Life In 19x19
http://www.lifein19x19.com/

Judging the early health of a corner
http://www.lifein19x19.com/viewtopic.php?f=11&t=7964		 Page 1 of 1
Author:  PeterPeter [ Mon Feb 25, 2013 9:23 am ]
Post subject:  Judging the early health of a corner
Early on in the game, when there are perhaps only a few stones in a corner, it is really useful to have an idea of whether that corner is probably safe, probably dead, or in the balance.

With so many possible variations, it is well beyond my ability to try to read it out. So, are there are any rules of thumb that usually give a good indication? What signs do you look for?

Just as an example, I was black in this corner of a 13x13 board:

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O . |
$$ . X . . . . . . O X . |
$$ . . . X . . . O . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]

I thought white looked safely alive there, so I played in another corner, but I am now wondering if I should have attempted to kill it. For what it’s worth, the computer later had it marked down as black territory, which is what prompted me to re-consider it.
Author:  Uberdude [ Mon Feb 25, 2013 9:36 am ]
Post subject:  Re: Judging the early health of a corner
PeterPeter wrote:
Early on in the game, when there are perhaps only a few stones in a corner, it is really useful to have an idea of whether that corner is probably safe, probably dead, or in the balance.

With so many possible variations, it is well beyond my ability to try to read it out. So, are there are any rules of thumb that usually give a good indication? What signs do you look for?


Well, it sounds rather facetious, but if you have the corner territory/eyespace you are probably safe, if you don't then you are probably (locally) dead. In your example it is still undecided who has the corner: if white gets to play atari at 2-3 then it looks rather like a white corner, if black gets to extend there it looks rather black. Obviously you need a big enough space in the corner to live, that's where learning your L groups comes in handy to get a feel for how much space in a corner you need to live. Of course if you aren't surrounded then you can always run away if you can't get 2 eyes in the corner.
Author:  amnal [ Mon Feb 25, 2013 9:46 am ]
Post subject:  Re: Judging the early health of a corner
Whoah, white is potentially very weak here. Even if you can't kill, you have a choice of powerful sente moves and can certainly seal white in totally so it's very big to play here anyway. The exact best choice will depend on the board, but looking at this area alone I'd like to play like the following:

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O . |
$$ . X . . . . . . O X . |
$$ . . . X . . . O . 1 . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]


This is very big and reverse sente, and it clearly threatens white's life. A black move on the other side leaves white with nowhere to go.

So what can white do?

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O . |
$$ . X . . x . x . O X . |
$$ . . . X . 2 . O . 1 . |
$$ . . . . . . x . . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]


This is the simple obvious move, but now black has several options (the xs). All either kill white or harrass him or simply seal him in where he may or may not be alive (but who cares because thickness, right?). White probably doesn't have room to live locally if you don't want him to, because you're strong on both sides and can simply poke at his eyespace.

Lets contrast with what white can do if he gets to play first:

Click Here To Show Diagram Code
[go]$$W
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O 2 |
$$ . X . . . . x . O X . |
$$ . . . X . 3 . O . 1 x |
$$ . . . . . . . . x x . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]


White could also tenuki 3, but that leaves the corner unsettled and black gets to play on the shape points again and maybe even kill since there isn't very much room in the corner. Even after :w3: black has the xs to aim at, but he no longer can destroy white's local shape so much in a single move.

In any case, white's looking much stronger and with territory now, right...plus he can move into the centre a little. That's a big swing from the big annoyance you can cause him with just one move. Of course, he's still a bit worried because the x points really could cause a lot of trouble...the original position is locally excellent for black even if white gets to play first.

The key in this shape is that the corner really isn't settled because just one move can swing things so much. Everything else tends to be very situational, which is why we have reading and (as Uberdude has said) why it helps to know some basic shapes that you can aim your reading at.

