time it spent analysing, the more time it spent on that move. It did not move up enough to become the "decision move" in the time I spent on it, but it was very much a strong candidate move. None of the moves the younger pros put forward appeared in LZ's list of candidates. Of course the 9-dan, Yuki, was vindicated, too.

Let us switch now to psychology.

I think we can sum up the AI-human dichotomy as follows:

* The bot says "I want to choose the move that gives me the best winning ratio, full stop."
* The human says "I want to choose the move that gives me the best winning ratio, but which is also a move that I can understand."

In other words, my interpretation is that the human is willing to sacrifice some hidden subtlety in return for keeping some sort of grip on the game. But Iyama (and Yuki) saw deeper than younger pros and didn't have to sacrifice quite as much subtlety as them.

Reverting to kakoi, I think one reason it is so hard a concept is that it is hard to find out where its hidden subtlety lies. In part, I believe, this is because people tend to think about the present and the future more than the past. They look at a kakoi move (or any move, come to that) and say, "What effect does it have now?" or, if you are a pro, "What effect could this have in the future?"

Both questions are valid, but there is a third aspect, which, it seems to me, even pros ignore. "What effect will the stones I have already played have if I don't play this kakoi?" In other words, you have to protect what you have already invested. As I have repeated ad nauseam in the past, thickness is only thickness if you can use it as thickness. And the bots have shown us, via the Direct 3-3, that joseki thickness is often not true thickness. If you can attack it, with guerilla tactics, or like a pack of wolves, soon enough, it ends up as a pile of skeletons.

If I may go off on a bit of a tangent now, and pick up on mamori - again not the baby-sitting variety. Its commonest use is as a prophylactic defence of the corner. By chance, I just read this morning an article by Shibano Toramaru in which he explains why the tsukeosae joseki has gone right out of fashion in the AI era. He doesn't use the word mamori, but I think we can see this joseki as having an effect similar to a mamori move (that is the point of the osae block, after all). However, one reason the mamori is so easy to understand (and play) is that its effect is easy to understand: it affects the corner and not (or hardly) the opponent. That is why it is at the bottom of the prophylaxis nexus. Shibano's explanation is that the joseki is bad precisely because it doesn't have enough effect on the opponent.

So, in a nutshell, we come back to what grandpa says: "Always only play moves that have an effect on your opponent." Add subtlety for seasoning depending how many Michelin stars you want to win.
Top
User avatar
ez4u
Oza
Posts: 2414
Joined: Wed Feb 23, 2011 10:15 pm
Rank: Jp 6 dan
GD Posts: 0
KGS: ez4u
Location: Tokyo, Japan
Has thanked: 2351 times
Been thanked: 1332 times

Re: Wacky centre move training from bots.

Post by ez4u »

Uberdude wrote:Can you find the perfect move for White to reinforce his moyo?
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . X . . X O X . . . |
$$ | . . . , . . . . . , . . . O X , X . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . X X X . O . . . . . . . . . . . . . |
$$ | . . O , . . O X . , . . . . . O . . . |
$$ | . . . . X O . O O . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Shin Jinseo did,
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . X . . X O X . . . |
$$ | . . . , . . . . . , . . . O X , X . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . 1 . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . X X X . O . . . . . . . . . . . . . |
$$ | . . O , . . O X . , . . . . . O . . . |
$$ | . . . . X O . O O . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Yeah, this one is not so wacky, but I was impressed Shin hit the sweet spot. Reminds me of a move of Yoda from 3-4 shimari we discussed a while back.
Katago 1.4.1 with the newest 40-block at 800K playouts...
Blue is O7, one space to the right of Shin's choice.
Click Here To Show Diagram Code
[go]$$Wc
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . X . . X O X . . . |
$$ | . . . , . . . . . , . . . O X , X . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . 1 . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . X X X . O . . . . . . . . . . . . . |
$$ | . . O , . . O X . , . . . . . O . . . |
$$ | . . . . X O . O O . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Dave Sigaty
"Short-lived are both the praiser and the praised, and rememberer and the remembered..."
- Marcus Aurelius; Meditations, VIII 21
Top
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Re: Wacky centre move training from bots.

Post by Bill Spight »

John Fairbairn wrote:In 2011 (i.e. pre AI times) Iyama played the triangle move below. It caused a bit of a sensation at the time as it was a title match and the analysis team in the press room were astounded by it. It was dubbed an "Ear reddening move of the Modern Era".



In Iyama's game, he played the triangled stone to force A and then he surrounded at B.

This same position was recently put to a group of pros. None of them got Iyama's move, but a couple of stronger pros (e.g. an 8-dan) got a similar idea - a move in the same area but not s effective, e.g. G).
The game record in the GoGoD database shows Iyama playing at G. I suppose that is a typo in the database. (?)
I didn't spend much time on it, but LeelaZero naturally liked the 3-3 point in the lower left. Apart from being a stylistic quirk this can be seen as an extemporising move - at some point attention has to revert to the top of the board. And even at this stage, LZ did rate moves in that area highly.

It liked B and C very much. But it also had a soft spot for Iyama's move. The more time it spent analysing, the more time it spent on that move. It did not move up enough to become the "decision move" in the time I spent on it, but it was very much a strong candidate move.
Elf in 2018 also chose the 3-3 invasion, with 94.6k rollouts; B was next, with 76.0k rollouts, and C was third, with only 1.6k rollouts. B and C are within Elf's margin of error, I believe. No other candidate got as many os 1.5k rollouts, and so did not show up in the commentary. Elf thought G, which is what it thought Iyama played, lost 13% by comparison with the invasion, making it a definite error.
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
Top
Uberdude
Judan
Posts: 6727
Joined: Thu Nov 24, 2011 11:35 am
Rank: UK 4 dan
GD Posts: 0
KGS: Uberdude 4d
OGS: Uberdude 7d
Location: Cambridge, UK
Has thanked: 436 times
Been thanked: 3718 times

Re: Wacky centre move training from bots.

Post by Uberdude »

ez4u, yes, one to the right was a promising winrate move and when I gave my LZ more than a few thousand playouts it became #1 too.
Top
John Fairbairn
Oza
Posts: 3724
Joined: Wed Apr 21, 2010 3:09 am
Has thanked: 20 times
Been thanked: 4672 times

Re: Wacky centre move training from bots.

Post by John Fairbairn »

The game record in the GoGoD database shows Iyama playing at G. I suppose that is a typo in the database. (?)
No, it was my error here. Sorry. I got misled by the triangled point also being marked in the Japanese article because they were commenting that Iyama showed his genius by not making the obvious cap there.

Compounding my error. the player who I said chose G (Yu Zhengqi) did in fact choose the same move (G) as Iyama.

For the record, two pros chose Direct 3-3, but Hirata Tomoya (one of them) called it a probe, and after a couple of shimmy moves vaguely hinting at a ko, he said he would then switch back to the top side at B. (Kyo Kagen was the other 3-3 player).

Assuming the designation of 3-3 as a probe stands up (and I have seen that several times in other such positions; plus I assume Hirata had AI support for his variation)), I do find it hard to believe that omitting a probe really makes such a HUGE difference, in this sort of position, as Elf claims.
Top
Post Reply

Return to “Study Group”