This 'n' that

Kirby · Post by **Kirby** » Thu Jun 22, 2017 7:47 pm

Bill Spight wrote:Another possible advantage is what moha points out, that the MC probabilities are based upon the actual position at hand, not some aggregate of more or less similar positions.

The value network component of a leaf node evaluation is also based on the actual position at hand, isn't it? It's true that the value network was trained and refined by self-play, thereby adjusting itself over time from similar game experiences, but the weighted function that's produced gives a unique value for that particular position, as I understand it.

I suppose MC gives the advantage of actually continuing to play out until the end of the game, whereas the value network produces a value based on past experience, which is agnostic to any hypothetical playouts...

Bill Spight · Post by **Bill Spight** » Thu Jun 22, 2017 8:54 pm

Kirby wrote:
Bill Spight wrote:Another possible advantage is what moha points out, that the MC probabilities are based upon the actual position at hand, not some aggregate of more or less similar positions.

The value network component of a leaf node evaluation is also based on the actual position at hand, isn't it? It's true that the value network was trained and refined by self-play, thereby adjusting itself over time from similar game experiences, but the weighted function that's produced gives a unique value for that particular position, as I understand it.

I suppose MC gives the advantage of actually continuing to play out until the end of the game, whereas the value network produces a value based on past experience, which is agnostic to any hypothetical playouts...

Good point.

What I had in mind was that the MC estimate is based solely upon the given position, where the value net estimate is based upon the given position plus (implicitly) many similar positions. The MC play outs could lead to the construction of a game tree that finds the exception in the current position, where it is significantly different from the other positions, which is what moha seems to be talking about, or least provide some correction to a mistaken value net estimate. OC, it may lead to a worse estimate, which I suspect will usually be the case. But the players are not laying bets on the game, (

), they are trying to win, so a worse estimate may have no effect on the outcome.

moha · Post by **moha** » Fri Jun 23, 2017 2:01 am

Bill Spight wrote:What I had in mind was that the MC estimate is based solely upon the given position, where the value net estimate is based upon the given position plus (implicitly) many similar positions. The MC play outs could lead to the construction of a game tree that finds the exception in the current position, where it is significantly different from the other positions, which is what moha seems to be talking about, or least provide some correction to a mistaken value net estimate. OC, it may lead to a worse estimate, which I suspect will usually be the case.

Their paper mentioned that a version with MC rollouts only (for evaluation, with otherwise similar search I suppose) is significantly stronger than a version with value net only. You guys expect way too much from NNs (esp wholeboard-wise).

Although this was only when using the best policy net - presumably since MC depends on quality moves as I mentioned.

Bill Spight · Post by **Bill Spight** » Fri Jun 23, 2017 7:45 am

moha wrote:
Bill Spight wrote:What I had in mind was that the MC estimate is based solely upon the given position, where the value net estimate is based upon the given position plus (implicitly) many similar positions. The MC play outs could lead to the construction of a game tree that finds the exception in the current position, where it is significantly different from the other positions, which is what moha seems to be talking about, or least provide some correction to a mistaken value net estimate. OC, it may lead to a worse estimate, which I suspect will usually be the case.

Their paper mentioned that a version with MC rollouts only (for evaluation, with otherwise similar search I suppose) is significantly stronger than a version with value net only.

Interesting. Thanks.

You guys expect way too much from NNs (esp wholeboard-wise).

Although this was only when using the best policy net - presumably since MC depends on quality moves as I mentioned.

Well, back in the 90s there were go playing programs that used Monte Carlo rollouts and programs that used neural networks. Neither approach was successful. The programs were very weak. I played around a bit with Monte Carlo rollouts myself, and I know that Monte Carlo evaluation sucks. But there was a breakthrough early in this century with Monte Carlo Tree Search, which uses Monte Carlo methods for both evaluation and exploration (building the search tree). As you might expect, it was guiding the search that was the breakthrough. The evaluation still sucked. (As I said, I think I could get rich betting against Monte Carlo evaluation.

) But for winning the game what counts is relative evaluation. If the relative evaluation orders plays well, that is what matters for winning the game, even if you could lose your shirt placing bets.

Now there is a new breakthrough, Deep Learning neural networks. AlphaGo's policy network alone, with no search tree, plays as well as the first MCTS programs. Hell, yes, that's impressive! In a short time AlphaGo reached pro level, then top pro level, and now has left pros in the dust, reaching a level surpassing even the go geniuses of all time. Furthermore, its rate of improvement has shown no signs of slowing down, which indicates that programs of the future have a lot of room to get even stronger. Yes, it was search that brought very weak Monte Carlo methods to the fore, but it is deep learning that has made an even more impressive improvement possible. We still do not know how much of an advance is possible via deep learning neural networks.