Edit: To be clear, I'm handwaving away some stuff above. I think black can probably kill white outright, and white possibly can't live locally even if he plays first, but I'm aiming mainly to show why the choice of who moves first makes such a big difference.
Author:  Uberdude [ Mon Feb 25, 2013 10:00 am ]
Post subject:  Re: Judging the early health of a corner
amnal wrote:
The key in this shape is that the corner really isn't settled because just one move can swing things so much. Everything else tends to be very situational, which is why we have reading and (as Uberdude has said) why it helps to know some basic shapes that you can aim your reading at.


You don't necessarily even need to read, I was just talking about a feeling for how big a corner needs to be to live, so in this example:

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O . |
$$ . X . . . . . . O X . |
$$ . . . X . . . O . 1 . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]


White can't even make an L group in the corner as he doesn't have any of the corner, so if L is dead and this is much less space its ultra-dead.

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X . X |
$$ . X . . . . . . O X . |
$$ . . . X . . . @ S @ . |
$$ . . . . . . . S . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]


This looks about L shape sized, so if white tenukis it's probably dead.

Click Here To Show Diagram Code
[go]$$W
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X . X |
$$ . X . . . . . . O X . |
$$ . . . X . @ S @ S @ . |
$$ . . . . . S . . . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]


This looks bigger than an L so I am hopeful it can live, but the problem is all those weak points amnal mentioned so meh, now it's time to read instead of feel.
Author:  PeterPeter [ Mon Feb 25, 2013 10:22 am ]
Post subject:  Re: Judging the early health of a corner
Is this a good question to ask:

Can I keep White out of the (4? 6? 8? 9?) cornermost points?
Author:  Uberdude [ Mon Feb 25, 2013 10:52 am ]
Post subject:  Re: Judging the early health of a corner
PeterPeter wrote:
Is this a good question to ask:

Can I keep White out of the (4? 6? 8? 9?) cornermost points?


I don't think like that. It's more just a feeling of how much space you have in the corner. Once 100 of your groups that were too small have died, and 100 of your opponent's groups that were big enough have lived, you too will have a better feeling for how much space you need.
Author:  Bill Spight [ Mon Feb 25, 2013 11:44 am ]
Post subject:  Re: Judging the early health of a corner
PeterPeter wrote:
Early on in the game, when there are perhaps only a few stones in a corner, it is really useful to have an idea of whether that corner is probably safe, probably dead, or in the balance.

With so many possible variations, it is well beyond my ability to try to read it out. So, are there are any rules of thumb that usually give a good indication? What signs do you look for?

Just as an example, I was black in this corner of a 13x13 board:

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O . |
$$ . X . . . . . . O X . |
$$ . . . X . . . O . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]

I thought white looked safely alive there, so I played in another corner, but I am now wondering if I should have attempted to kill it. For what it’s worth, the computer later had it marked down as black territory, which is what prompted me to re-consider it.


What do you need to live? Seki aside, you need eyes. Where are White's eyes?

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . 3 . X X O . |
$$ . X . . . . . . O X . |
$$ . . . X . 2 . O . 1 . |
$$ . . . . . . C . . . . |
$$ . . . . . . C . . . . |
$$ ----------------------[/go]


If :b1:, :w2: makes a potential eye at the marked points. :b3: prevents White from escaping to the center. Can White now make another eye, as well as firming up his potential eye?

One rule of thumb is One eye plus access to the center. Early in the game, that is usually good enough for life. Later in the game the opportunities to make an eye in the center are reduced. Even granting White an eye here, White has no access to the center, and Black is strong on both sides. Without reading, you can tell that it would be difficult for White to live.

Click Here To Show Diagram Code
[go]$$W
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O 2 |
$$ . X . . . . . . O X . |
$$ . . . X . 3 . O . 1 . |
$$ . . . . . . C C C . . |
$$ . . . . . . C C C . . |
$$ ----------------------[/go]


After :w1: - :w3: White has an eye on the side and a possible escape to the center.

