Opening problems for AI: Problem 36

Bill Spight · #1

Another ancient game. But assume komi, as always.

White to play.

Click Here To Show Diagram Code: [go]$$Wcm28 White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . X X . . O O . . . . . . . . . . . | $$ | . . . . . . O X X . . . . . . . . . . | $$ | . . . X . . . O X . . . . . . . . . . | $$ | . . X , O . . O X , . . . . . , . . . | $$ | . . . X O . . . X . . O . . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Enjoy!

pwaldron · #2

Bill Spight wrote:

Click Here To Show Diagram Code: [go]$$Wcm28 White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . b . . . . . . . . . | $$ | . . X X . . O O . . . . . . . . . . . | $$ | . . . . . . O X X . . . . . . . . . . | $$ | . . . X . . . O X . . . . . . . . . . | $$ | . . X , O . . O X , . . . . . , . . . | $$ | . . . X O . a . X . . O . . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

My eyes are drawn to the bottom left area of the board. Black has a string of stones without much of a base, and White's base is iffy as well. To me the choice is between 'a' and 'b', and while the gospel of Leela doesn't say much, I think the gospel of Fairbairn's barmkins would advocate for 'a'. Not only does it strengthen Black and directly weaken White's adjacent group, it will also support activities against White's lower right and possibly provide a useful ladder breaker for the top right.

'b' looks like the vital point for both group's bases, but I'm a little concerned white might just play 'a' himself.

EDIT: Realized after writing the above that it's White to play. 'a' is still my choice for White, though, for all the previously stated reasons.

Joaz Banbeck · #3

The two upper corner stones each have multiple ways to stay alive. So They are not critical.
I can give up the two X-ed stones. Especially when I get sente for it.
But the right side has a huge unclaimed area.

Click Here To Show Diagram Code: [go]$$Wcm28 White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . ? ? ? ? | $$ | . . O . . . . . . . . . . . . ? ? ? ? | $$ | . . . , . . . . . , . . . . . ? ? ? ? | $$ | . . . . . . . . . . . . . . . ? ? ? ? | $$ | . . O . . . . . . . . . . . . ? ? ? ? | $$ | . . X X . . O O . . . . . . . ? ? ? ? | $$ | . . . . . . O X X . . . . . . ? ? ? ? | $$ | . . . X . . . P X . . . . . . . ? ? ? | $$ | . . X , O . . P X , . . . . . , . ? ? | $$ | . . . X O . . . X . . O . . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ --------------------------------------- $$ {AR D17 B18} $$ {AR D17 F15} $$ {AR D17 F17} $$ {AR P17 R18} $$ {AR P17 P15}[/go]

So my choice is this:

Click Here To Show Diagram Code: [go]$$Wcm28 White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . W . . | $$ | . . X X . . O O . . . . . . . . . . . | $$ | . . . . . . O X X . . . . . . . . . . | $$ | . . . X . . . O X . . . . . . . . . . | $$ | . . X , O . . O X , . . . . . , a . . | $$ | . . . X O . . . X . . O . . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

It is almost in the center of the unclaimed area, and just close enough to the lower right corner to make a black placement in the corner very uncomfortable. If black does not dispute the corner, a white play at 'a' secures it on a large scale.

Knotwilg · #4

Very interesting position! I changed my mind a few times while analyzing it:

As usual these exercises are about deciding to make a "big play" (= apply Uberdude's AI opening gospel) or an "urgent play" (= an existing group needs to be defended or can be attacked for profit). I'm also going to inject some barmkin esoterics in the analysis

... to an unexpected effect!

So my eyes were drawn to Black's structure on the lower side. Is it rather thick or heavy? And are White's stones to the left of it important to keep the Black stones under pressure, or can part of it be considered as disposable?

Click Here To Show Diagram Code: [go]$$W White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . X X . . O O . . . . . . . . . . . | $$ | . . . . . . O X X . a . . . . . . . . | $$ | . . . X . . . O X . . . . . . . . . . | $$ | . . X , O . . O X , . . . . . , . . . | $$ | . . . X O . . b X . . O . . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

In terms of moves: is A a meaningful attack on the black stones? Or should White even strengthen at B first, to make a "barmkin"?

Click Here To Show Diagram Code: [go]$$W White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . X X . . O O . . . . . . . . . . . | $$ | . . . . . 6 O X X . 1 . . . . . . . . | $$ | . . . X 5 4 2 O X . . . . . . . . . . | $$ | . . X , O . 3 O X , . . . . . , . . . | $$ | . . . X O . . . X . . O . . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Then I noticed that :w1:

can be answered by a cut at :b2:

which White can not resist, like in this diagram, so White will have to give up the two stones.

Click Here To Show Diagram Code: [go]$$W White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . X X . . O O . . . . . . . . . . . | $$ | . . . . . . O X X . . . . . . . . . . | $$ | . . . X . . . O X . . 2 . . . . . . . | $$ | . . X , O . . O X , . . . . . , . . . | $$ | . . . X O . . 1 X . . O . . O . . . . | $$ | . . . . . . . a . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

While

may create a "barmkin", :b2:

makes for an easy way out into the most open area and Black can still hane at the head of three later.

So the conclusion is that Black is rather thick than heavy and White's two stones are rather disposable than important.

Can we then be helped by John's other aspect of "equilibrium", the count of stones in each area? Since we don't know which areas John has identified, I'll think about the usual quadrants. UL: even, LL: -2, UR: -1; LR: +2. From that it follows that White surely shouldn't add another stone in the lower right, nor in the upper left, maybe in the upper right but probably she still needs to play in the lower left. But where? Let me apply my amateur intuition, like John does.