EdLee · Post by **EdLee** » Fri Jun 23, 2017 9:51 am

We still do not know how much of an advance is possible via deep learning neural networks.

Interesting to guesstimate how far below perfect play is the top human level: 5 stones ? Maybe that's too much; 4 ?

Bill Spight · Post by **Bill Spight** » Fri Jun 23, 2017 10:13 am

EdLee wrote:
We still do not know how much of an advance is possible via deep learning neural networks.
Interesting to guesstimate how far below perfect play is the top human level: 5 stones ? Maybe that's too much; 4 ?

Before AlphaGo I would have guessed 4 stones. They say that AlphaGo can give handicaps, despite not having trained for those conditions. Maybe AlphaGo can give today's 9 dans 3 stones? And remember that a 3 stone handicap already takes away much of the advantage White has in the opening, which is a good bit of the difference between AlphaGo and humans.

Kirby · Post by **Kirby** » Fri Jun 23, 2017 10:22 am

EdLee wrote:
We still do not know how much of an advance is possible via deep learning neural networks.
Interesting to guesstimate how far below perfect play is the top human level: 5 stones ? Maybe that's too much; 4 ?

This brings about an interesting question regarding what we mean by "perfect play". I presume that the intended definition was a set of moves that are optimal in the sense that they are guaranteed to produce the best result, no matter how the opponent responds. However, this way of playing may not bring about the best results against humans. In the same way that some high-dan amateurs can sometimes give weaker players more stones than professionals, there may exist strategies for computer programs to play in such a confusing (but non-optimal) way that they are able to beat top pros with a greater number of stones.

For example, against a computer program playing "properly", only playing moves that cannot possibly be refuted, let's say that a human can win with 4 stones. A different computer program that plays in a confusing way, or perhaps a way that is not aligned with that particular pro's style, may be able to achieve wins with a greater handicap.

Answering the question of how many stones a top pro would need to win against a computer playing properly, with moves that are guaranteed to be optimal, is a difficult question to answer. But I think it's even more challenging to determine the maximum number of stones that a computer program could give a top pro and win, given an arbitrary strategy that may be particularly difficult for the human.

EdLee · Post by **EdLee** » Fri Jun 23, 2017 10:38 am

Hi Bill, even if it's 5 stones, it seems very impressive to me for this (stone age?) brain.

Kirby, interesting point !

Bill Spight · Post by **Bill Spight** » Fri Jun 23, 2017 11:22 am

Kirby wrote:
EdLee wrote:
We still do not know how much of an advance is possible via deep learning neural networks.
Interesting to guesstimate how far below perfect play is the top human level: 5 stones ? Maybe that's too much; 4 ?

This brings about an interesting question regarding what we mean by "perfect play". I presume that the intended definition was a set of moves that are optimal in the sense that they are guaranteed to produce the best result, no matter how the opponent responds.

That's not the definition used by AlphaGo or MCTS programs. Rather, they do not aim at perfect play, and have no definition, explicit or implied, for it.

However, this way of playing may not bring about the best results against humans. In the same way that some high-dan amateurs can sometimes give weaker players more stones than professionals, there may exist strategies for computer programs to play in such a confusing (but non-optimal) way that they are able to beat top pros with a greater number of stones.

{snip}

Answering the question of how many stones a top pro would need to win against a computer playing properly, with moves that are guaranteed to be optimal, is a difficult question to answer. But I think it's even more challenging to determine the maximum number of stones that a computer program could give a top pro and win, given an arbitrary strategy that may be particularly difficult for the human.

Well, if a program perceives itself to be far behind, as in giving 5 stones, will it make desperation moves that the human can take advantage of? IIUC, some programs today use variable komi to adjust the final score of play outs in the computer's favor, reducing the komi as the game progresses, to avoid that kind of thing.

Bill Spight · Post by **Bill Spight** » Sun Jun 25, 2017 8:57 am

Top attachment

Click Here To Show Diagram Code: [go]$$Wcm16 Top attachment $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O O . . . . | $$ | . . . . . . . . . . . . X X X O O . . | $$ | . . . O . . . . . , . . . . . X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . 2 1 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , X . . | $$ | . . . . . X . . . . . X . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

We teach beginners not to attach to stones they are attacking, but once you see this top attachment, it looks very nice, doesn't it?

Top attachments against pincers and pincered stones are not unusual, but I think that the top attachment against a wedge is another AlphaGo innovation. It avoids the problems of a third line approach from either side. It also indicates the value of the center. The center is important, but not so important as to play a boshi.

