It is late in the opening.
Black to play.
As always, assume komi.
Which do you choose among a, b, or c? One is the human's play, one is Elf's top choice, and one is Elf's second choice, with basically the same winrate as its top choice. The human play is very likely a blunder, losing 23½% to Elf's top choice.
Enjoy!
Opening problems for AI: Problem 45
-
Bill Spight
- Honinbo
- Posts: 10905
- Joined: Wed Apr 21, 2010 1:24 pm
- Has thanked: 3651 times
- Been thanked: 3373 times
Opening problems for AI: Problem 45
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
-
Bill Spight
- Honinbo
- Posts: 10905
- Joined: Wed Apr 21, 2010 1:24 pm
- Has thanked: 3651 times
- Been thanked: 3373 times
Re: Opening problems for AI: Problem 45
jlt wrote:Just stopping by to say that I like this series, even though I get most answers wrong.
Truth to say, I was about ready to end it some weeks ago for lack of interest. 
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
-
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:
-
Bill Spight
- Honinbo
- Posts: 10905
- Joined: Wed Apr 21, 2010 1:24 pm
- Has thanked: 3651 times
- Been thanked: 3373 times
Re: Opening problems for AI: Problem 45
was a likely blunder, even with no komi. With komi, Elf reckons that it lost 23½% to the push at 44. It also reckons that 
was a minor error, losing 7½% to occupying the last open corner at a. 
attacks with a boshi, but a was also playable.
and 
create some aji.
plays a shoulder hit against the corner. Then 
continues a large scale center oriented attack. 
pushes down and 
blocks from the corner. Then 
crawls on the 2d line, threatening to connect to the White corner stones. (Because Elf does not show its plays that get fewer than 1500 rollouts, we cannot be sure of its next choice for White.)
solidly connects on the 3-3.
pushes again in the center and then 
- 
push out of the corner into the center in exchange for Black territory on the top left side. The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
-
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: Opening problems for AI: Problem 45
C is not a probe, white has not several options to respond
C loses a ko threat
C is a bad move that does no harm immediatly
Humans can recognize C as a bad move
AI has problems to see C is bad
C loses a ko threat
C is a bad move that does no harm immediatly
Humans can recognize C as a bad move
AI has problems to see C is bad
Last edited by Gomoto on Wed Aug 19, 2020 2:40 am, edited 1 time in total.
-
John Fairbairn
- Oza
- Posts: 3724
- Joined: Wed Apr 21, 2010 3:09 am
- Has thanked: 20 times
- Been thanked: 4672 times
Re: Opening problems for AI: Problem 45
It is a probe. He does not need the opponent to have several options - two is enough: O2 and R2 will doC is not a probe, white has not several options to respond
(a) loses - Dutch people, please note. No offence intended, but the non-word 'looses' really sounds like scratching down a window pane to many natives who would assume it rhymes with 'mooses'. American mis-spellings such as center when made by Brits are even more irritating. And more English speakers in the world follow British spelling then US spellings, I believe. At any rate it's more of an even split than you might imagine.C looses a ko threat
(b) but it also makes future ko threats bigger
Lack of immediate effect sounds plausible but that doesn't make it bad - just wishy-washyC is a bad move that does no harm immediatly
Humans can also recognise C as not a bad moveHumans can recognize C as a bad move
On behalf of the bots: "Sez you and whose army?"AI has problems to see C is bad
My own perception is that A is not good. Many people amateurs play this automatically as a shape move, but pros seem to play it only when it is essential, and I see no urgency here.
I favour B for a slightly odd reason. Until last night I would have assumed a move in the upper-left corner was mandatory for a bot. But I saw a vaguely similar situation in a game by the utterly brilliant Huang Longshi where he played a B-like move, treating it as a probe to see which side the opponent would start off in the corner. Because this is not the fuseki, the principle of prior occupancy of an empty corner does not carry the same force as it would there. It can be treated as a miai situation here. And of course the elsewhere move, B, has a lot going for it in addition to its probing function. (That said, I'd still I'd be very surprised if a Black move in the UR corner is a bad move.)
-
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: Opening problems for AI: Problem 45
O2 is not an option but a losing move. In this sense C would be a probe into the ability of the opponent or should we say a trick move.
I am fine with calling it wishy-washy over bad.
