Life In 19x19

(Sorry for the thread hijacking)

Pincers (are) were overrated in Go theory.

My view since the advent of neural nets is:
You do not need to know why a move is good, you just need to know the good move.
The request to explain a move is therefore superfluous.
Just know your moves, if you want to play strong go

John Fairbairn wrote: I'm also very dubious about White D6. Thickness is not thickness unless it functions as thickness. That means attacking. This D6 is a pure gote defence move and smacks of overconcentration and of thinking only of territory. You make territory in the region of thickness by surrounding it dynamically, i.e. by attacking and keeping sente. A positive attitude such as F5 or a pincer seems called for here.

I believe playing on the left side (around D6) is good in this case for white.
White's thickness is too far away to use directly, so he needs to build a moyo first, lure black in, so he can attack and use the thickness.

If white plays pincer on the lower side instead, when black approaches on the left side (around C6) he will settle in there easily (discarding the stone on the lower side) and white's thickness above goes to waste.
Alternatively, black can reply to white's pincer by simply entering 3x3 and again settles quickly if white follows the common-sense way of blocking at D3, etc.

John Fairbairn wrote:First the tewari as done here seems dubious to me. Even Bill's. I'm at a natural disadvantage here as the position needs more fingers than I've got, but in tewari (ii) I don't see how you can justify ending up with a position where locally both sides have the same number of stones but Black has one more stone on the board as a whole, whereas in the game they have played an equal number of stones overall.

Yes, there are questions with my tewari. I have actually thought a bit more about it. I was trying an experiment of averaging tewari after each side played a small gote. I think that that makes sense, but it adds difficulty. As for tewari (ii), Black has made an extra move that gains only 3.5 pts., so it is not surprising that the position is inefficient for Black during the opening. How do you average tewari (i) and tewari (ii)? Not easy.

Also, I am doing tewari at a different position than some are evaluating.

Click Here To Show Diagram Code

[go]$$Wc Position to tewari
$$ ---------------------------------------
$$ | . O . X . . . . . . . . . . . . . . . |
$$ | X O O O B X X X . . . . . . . . . . . |
$$ | O O X O O X O . . . . . . . . . . . . |
$$ | X X X X X O C O . , . . . . . X . . . |
$$ | . W . X O O O . . . . . . . . . . . . |
$$ | . X X O X . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . . . . . |
$$ | . O . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . W . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . B . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

Not everybody else is including the marked stones. And we should remember that Black has played a stone at .

Click Here To Show Diagram Code

[go]$$Wc White sente
$$ ---------------------------------------
$$ | . O 4 X 2 . . . . . . . . . . . . . . |
$$ | X O O O X X X X . . . . . . . . . . . |
$$ | O O X O O X O . . . . . . . . . . . . |
$$ | X X X X X O . O . , . . . . . X . . . |
$$ | . O . X O O O . . . . . . . . . . . . |
$$ | 1 X X O X . . . . . . . . . . . . . . |
$$ | . 3 . O . . . . . . . . . . . . . . . |
$$ | . O . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

We humans can simplify the position to evaluate by playing White's sente. The top left side is still unsettled, in the sense that there is a small gote left. However, we can simplify further.

Click Here To Show Diagram Code

[go]$$Wc Position for tewari
$$ ---------------------------------------
$$ | B O X X X . . . . . . . . . . . . . . |
$$ | X O O O X X X X . . . . . . . . . . . |
$$ | O O X O O X O . . . . . . . . . . . . |
$$ | X X X X X O C O . , . . . . . X . . . |
$$ | . O . X O O O . . . . . . . . . . . . |
$$ | O X X O X . . . . . . . . . . . . . . |
$$ | . O W O . . . . . . . . . . . . . . . |
$$ | . O . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

In the yose, regardless of who plays first in the top left, Black will play and capture the White stones and White will play . So this is the simplified position to evaluate.

In this position there is still a 3.5 pt. gote in the corner, but it is contained. Locally it is worth 1.5 pts. for White. Black to play can make 2 pts.; White to play can make 5 pts.

Click Here To Show Diagram Code

[go]$$Wc Tewari
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . X X X . . . . . . . . . . . |
$$ | . . . . . X O . . . . . . . . . . . . |
$$ | X X X X X O . O . , . . . . . X . . . |
$$ | . O . X O O O . . . . . . . . . . . . |
$$ | O X X O B . . . . . . . . . . . . . . |
$$ | . O O O . . . . . . . . . . . . . . . |
$$ | . W . @ . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

Black has captured 8 White stones in the corner and White has captured 1 Black stone. We balance the captures by removing 7 Black stones from the corner. I think that this tewari clearly favors Black in the top left corner. The stone has aji, but the stone is plainly badly placed. Those two stones are not equivalent, but if we treat them as equivalent and remove them, the stone is badly placed. Definitely good for Black.

Gomoto wrote:(Sorry for the thread hijacking)

I don't think it is hijacking.

Pincers (are) were overrated in Go theory.