Click Here To Show Diagram Code: [go]$$Wcm16 Stretch $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . O O . . . . | $$ | . . . . . . . . . . . . X X X O O . . | $$ | . . . O . . . . . , . . . . . X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . b 3 . . | $$ | . . . . . . . . . . . . . . . 2 1 . . | $$ | . . . . . . . . . . . . . . . a c . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , X . . | $$ | . . . . . X . . . . . X . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Neither hane-counterhane sequence looks good. Wa - :b18:

lets the Black wall work, either in a fight or with the :b17:

and

stones, and White does not have much room to develop on the bottom right. If Wb - Bc instead, Black gets good development of his moyo. :w18:

is solid, taking away Black's preferred hane.

Click Here To Show Diagram Code: [go]$$Wcm16 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . 5 O O . . . . | $$ | . . . . . . . . . . . 6 X X X O O . . | $$ | . . . O . . . . . , . . . . . X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , 3 . . | $$ | . . . . . . . . . . . . . . b 2 1 . . | $$ | . . . . . . . . . . . . . . . . a . . | $$ | . . . . . . . . . . . . . . . 4 . 7 . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , X . . | $$ | . . . . . X . . . . . X . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

is very nice, and shows the utility of :b17:

. If Black had played :b19:

first, White could have jumped to "b". :b17:

is perfectly placed.

relieves the pressure on the White corner with sente, and then White slides to :w22:

.

is not an AlphaGo style play, a human could easily have played it. It prevents a Black hane at "a", which is now senseless. It also, OC, gives the White stones some eye potential. But more important, I think, is that it offers potential for further White play in the bottom right. By contrast, an extension towards the Black wall does not have much potential development precisely because of the wall.

Large scale development

Click Here To Show Diagram Code: [go]$$Bcm23 One space jump $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . O O O . . . . | $$ | . . . . . . . . . . . X X X X O O . . | $$ | . . . O . . . . . , . . . . . X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . X . O . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . 1 . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , X . . | $$ | . . . . . X . . . . . X . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

is one of the oldest plays in go — against the large knight's enclosure. Against the small knight's enclosure it does not threaten the corner as much. But it is very much in the AlphaGo - Go Seigen style of rapid but thin development. Surely White has to play inside Black's huge sphere of influence, but where?

Click Here To Show Diagram Code: [go]$$Bcm23 Deep invasion $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . O O O . . . . | $$ | . . . . . . . . . . . X X X X O O . . | $$ | . . . O . . . . . , . . . . . X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 4 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . X . O . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . X . . . . . a . . . 3 . . . | $$ | . . . O . . . . . , . . . . . , X . . | $$ | . . . . . X . . . . . X . . 2 . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

invades the bottom right. After :b25:

,

settles the White group on the right side. Note that White extends only one space to forestall any attack. Now :w24:

looks like a gift. And indeed, it dies later in the game, along with other stones White plays in the vicinity. But it has aji. It exemplifies the AlphaGo - Kitani style of salting the opponent's position with stones which often die, but which have aji.

Where does Black play now? "a" would surely kill :w24:

, but it is a single purpose play, and kind of small looking at that. Black has better plays.

Click Here To Show Diagram Code: [go]$$Bcm23 Prelude to a fight $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . O O O . . . . | $$ | . . . . . . . . . . . X X X X O O . . | $$ | . . . O . . . . . , . . . . . X X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . X . O . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . B . . . . . . . . . X . . . | $$ | . . . O . . . . . , . . . . . , X . . | $$ | . . 5 . . X . . . . . X . . W . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

The invasion, :b27:

, surprised me. Surely in exchange for the corner White will go into Black's bottom left position, after which Black will have the difficult task of keeping White from utilizing the aji of :wc:

while at the same time making use of :bc:

. And indeed that is how the game developed. The fight is quite interesting, but any commentary is above my pay grade.

I'll leave that to the pros.

Bill Spight · Post by **Bill Spight** » Mon Jun 26, 2017 9:51 am

AlphaGo vs. AlphaGo Game 46

Sanrensei

Click Here To Show Diagram Code: [go]$$Wcm6 Sanrensei $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 2 3 . . . . . . . . . . . . . . . | $$ | . . 4 O . . . . . a . . . . . X . . . | $$ | . . 6 7 . . . . . . . . . . . . . . . | $$ | . 8 5 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . 1 . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Black played sanrensei, which surprised me a little, since it has gotten a bad rap in recent years. At first glance, it doesn't seem bad in theory, but its stats have not been so great. One problem is how to respond to the approach, :w6:

? Local replies have not fared too well. Maybe yonrensei at "a" is good. AlphaGo offers an answer, the AlphaGo 3-3 invasion in the top left corner.

Which side to block on? I think that we can consider some of these invasions as probes. White chooses to block on the side facing the sanrensei instead of the side facing the White corner stone. :w10:

is perhaps the usual choice, instead of playing hane at the head of two stones. It is not an AlphaGo innovation, but the subsequent play is. Humans have been playing the keima to take sente. White does take sente, but after a new sequence. :b7:

-

is a new, AlphaGo joseki.