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . 1 . X X . X |
$$ . X . . . . . . O X . |
$$ . . . X . O . O . O 2 |
$$ . . . . . . C C C C C |
$$ . . . . . . C C C C C |
$$ ----------------------[/go]


If :b1: blocks White out of the center, :w2: enlarges White's potental eye space. There is another rule of thumb. If you make at least 8 points of territory you are alive. There are exceptions, of course. ;)

Here White looks good, with 10 potential points.

Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . X . X X . X |
$$ . X . . . . . . O X . |
$$ . . . X . O . O . O O |
$$ . . . . . . C . C . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]


The trouble is, White has weaknesses. For instance, Black has possible peeks at the marked points. The position is too complicated to rely upon rules of thumb. It is also not so easy to read.
Author:  PeterPeter [ Mon Feb 25, 2013 2:19 pm ]
Post subject:  Re: Judging the early health of a corner
Click Here To Show Diagram Code
[go]$$B
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . X . |
$$ . . . . . . . X X O 1 |
$$ . X . . . . . . O X . |
$$ . . . X . . . O . . . |
$$ . . . . . . . . . . . |
$$ . . . . . . . . . . . |
$$ ----------------------[/go]

If, as black, I was going to have a go at killing white, my instinct was to start at :b1: above. With that start, white does look a little more healthy.

Seems I still have a long way to go.
Author:  xed_over [ Mon Feb 25, 2013 2:39 pm ]
Post subject:  Re: Judging the early health of a corner
PeterPeter wrote:
If, as black, I was going to have a go at killing white, my instinct was to start at :b1: above. With that start, white does look a little more healthy.

yeah, but that stone is already dead, you want to kill it more? :)

its the other stones you have to worry about. you've got to steal their chances for survival
Author:  yoyoma [ Mon Feb 25, 2013 3:18 pm ]
Post subject:  Re: Judging the early health of a corner
PeterPeter wrote:
For what it’s worth, the computer later had it marked down as black territory, which is what prompted me to re-consider it.


What does this mean? Did you play till the end, both sides pass, and White (the computer I guess) marked those stones dead? Can you post the end position or the entire game?
Author:  PeterPeter [ Mon Feb 25, 2013 3:32 pm ]
Post subject:  Re: Judging the early health of a corner
yoyoma wrote:
PeterPeter wrote:
For what it’s worth, the computer later had it marked down as black territory, which is what prompted me to re-consider it.


What does this mean? Did you play till the end, both sides pass, and White (the computer I guess) marked those stones dead? Can you post the end position or the entire game?

Bizarrely, the app I am using (AI Factory Go on Android) doesn't have the option to save or export games (I really need to find a better app). What it does do, at the end of each game, is let you flick backwards and forwards between the moves, with its territory forecast overlaid. It is quite interesting to see the changes during a quick review of a 13x13 game.

The final position had that corner alive for white, because I switched attention to another corner early on. I ended up losing that game.
Author:  PeterPeter [ Mon Feb 25, 2013 3:37 pm ]
Post subject:  Re: Judging the early health of a corner
xed_over wrote:
PeterPeter wrote:
If, as black, I was going to have a go at killing white, my instinct was to start at :b1: above. With that start, white does look a little more healthy.

yeah, but that stone is already dead, you want to kill it more? :)

its the other stones you have to worry about. you've got to steal their chances for survival

Having a ponnuki so near enemy stones is usually helpful in fights.

I can see that it is wrong now though.
Author:  Dusk Eagle [ Mon Feb 25, 2013 3:54 pm ]
Post subject:  Re: Judging the early health of a corner
PeterPeter wrote:
yoyoma wrote:
PeterPeter wrote:
For what it’s worth, the computer later had it marked down as black territory, which is what prompted me to re-consider it.


What does this mean? Did you play till the end, both sides pass, and White (the computer I guess) marked those stones dead? Can you post the end position or the entire game?

Bizarrely, the app I am using (AI Factory Go on Android) doesn't have the option to save or export games (I really need to find a better app). What it does do, at the end of each game, is let you flick backwards and forwards between the moves, with its territory forecast overlaid. It is quite interesting to see the changes during a quick review of a 13x13 game.