Apparently a lesson from AlphaGo Master. However, AlphaGo Zero pincers more often than Master. Probably less than humans do now, though. My impression is that Zen pincers even more often. Zen obviously favors the one space reply over the other moves, but only the one space pincer is outside the margin of error, IMO.

My view since the advent of neural nets is:
You do not need to know why a move is good, you just need to know the good move.
The request to explain a move is therefore superfluous.
Just know your moves, if you want to play strong go

The idiot savant theory of go.

John Fairbairn wrote:Next, applying the pro rule of thickness being worth (n * n+1)/2 with n being the length of the wall (but not usually counting stones on the edges - the rule of each stone in a wall being worth 3 points is an approximation of this for the commonest lengths of wall).

Prone to overestimation, IMO — and I like thickness! I think that a 3.5 pt. approximation is better, as a rule.

On that basis I assume a wall of 8 gives White about 36 points, with an averaged few extra points for the late boundary plays on the upper left side. Black has just over 20. Since Black as an extra stone elsewhere on the board which can be counted as 15 points, White has significant advantage.

Huh? 20 + 15 = 35, so White by that reckoning has a slight advantage.

FWIW, my influence function pretty much agrees. It thinks that White is around 7 pts. ahead. It is biased towards the outside, so if we take that into account, White has a slight advantage.

I'm also very dubious about White D6. Thickness is not thickness unless it functions as thickness. That means attacking. This D6 is a pure gote defence move and smacks of overconcentration and of thinking only of territory. You make territory in the region of thickness by surrounding it dynamically, i.e. by attacking and keeping sente. A positive attitude such as F5 or a pincer seems called for here.

I also like F-05. But D-06 makes a moyo, not territory. Strong go players of yore made similar moyo. One advantage of D-06 is that it reduces uncertainty. Since Zen thinks that White is ahead, that is a plus.

Edit: OC, Zen is assuming a 7.5 komi to get that advantage for White.

I am following your post with incredible pleasure... pretty much interesting tewari analysis and thickness approximations...

My view since the advent of neural nets is:
You do not need to know why a move is good, you just need to know the good move.
The request to explain a move is therefore superfluous.
Just know your moves, if you want to play strong go

The idiot savant theory of go.

Indeed, neural networks show perhaps the secret to the power of some savants.

But I really think there is a lesson to be learned for the average (Joe) Go aficionado .

1. Rule based knowledge is to be treated with a certain amount of mistrust.
2. Training routine should be heavily based on intuition building activity (aka: spam games and tsumego, rehearsal of pro games, ... )


Edit: Something I found on Wikipedia

In savants, says Snyder, the top layer of mental processing —conceptual thinking, making logical deductions— is somehow deactivated. His working hypothesis is that once this layer is inactivate, one can access a startling capacity for recalling the most minute detail or for performing lightning-quick calculations. Snyder's theory has a conclusion of its own: He believes it may be possible someday to create technologies that will allow any non-autistic person to access these abilities.

A little bit of willpower could also be used instead of the technologies

Edit: OC, Zen is assuming a 7.5 komi to get that advantage for White.

(Not in my case, I used a 6.5 komi setting with Zen 7.)

Gomoto wrote:(Sorry for the thread hijacking)

Pincers (are) were overrated in Go theory.

My view since the advent of neural nets is:
You do not need to know why a move is good, you just need to know the good move.
The request to explain a move is therefore superfluous.
Just know your moves, if you want to play strong go

I hope you're being sarcastic.

If you don't know why a move is good, how will you be able to figure out what move to play in a position that's similar to, but not quite the same as, other positions you've seen before? (Or if your opponent doesn't follow joseki as another example.)

Gomoto wrote:Something I found on Wikipedia

In savants, says Snyder, the top layer of mental processing —conceptual thinking, making logical deductions— is somehow deactivated. His working hypothesis is that once this layer is inactivate, one can access a startling capacity for recalling the most minute detail or for performing lightning-quick calculations. Snyder's theory has a conclusion of its own: He believes it may be possible someday to create technologies that will allow any non-autistic person to access these abilities.

A little bit of willpower could also be used instead of the technologies

Willpower? Liquor is quicker.

Actually, I doubt that these abilities are particularly dormant in regular people. Most pros see the right move immediately. Conscious calculation is for checking. (Recall is a different question.)

Fedya wrote:

Gomoto wrote:(Sorry for the thread hijacking)

Pincers (are) were overrated in Go theory.

My view since the advent of neural nets is:
You do not need to know why a move is good, you just need to know the good move.
The request to explain a move is therefore superfluous.
Just know your moves, if you want to play strong go

I hope you're being sarcastic.

If you don't know why a move is good, how will you be able to figure out what move to play in a position that's similar to, but not quite the same as, other positions you've seen before? (Or if your opponent doesn't follow joseki as another example.)

Neural nets “know” why a move is good. It’s just a representation that is different from how humans think.

I prefer black, the top side has little value, black has still two 4-4 stones and the left side is easily reducible or at least can be made overconcentrated in some way

(im overconfident in this post)