White gets a wall, but one without either eye shape or an extension. No matter, says AlphaGo. :cool:

Click Here To Show Diagram Code: [go]$$Wcm14 3-3 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . . X O . . . . . , . . . . . X . . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . X O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . 8 9 . a . . . | $$ | . . . . . . . . . . . . 7 6 0 . . . . | $$ | . . . O . . . . . , . . . 5 4 X 2 . . | $$ | . . . . . . . . . . . . . O . 3 1 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

jumps into the corner. I don't think that we can conclude that AlphaGo's choices are the only good plays, or even the best. The double approach at "a" has good stats, and I doubt if humans will give up on it or other plays. The play through :b23:

is joseki.

Click Here To Show Diagram Code: [go]$$Wcm24 The fight begins $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . . X O . . . . . , . . . . . X . . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . X O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . 7 . . 6 . . . . . | $$ | . . . . . . . . . . 3 2 X W . . . . . | $$ | . . . . . . . . . 8 4 1 O X X . . . . | $$ | . . . O . . . . . , . . . O X X X . . | $$ | . . . . . . . . . . . . . O . O O . . | $$ | . . . . . . . . . . . . 5 . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

White might consider playing :w26:

at 27. How bad can it be to crawl along the 5th line?

Besides,

is a good play, necessitating :w28:

. But

takes advantage of the aji of :wc:

.

and

are good.

Click Here To Show Diagram Code: [go]$$Wcm32 Kikashi, Eye stealing tesuji $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . . X O . . . . . , . . . . . X . . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . X O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . 5 . . O 3 1 X . . . . . | $$ | . . . . . . . . . . O X X W 2 . . . . | $$ | . . . . . . . . . B B O O X X . . . . | $$ | . . . O . . . . . , . . a O X X X . . | $$ | . . . . . . . . . . . 6 . O . O O . . | $$ | . . . . . . . . . . . . O . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

fills at

White takes his kikashi and then :w36:

bolsters White's central stones while attacking the :bc:

stones and building central influence.

:b37:

is the eye stealing tesuji, threatening to cut at "a", which captures two white stones and saves the two endangered :bc:

stones.

Click Here To Show Diagram Code: [go]$$Wcm38 Be my guest! $$ --------------------------------------- $$ | . . 8 . . . . . . . . . . . . . . . . | $$ | . 4 1 6 7 . . . . . . . . . . . . . . | $$ | . a X O . . . . . . . . . . . . . . . | $$ | . . X O . . . . . , . . . . . X . . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . X O 3 . . . . . . . . . . . . . . . | $$ | . 5 . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . O . . O O O X . . . . . | $$ | . . . . . . . . . . O X X X X . . . . | $$ | . . . . . . . . . B B W W X X . . . . | $$ | . . . O . . . . . , . . 2 O X X X . . | $$ | . . . . . . . . . . . B . O . O O . . | $$ | . . . . . . . . . . . . O . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

is surprising. Who doesn't connect to a peep? Especially as connecting leaves the :bc:

stones weak and attackable. Maybe this was a mistake. (Black did win the game, after all.) OTOH, connecting gives White bad shape, and the :wc:

stones have done work, permitting the White kikashi with :w32:

and

. So maybe they are expendable.

AlphaGo as White chose to sacrifice the :wc:

stones and play kikashi against the top left corner, in the process building influence in the center and left side. :w40:

is very nice in that regard, as opposed to, say, the hane at "a", which looks tempting.

prepares for :w42:

, which is also big.

Click Here To Show Diagram Code: [go]$$Wcm46 AlphaGo joseki $$ --------------------------------------- $$ | . . X . . . . . . . . . . . . . . . . | $$ | . X . X O . . . . . . . 9 7 . . . . . | $$ | . . X O . . . . . . . 0 8 4 5 3 1 . . | $$ | . . X O . . . . . , . . . . 6 X 2 . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . X O O . . . . . . . . . . . . . . . | $$ | . O . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . O . . O O O X . . . . . | $$ | . . . . . . . . . . O X X X X . . . . | $$ | . . . . . . . . . X X O O X X . . . . | $$ | . . . O . . . . . , . . X O X X X . . | $$ | . . . . . . . . . . . X . O . O O . . | $$ | . . . . . . . . . . . . O . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Next, White plays another AlphaGo special in the top right corner.

This time

is played with the idea of building thickness, not of taking sente. (Anthropomorphizing a bit. AlphaGo does not make plans, OC.