It will be interesting if further AI development will change the evaluation of this and similar moves.
I am fine with calling it wishy-washy over bad.
It will be interesting if further AI development will change the evaluation of this and similar moves.
-
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: Opening problems for AI: Problem 45
I'd say the most likely 2nd choice that makes this a probe is the tenuki choice (with O2 3rd), which links to the making a bigger ko threat point. This black 1 is a few points bigger as a ko threat than q2 as a ko threat (say 34 points versus 30 points, pulling numbers out of the air), so if there's a ko happening which needs a >30 and <=34 point threat then it's good if you have already made this exchange because it generated a valid threat where you otherwise wouldn't have one, but if it's a ko for which <30 is valid then it's bad to have wasted 1 threat, and if the ko needs >34 then it didn't matter now and depends on the size of future kos. But I don't yet foresee any such 30-34 point kos developing and doubt Elf does either, so it's probably more a horizon effect of making incidental exchanges to get more stones on the board and increase the confidence of the winrate.John Fairbairn wrote:It is a probe. He does not need the opponent to have several options - two is enough: O2 and R2 will doC is not a probe, white has not several options to respond
(b) but it also makes future ko threats bigger
The probe-do-you-wanna-tenuki is similar to why bots like to make this exchange early: if you do it latter when there's a big fight that you want it to be a sente exchange in there's a danger it's not big enough and gets tenukid.
I ne-er thought I'd see the day!John Fairbairn wrote:On behalf of the bots: "Sez you and whose army?"AI has problems to see C is bad
C'est moi?!John Fairbairn wrote: American mis-spellings such as center when made by Brits are even more irritating.
-
Bill Spight
- Honinbo
- Posts: 10905
- Joined: Wed Apr 21, 2010 1:24 pm
- Has thanked: 3651 times
- Been thanked: 3373 times
Re: Opening problems for AI: Problem 45
John Fairbairn wrote:American mis-spellings such as center when made by Brits are even more irritating. And more English speakers in the world follow British spelling then US spellings, I believe. At any rate it's more of an even split than you might imagine.
Online Etymology Dictionary wrote: The spelling with -re was popularized in Britain by Johnson's dictionary (following Bailey's), though -er is older and was used by Shakespeare, Milton, and Pope.
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
-
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: Opening problems for AI: Problem 45
Is this a test for the Dutch speakers, John?John Fairbairn wrote:
(a) loses - Dutch people, please note. No offence intended, but the non-word 'looses' really sounds like scratching down a window pane to many natives who would assume it rhymes with 'mooses'. American mis-spellings such as center when made by Brits are even more irritating. And more English speakers in the world follow British spelling then US spellings, I believe. At any rate it's more of an even split than you might imagine.
-
John Fairbairn
- Oza
- Posts: 3724
- Joined: Wed Apr 21, 2010 3:09 am
- Has thanked: 20 times
- Been thanked: 4672 times
Re: Opening problems for AI: Problem 45
Well, Shakespeare couldn't even spell his own name consistently. And let us not forget that American spellings were popularised by a Scot.The spelling with -re was popularized in Britain by Johnson's dictionary (following Bailey's), though -er is older and was used by Shakespeare
Sorry, but I don't understand the question. I'm just observing that Dutch people seem particularly prone to loose usage of lose. I've seen it picked up on several other non-go forums, so it's not just me being picky.Is this a test for the Dutch speakers, John?
It's not really that it irritates me. It's more of a fascination. Dutch people are generally so superb at English (and other languages) that to see this blindspot over and over again makes me wonder what is going on. The irritation I feel is that, even as a linguist, I don't understand how it comes about. (The only other constant hiccup I can think of by Dutch people is "an applause". This is one of Andre Rieu's solecisms, who overuse of it makes me cringe, even though he is otherwise one of my favourite entertainers. But at least I can see where it comes from.)
-
Bill Spight
- Honinbo
- Posts: 10905
- Joined: Wed Apr 21, 2010 1:24 pm
- Has thanked: 3651 times
- Been thanked: 3373 times
Re: Opening problems for AI: Problem 45
Neither could Maleore.John Fairbairn wrote:Well, Shakespeare couldn't even spell his own name consistently.The spelling with -re was popularized in Britain by Johnson's dictionary (following Bailey's), though -er is older and was used by Shakespeare
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.