Click Here To Show Diagram Code: [go]$$W White to play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . X . . X . . O . . . . | $$ | . . . , . . . . . , . . . . . , X . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O 1 . . . . . . . . . . . . . . . | $$ | . . X X . . O O . . . . . . . . . . . | $$ | . . . . . . O X X . . . . . . . . . . | $$ | . . . X . . . O X . . . . . . . . . . | $$ | . . X , O . . O X , . . . . . , . . . | $$ | . . . X O . . . X . . O . . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Then I saw this move: it helps White's group. It helps against aforementioned cutting sequence. If Black hanes, then White can cut. If Black extends, his stones become inefficient, White takes "control of the entire board" and her bottom stones become truly disposable.

Could it be ...?

Edit: alas, it was not to be, or I applied the Go theory of the 22nd century in the wrong way. Someone else got it right though and the theory might hold!

John Fairbairn · #5

Quote:

Edit: alas, it was not to be, or I applied the Go theory of the 22nd century in the wrong way. Someone else got it right though and the theory might hold!

Simplifying quite a bit, my program tells be to look for a move in the lower half formed by the A1/T19 diagonal, and with preference in the lower SE quadrant.

The move I would then choose is not the one Lizzie recommends and is about 3% ~ 4% worse.

But if I had not been prompted by the program, I would have chosen a move on the left side, and my choice, according to Lizzie, would have been about 7% worse. In other words, me + program appears to be stronger than me + me (the strong trend I have already reported).**

I remain oblivious of why the recommended areas are better, and it is in these cases, and only these cases, that I try out my X + barmkin explanations, looking also at the final position. X + B indeed stands up here, but what is amusing is that the winning move (move 71) is Lizzie's choice in this position, and Dosaku resigned as soon as he saw it.

** Thinking about this further, it occurred to me that the same sort of trend would probably show if a human pro told me the area to look in, in which case my "improvement" may be nothing to do directly to do with AI behaviour. It may just be that the extra topographical help I am getting is simply eliminating some of my bad habits. In particular, I think one of the worst habits all amateurs have is staying in one area for too long. I am certainly guilty. I think some of this is due to an "eyes down, follow your nose" approach, combined perhaps with a touch of OCD - you want to finish off each area tidily. But in broader terms, it may also be that amateurs shy away from open virgin areas simply because they don't know what to do there. If that's on the right lines, maybe a quick fix for amateurs is just to suck and see and play in those virgin areas come what may. I've noticed in quite a few cases in Bill series that this is already going on to some extent: posters say something like "I know I would play in area B but I know Bill is telling us the text book is flawed, so where else could I possibly play." Head scratching follows as poster tries to imagine what a bot would do. That's not much different from what I am doing except that my program seems to provide some refinement as regards the unexpected area to play in.

Bill Spight · #6

Bill Spight · #7

John Fairbairn wrote:

I've noticed in quite a few cases in Bill series that this is already going on to some extent: posters say something like "I know I would play in area B but I know Bill is telling us the text book is flawed, so where else could I possibly play."

Some people may have noticed that, particularly as I go back to a time before those textbooks were written, the AI move is often the textbook play. As in this example.

Quote:

Head scratching follows as poster tries to imagine what a bot would do.

An excellent question, IMO.

dhu163 · #8

Speculatively:
Current position:
1. H3 is the key point on the lower side especially as B is so thick in the lower left. However, there isn't much attack on B as the centre is open and the lower right is low.
2. Otherwise, the upper left is hottest as W is too weak there. It seems natural to play kick and jump for eyespace. C16 D15 F17. For now, the left is important for the fight on the lower side, but locally low temperature as it has some side eyespace next to B thickness and can attack C15. So this has some flow. E16 is the more solid alternative to take care of the central fight better, but it seems to not make much territory.
3. W R4 is always big, but doesn't seem that appropriate when W may get life from attacking M3.
4. Upper right Q15 is playable but heavy as B may get strong in the centre from the lower right, so W may remain weak. May be better to turn to the right instead.
5. Overall, I have to choose H3. If B gets G3, then F5 and F2 are next. Then R4 becomes a threat. Once W has some eyespace (from E2 later and attacking B) for the heaviest fight, the fights on the rest of the board become lighter for W as W is flexible, able to choose the centre instead of relying on side eyespace, or small side left without needing a connection. Next, I anticipate K8 (though the shape with W L6 is uncomfortable, W is making use of having a stronger centre of 2 stones there after B played the wedge). Then R4 seems biggest. W needs to be alive to profit from the attack. The upper left seems ok to tenuki as locally not big when neither side can kill or even really attack.

Opp to play
1. For the reasons above, G3.
2. In the upper left E16 is natural, but J17 isn't in the best position.

Past:
1. Most already mentioned, so I'm not convinced by W's position, but I'm also not sure about J17, though it does make sense. W is too thin. Perhaps this is W's attempt at "shinogi", but AI tends to say that although shinogi may be good against a weaker opponent, in an even game, W should play more solidly for influence with less groups rather than thinly. The problem is that B has 2 corners and W doesn't even have 1 clear corner.

Check:
1. Not quite. In this case, this AI is saying that W isn't yet alive after H3 and it may help B make shape since it is tight on B's strong shape on both sides, and in Bill's variation, W F2 is good endgame and still threatens H3. Regardless, Bill's variation is scary to choose, even if I am already convinced it is better. Because you need to play carefully on the left.
2. Very interesting variation. We see that :b33:

isn't exactly good shape so W avoids spending a bad shape move at H3. Of course the cost is being attacked on the left with another cut or shape point around H3. However, all W is going for in this variation is shinogi. Living small with each group is ok. It isn't as if B can surround in the centre.

Opening problems for AI: Problem 36

Who is online