)

Click Here To Show Diagram Code: [go]$$Wcm56 Central moyo $$ --------------------------------------- $$ | . . X . . . . . . . . . . . . . . . . | $$ | . X . X O . . . . . . . O O . . . . . | $$ | . . X O . . . . . . . X X X O O O . . | $$ | . . X O . . . . . , . . . . X X X . . | $$ | . . X O . . . . . . . . . . . . . . . | $$ | . X O O . . . . . . . . 9 . . . . . . | $$ | . O . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . 1 . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . 3 . 7 . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 2 . 6 . . O . . O O O X . . . . . | $$ | . . . . . . . . . . O X X X X . . . . | $$ | . . . . . . . . . X X O O X X . . . . | $$ | . 4 . O . . . . . , . . X O X X X . . | $$ | . 8 5 . . . . . . . . X . O . O O . . | $$ | . . . . . . . . . . . . O . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

builds the White central moyo while preventing Black from extending her moyo on the right. A Black play at or around 56 would be good, eh?

enters the left side. White pincers it and builds up his huge moyo while (hopefully) holding the fort in the bottom left corner. :w64:

reduces Black's moyo while extending White's. White's center oriented play in this game is reminiscent of Go Seigen's.

Bill Spight · Post by **Bill Spight** » Tue Jun 27, 2017 11:46 am

AlphaGo vs. AlphaGo Game 48

Click Here To Show Diagram Code: [go]$$Bc Game 48 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . 4 . 0 . . . . . . . . . . . . . | $$ | . . 8 , . . . . . , . . . . . , 1 . . | $$ | . . 6 5 . . . . . . . . . . . . . . . | $$ | . . 7 9 . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . a d . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . b . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . 2 . . . . . , . . . . . , . . . | $$ | . . . . . c . . . . . . . . . 3 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Nothing unusual so far.

When I was learning go it was already known that there is a problem with this joseki ending with a Black play at "a" to make a base — at least, early in the opening. One way to avoid the problem is to omit :b9:

and play elsewhere, for instance, enclosing the top right corner. Another way is to follow with Bb - Wc, Bd, staking out the left side and extending on the fourth line instead of the third.

Click Here To Show Diagram Code: [go]$$Bcm11 What, me worry? $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . 3 . , X . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . a . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 1 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . , . . . | $$ | . . . . . 2 . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

AlphaGo shows another way. Play the exchange, :b11:

-

, and then play elsewhere. :shock:

In this case Black plays the two space high enclosure in the top right corner. AlphaGo likes this enclosure, and it will probably become more popular as a result.

Black allows White the good play at "a", but is unconcerned.

Click Here To Show Diagram Code: [go]$$Wcm14 No rush $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . X . , X . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . 3 . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 1 . . | $$ | . . . O . . . . . , . . . 2 . , . . . | $$ | . . . . . O . . . . . . a . . B . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

White does not hurry to make the good play at :w16:

, but approaches the bottom right corner, instead. Black could go back and play at 16, OC, but then White might pincer the :bc:

stone. If Black pincers :w14:

, White might play a counter pincer. :b15:

at "a" would allow White to press down at 15. I like :b15:

.

(Not that I would have left the left side wide open.

)

Click Here To Show Diagram Code: [go]$$Wcm14 Still not worried $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . X . , X . . | $$ | . . O B . . . . . . . . . . . . . . . | $$ | . . B B . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . 3 . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . 4 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 1 . . | $$ | . . . O . . . . . , . . . 2 . , . . . | $$ | . . . . . O . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Even after the double pincer at :w16:

, Black does not respond, but makes a pincer on the right side. Now, the single Black stone is not of much concern, as it is just one stone. But what about the — ahem — Black wall, such as it is, in the top left. Those three :bc:

stones have no eye shape and no base. How does Black keep those stones from becoming heavy baggage, if they aren't already?

Now, in my comments I have characterized some of AlphaGo's plays or stones as thin or heavy. I really should not do that, because those are go terms indicating incorrect play, and I cannot apply them to AlphaGo's play. Not that AlphaGo's play is perfect, OC, but I cannot show those errors. A better term for AlphaGo's play in these cases is light. Light and thin are cousins, so it is easy to see how a play that looks thin could actually be light. But light and heavy are opposites. How to keep eyeless stones from becoming heavy? It is important, I think, to treat them lightly, to be flexible and to be prepared to sacrifice them.

I think that omitting D-10 in this position is an AlphaGo innovation, one which I think pros will copy. But omitting the extension from the :bc:

stones is not new, going back at least a couple of centuries, although the extension is by far the usual play.