The final position had that corner alive for white, because I switched attention to another corner early on. I ended up losing that game.


If you're relying on the computer to tell you whether a group is alive or dead, then clearly you're passing too early. If we just suppose for a second that the group is 100% alive no matter what you do, then you still need to play all of the endgame associated around that shape. Or if you're not sure whether the group is dead or not, why did you not kill it during the game? What if white tried to save it and lived? Surely that would be worth a lot more points than the 1-2 point endgame moves you were playing elsewhere, which means you should've killed it before playing those endgame moves.
Author:  PeterPeter [ Mon Feb 25, 2013 4:23 pm ]
Post subject:  Re: Judging the early health of a corner
Dusk Eagle wrote:
If you're relying on the computer to tell you whether a group is alive or dead, then clearly you're passing too early.

I wasn't exactly relying on the computer to tell me whether it is alive or dead. During the game, my judgment was that white was probably alive there, so I played a big move in another corner. On review afterwards, the computer said that territory should have been mine (I doubt it was 100% certain either). The consensus on this thread is that my judgment was wrong. Given that I cannot read out life-and-death problems like that yet, I am hoping to improve my judgment on whether, on balance, something looks alive or dead.

Dusk Eagle wrote:
If we just suppose for a second that the group is 100% alive no matter what you do, then you still need to play all of the endgame associated around that shape. Or if you're not sure whether the group is dead or not, why did you not kill it during the game?

At that early stage of the game, given that I thought it was alive, I thought I had bigger moves elsewhere. I thought that it looked reasonably well contained for a while. In the latter half of the game, I gave more priority to containing it (it was too late to kill it by then).
Author:  billywoods [ Mon Feb 25, 2013 9:23 pm ]
Post subject:  Re: Judging the early health of a corner
I haven't read through this thread very carefully, but here's a point worth making if it hasn't been made already:

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


Here's how one continuation might go. Notice how desperate white's moves are, and that black's 1, 3, 5 and 7 are all sente. (To be clear: 1, 2 and 3 are good. 4-8 are squirmy eyespace moves, and I'm not necessarily saying they're the best moves for either side!)

Even if white plays some awful squirmy moves and lives, you've still profited here. Usually, when thinking about killing a group, I first think about taking away its base, narrowing its eyespace and sealing it in, and then check to see whether it looks killable. If it turns out to be too big to kill, maybe that's annoying - but here, white ended up with about 4 points of territory, got sealed in and lost the corner. Black has more than 4 points in the lower right corner alone(!), plus gained thickness on the outside and towards the lower side, and has sente. White lived, but this is a complete failure for white anyway.
Author:  RobertJasiek [ Mon Feb 25, 2013 10:18 pm ]
Post subject:  Re: Judging the early health of a corner
Instead of speaking of "probably" alive / dead, speak of the "n-alive" degree! This is much simpler, because you are not confronted with infinitely many possible values of probabilities but only with the possible values for n = -2 (or smaller), -1, 0, 1, 2 (or bigger), *.

http://senseis.xmp.net/?NAlive

Ask yourself "Is the group at least 0-alive?" and answer that by reading. If yes, answer 1-alive. If no, answer -1-alive etc.

Yes, reading is necessary, of course.
Author:  Dusk Eagle [ Mon Feb 25, 2013 11:51 pm ]
Post subject:  Re: Judging the early health of a corner
Quote:
Ask yourself "Is the group at least 0-alive?" and answer that by reading.
Given that this is where the "probably" part of the discussion comes from, I don't think your explanation is any simpler.
Author:  RobertJasiek [ Tue Feb 26, 2013 12:49 am ]
Post subject:  Re: Judging the early health of a corner
Reading is not simple, but the following considerations for the attacker prune the complexity:

- surround the group
- reduce the eyespace while keeping the reduction strings string- or directly connected
- play functional sente (such as threats to cut)
- play reverse sente to prevent forcing moves
- prevent partition of the eyespace
- construct nakade
- keep the eye value below 2
- avoid seki and ko
Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/