Click Here To Show Diagram Code: [go]$$Wcm18 Double attack $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . X . , X . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . 8 . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . 7 . . . , . . . . . , . . . | $$ | . . . 5 . . . . . . . . . . . . X . . | $$ | . . 3 4 6 . 9 . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X 2 . a 0 . . . . . . . . . . . . | $$ | . . 1 . . . . . . . . . . . . . O . . | $$ | . . . O . . . . . , . . . X . , . . . | $$ | . . . . . O . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

The kick at :w18:

launches a double attack. I like to play White in these situations, but AlphaGo as Black invited the attack. :b25:

jumps out from the upper group, perhaps because the lower one has more flexibility. :w26:

is strong, topping the tree. Note :b27:

. Black could have made a table shape with "a", but :b27:

is lighter. Black looks ready to sacrifice two of the stones in the lower left.

Click Here To Show Diagram Code: [go]$$Wcm28 Sideways attacks $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . 3 . . , . . . X . , X . . | $$ | . . O X . . 4 . . . . . . . . . . . . | $$ | . . X X . . X . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . 6 . , . . . . . , . . . | $$ | . . . O . . . . . . . . . . . . X . . | $$ | . . O X X . O . 5 . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X X . . X . 2 . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . O . . | $$ | . . . O . . 1 . . , . . . X . , . . . | $$ | . . . . . O . . . . . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Often in an attack you want to get ahead of your opponent. Pushing the cart from behind usually makes it easy for him to escape. One thing I noticed in Okigo Jizai is how often Hattori plays an attack from the side. Both :w28:

and

are sideways attacks. Because of the solid White corner, the one in the top left looks slow, but AlphaGo likes it. Both of them prevent shoulder blows, which would be big. :w32:

is another sideways play, but :b33:

launches a counter-attack.

White eventually won the game, but Black's light treatment of the three Black stones in the upper left is characteristic of AlphaGo's play. Such flexibility is not unique in the history of go, but, looking at AlphaGo's play against today's pros, I think that its skill in that aspect of the game is supreme. Much food for thought there.

Bill Spight · Post by **Bill Spight** » Thu Jun 29, 2017 1:18 am

AlphaGo vs. AlphaGo Game 4

Click Here To Show Diagram Code: [go]$$Bc Opposing 3-4s $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . 2 . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . 1 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . 4 . . . | $$ | . . . 3 . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

By symmetry we know that the net count of this position is 0. But the opposing 3-4s on the left side are supposed to favor the first player. As we would say now, the global temperature is greater than that of the empty board, so Black can overcome komi. The argument, which I learned as a beginner, goes like this.

Click Here To Show Diagram Code: [go]$$Bc Good move for Black $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . 2 . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . 1 . . . | $$ | . . 5 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . 7 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 6 . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . 4 . . . | $$ | . . . 3 . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

After

,

preserves symmetry, but then Black gets the good play, :b7:

, a combined extension and pincer. OC, White does not have to approach the bottom left corner with :w6:

, but then Black can enclose it.

Well, that's the theory, anyway.

But pros have still played opposing 3-4s from time to time. And White won this game. Not that :b7:

isn't a good move, but the theory remains unproven. Something else we won't be teaching beginners anymore.

Click Here To Show Diagram Code: [go]$$Bcm5 Will he or won't he? $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . 6 . . . . . . . . . . . . . | $$ | . . 4 , . . . . . , . . . . . X . . . | $$ | . . 2 1 . . . . . . . . . . . . . . . | $$ | . . 3 5 . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . a b . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . O . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Black played the high approach, and White made the popular response. Joseki so far. But will Black play the extension to "a", or maybe "b"? Or maybe Black will make no extension at all, and enclose the bottom left corner.

Click Here To Show Diagram Code: [go]Bcm11 Mini-Chinese? $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 1 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . O . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Of course not. :b11:

makes a kind of Mini-Chinese shape. This is, I think, another AlphaGo innovation, but it would not have appeared strange to the players of two centuries ago. Not only did they play the Mini-Chinese in handicap games, they often played extensions in the opening that are wider than the extensions to make a base. I expect that AlphaGo is reviving that style of play.

As I have mentioned before ( viewtopic.php?p=211817#p211817 and following posts) the idea of making a base in the opening is quite ancient. Long ago, players even preferred to play a wedge, leaving enough room to make a base on either side, in situations where modern players would most likely pincer. Also, extensions often leave room for a two space extension, to be able to make a base if the opponent invades. And six space extensions are considered questionable, as they leave room for an invader to make a two space extension. Takagawa, known for his skill in the opening, said that the main reason for playing in the corner was not because it is easier to make territory than elsewhere, although that is so, but because it is easier to make a base. Much of traditional opening theory derives from the importance of making a base. But, early in the game, AlphaGo does not seem to place much emphasis on making a base. It often makes longer extensions. This is not exactly new, as the play of a couple of centuries attests, but AlphaGo's play threatens the very foundations of traditional opening theory. Fan Hui said that AlphaGo stresses efficiency over making territory, but of even more theoretical significance, I think, is that it stresses efficiency over making a base.

Click Here To Show Diagram Code: [go]$$Wcm12 Go Seigen shoulder blow $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . a . . 2 3 . O . . . | $$ | . . . X . . . . . . . . . 1 . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

is the modern, quasi-enclosure in the bottom right corner. (White makes a base instead of extending to "a", what can I say?

) Some people would call :b13:

an AlphaGo shoulder blow, but Go Seigen advised this play, when Black has a stone in the bottom left corner, in his 21st Century Go books.

responds in the corner. Where does AlphaGo play next?

Click Here To Show Diagram Code: [go]$$Bcm15 Close approach $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . 1 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . a . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . X O . O b . . | $$ | . . . X . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Who would have guessed this play? Oh, the main point is easy to grasp. :b15:

spoils White's development on the right side. A White play at "a" would be over concentrated. But by the same token, a Black play at "a", while it makes a base, is smaller than the typical two space extension. Black would be cramped. And in the actual game Black did not play at "a" but make the attachment at "b", and did not make a base in the bottom right.

Click Here To Show Diagram Code: [go]$$Wcm16 Another Go Seigen play $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . 1 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . a . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . 2 . . . . . . | $$ | . . . , . . . . . , . . X O . O . . . | $$ | . . . X . 3 . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

approaches the top right corner and prevents Black's development here. Note that :w16:

has room to make a base at "a". :b17:

extends to the 5th line, a big play. :w18:

approaches the Mini-Chinese from the bottom side, another Go Seigen idea (although this approach to the 3-4 also goes back to antiquity).

Click Here To Show Diagram Code: [go]$$Bcm19 High pincer $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . X . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . 3 . . 2 . . . . . X . . . . . . | $$ | . . . , . . . 1 . , . . X O . O . . . | $$ | . . . X . O . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

is a high pincer. After :w20:

secures the corner. Where does White play now?

Click Here To Show Diagram Code: [go]$$Wcm22 Zwischenzug $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . X . . | $$ | . . . . 1 . . . . . . . . . . . . . . | $$ | . . . 2 . . . . . . . . . . . . . . . | $$ | . . . X . . O 3 . . . . X . . . . . . | $$ | . . . , . . . X . , . . X O . O . . . | $$ | . . . X . O . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Plainly, White wants to press with :w24:

. But first White plays kikashi, a reducing play at :w22:

. Note that after Black protects with :b23:

, White still threatens the left side.

Click Here To Show Diagram Code: [go]$$Wcm24 Choshi $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . X . . | $$ | . . . . W . . . 7 . . . . . . . . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . X . . O 1 3 6 . . X . . . . . . | $$ | . . . , . . . X 2 4 . . X O . O . . . | $$ | . . . X . O . . . . . . . O . . . . . | $$ | . . . . . . . . . . . 5 . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

-

are choshi, a term that has been variously translated. The general idea is to force the opponent to force you to make a play that you want to make. The keima, :w28:

, is a good play in itself, but after :b25:

-

, it is even better. Note how the :wc:

stone (

) helps to give White some eye potential in the center. Both sides have played a nice opening.

Now let's fast forward to the early middle game.

Click Here To Show Diagram Code: [go]$$Bcm63 Base? What base? $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . 3 . X . . . | $$ | . . O X . . . . . . . . . . . . 1 . . | $$ | . . X X . . . . . . . . 6 . . 2 W . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . O . . . . . . . . . . . . . | $$ | . . X . . O X X . . . . . . . 5 b . . | $$ | . . . , . O O X . , . . . . . , a . . | $$ | . . . . . . . O X . . . . . . . . . . | $$ | . . X . . . . O X . . . . . . . X . . | $$ | . . . . O . . X O X O . . . X f e 4 . | $$ | . . . X . . . . O O . . . O X O c d . | $$ | . . . X . . O O O X . . X O X X O O . | $$ | . . . , . X X X X X . . X O . O X X . | $$ | . . . X . O . . . . . . . O . O O . . | $$ | . . . . . . . . . . . O . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

The kick,

, initiates an attack against the :wc:

orphan. White could make a base with "a" or "b" instead of :w66:

. White prefers to allow the pincer, :b67:

, and then run out with :w68:

. Why White made that judgement, I cannot say. But note that :w66:

takes away Black's potential base here. Black could make a base with Bc - Wd, :b66:

- We, Bf, and White cannot connect.

AlphaGo's play in this game and others demonstrates that AlphaGo is less concerned with making a base than with efficiency. IMO that will have a profound effect upon go theory.

John Fairbairn · Post by **John Fairbairn** » Thu Jun 29, 2017 2:19 am

Bill

Interesting thoughts! First, just parenthetically, omitting D10 has been tried by humans before (Takagawa, Kobayashi, etc), but is admittedly rare.

Your most intriguing comments are about using "light" in a new sense. I likewise have come to the conclusion that we need to re-invent the vocabulary somewhat to start making sense of AlphaGo's plays.

The best approach we have both adopted seems to be to say it's just like Go Seigen (and it's fascinating and exciting that Yoda's forthcoming book on AG will focus on its similarities with Go and Dosaku). I think that means a lot to us and anyone else who has been through the corpus of Go's games, but it's very hard to be specific or to generalise.

Rather than the light/not-heavy focus you have been taking, I have been thinking on other lines which may well end up at the same place. My thoughts have focused on thickness, which I think has been the dominant theoretical theme in pro talk for close to a century now. I'm sensing that AG seems to eschew that and to focus rather on influence (from which thickness may develop, but that's by the by as far as AG is concerned).

However, AG seems to focus on influence in a quite different way from us. I think we see influence as tied in with thickness - they are on a continuum - and so I have an impression that most of us feel that subsequent play is somehow towards that influence/thickness (driving the enemy towards it), or staying out of its shadow (keeping away from it). Either way, it is the actual entity of influence/thickness that dominates the thought process.

AG, however, seems to see things more nebulously, in terms of an influence map. Rather than a wall and its shadow it sees a lake. Its focus is outwards. It seems to be able to see what effect outside stones will have elsewhere over a much wider area than humans do. What AG seems to be doing is what humans tried to analyse at the time of Shin Fuseki (e.g. the way they talked about the 5-5 point or tengen). The humans gave up because it was just computationally too hard, but at least they knew where to look!

My sense is that what I am saying does resolve into your thoughts on lightness. I know you've done some work on influence maps as well. Maybe we should revisit this? I feel sure even at this stage that we do need some new terminology, maybe based on influence maps. I'm attracted by the idea of basing it on water. Not just lakes, but fluidity, the notion of water being light and heavy, evaporating, freezing, mixing, diluting, etc. etc. Instead of food for thought, liquid for thought

(Mine's a double, neat, thank you.)

Uberdude · Post by **Uberdude** » Thu Jun 29, 2017 4:25 am

Bill Spight wrote:AlphaGo vs. AlphaGo Game 4

$$Bcm11 Mini-Chinese?
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . . . . . . |
$$ | . . O , . . . . . , . . . . . X . . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . 1 . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . O . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------

Click Here To Show Diagram Code
[go]$$Bcm11 Mini-Chinese? $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 1 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . O . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Of course not. makes a kind of Mini-Chinese shape. This is, I think, another AlphaGo innovation, but it would not have appeared strange to the players of two centuries ago. Not only did they play the Mini-Chinese in handicap games, they often played extensions in the opening that are wider than the extensions to make a base. I expect that AlphaGo is reviving that style of play.

There are 5 hits of the left side shape in waltheri, the oldest being Rin Kanketsu in 2010 and 4 in 2015 (other corners were usually black top right but not 4-4 and white bottom right). Making further extensions from the top left joseki when you have the corner beyond seems fairly standard to me, in fact I remember learning from some book (maybe [sl=ADictionaryOfModernFusekiTheKoreanStyle]A Dictionary Of Modern Fuseki, The Korean Style[/sl]) that if you do the hanging connection and normal joseki extension it's inefficient with a 4-4 below (I recall playing as white against this on OGS and got a good result):

Click Here To Show Diagram Code: [go]$$Bcm11 Joseki inefficient (4-4 not helped enough) $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . 1 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . X . . . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

So play a move which helps the corner too, and if white invades on the left around a you can jump out around b.

Click Here To Show Diagram Code: [go]$$Bcm11 Better balance $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . O . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X . . . b . . . . . . . . . . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . a . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 1 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . X . . . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

So anticipating this fight, in which black's hanging connection group is fairly flexible and can run out nicely, white may peep first, trying to make it heavier and less easy to jump out (on this half board on waltheri it has 68 games and 54.5% win, vs 89 games with 49.5% win for the one-space extension: not hugely significant but a valuable hint I think):

Click Here To Show Diagram Code: [go]$$Wcm10 White's counter $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . O , . . . . . , . . . . . X . . . | $$ | . . O X . . . . . . . . . . . . . . . | $$ | . . X . 1 . . . . . . . . . . . . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . X . . . . . , . . . . . O . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Edit: P.S.

if you do the hanging connection and normal joseki extension it's inefficient with a 4-4 below (I recall playing as white against this on OGS and got a good result)

Here's the game: how should black answer the approach the the 4-4? Maybe iron pillar? Because if you one point jump then 3-3 seems to be good for white as the smooth lead I got in the game seems to show.

Life In 19x19

This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re:

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that

Re: This 'n' that