Life In 19x19 http://www.lifein19x19.com/ |
|
What means "approach ko" ? http://www.lifein19x19.com/viewtopic.php?f=12&t=18561 |
Page 1 of 1 |
Author: | Gérard TAILLE [ Wed Jan 05, 2022 11:35 am ] |
Post subject: | What means "approach ko" ? |
How do you call the result of this semeai ? Assume black has at least two ko threats. The best sequence (I am not quite sure it is really the best sequence ![]() and only now we reach a "direct ko" Seeing that after ![]() ![]() But after ![]() |
Author: | dhu163 [ Wed Jan 05, 2022 11:54 am ] |
Post subject: | Re: What means "approach ko" ? |
After ![]() ![]() My understanding of terms is that n-move approach ko refers to a single direct ko locally that requires n moves by one player (more than the opponent) in order to start it. This refers to two distinct positions depending on which side has taken the ko. I equate this term with n-step ko. Hence a direct ko is a 0-step ko. An n-stage ko is something different and two nearby kos may be nonindependent though not adjacent, so they are not double kos, nor 2-stage kos but something different. Of course in theory, you might have any arrangement of kos with differing dependences but all other unnamed ones are very rare (1 in million? for each) |
Author: | Knotwilg [ Wed Jan 05, 2022 11:55 am ] |
Post subject: | Re: What means "approach ko" ? |
At SL we call a ko resulting in another ko, a "two stage ko", and a ko that can't be resolved by winning it but needs one additional move hence another ko, an approach ko. See https://senseis.xmp.net/?MultiStageKo and https://senseis.xmp.net/?ApproachKo But dhu seems to disagree here. |
Author: | dhu163 [ Wed Jan 05, 2022 11:59 am ] |
Post subject: | Re: What means "approach ko" ? |
I think I have agreed with SL on step/approach ko, but I feel SL's definition of stage ko is slightly ambiguous. I think GT is counting B's two passes, but forgets that B captured a stone which is like W passing once. Hence as they started with equal liberties (ignoring the ko liberties), it is a (2-1)=1 step ko. But I see that there is ambiguity. I would call the two positions that are n approach moves from a direct ko an n-move approach ko. Bill's paper evaluating them seems to use this. This is preferable in Bill's NTE model since this allows us to say that moves to end the n-step ko (including moving to (n-1 step ko) have the same gain as ko takes in an n-step ko. However, SL is a bit confusing in the suggestion of splitting positions by number of ko threats ignored in order to win the ko. This would mean say in a direct ko, you to play, there are two positions, one you can win immediately and the other you need to ignore a threat. Of course, ![]() ![]() |
Author: | Schachus [ Wed Jan 05, 2022 12:02 pm ] |
Post subject: | Re: What means "approach ko" ? |
I would call this a 1-move approach ko. I would have had both sides continuing play without passing until the ko is direct for one side. Sure black might tenuki for 6, but than white can also tenuki for 7. if we have b6 take a liberty then w7 takes the ko. Now the ko is direct for b but w needs one more approach move (7 in your variation to make a direct ko. That means b can tenuki for 9 and w needs to play the approach move if he wants to fight the local situation( if he also tenukis, he accepts any tenuki as a ko threat) |
Author: | Gérard TAILLE [ Wed Jan 05, 2022 3:48 pm ] |
Post subject: | Re: What means "approach ko" ? |
dhu163 wrote: I think I have agreed with SL on step/approach ko, but I feel SL's definition of stage ko is slightly ambiguous. I think GT is counting B's two passes, but forgets that B captured a stone which is like W passing once. Hence as they started with equal liberties (ignoring the ko liberties), it is a (2-1)=1 step ko. But I see that there is ambiguity. I would call the two positions that are n approach moves from a direct ko an n-move approach ko. Bill's paper evaluating them seems to use this. This is preferable in Bill's NTE model since this allows us to say that moves to end the n-step ko (including moving to (n-1 step ko) have the same gain as ko takes in an n-step ko. However, SL is a bit confusing in the suggestion of splitting positions by number of ko threats ignored in order to win the ko. This would mean say in a direct ko, you to play, there are two positions, one you can win immediately and the other you need to ignore a threat. Of course, ![]() ![]() I do not understand why ![]() ![]() Could you clarify in which kind of situations the move ![]() ![]() |
Author: | dhu163 [ Wed Jan 05, 2022 6:07 pm ] |
Post subject: | Re: What means "approach ko" ? |
Thanks for the exercise. Did I get it wrong? Well B plays first with one more liberty so I expected them to win the shape unconditionally rather than getting to a ko even if it is 1 step ko. However, you are right that deciding between them is difficult. This is a 1 step ko in B's hand in gote, but B's forcing moves at ![]() ![]() Note that if ![]() ![]() ![]() ![]() I think that ![]() ![]() ![]() ![]() ![]() ![]() The average is before ![]() ![]() B gets a 2 step (B's favour) ko in W's hand in gote. The swing is now increased to 28+1 = 29, but the zero point hasn't shifted. At temperatures above the ko gain 29/8 = 3 5/8, this is close to but slightly worse for B than a 1 step ko in B's hand since 2K/8 > K/5. This is 2 factors, both worse for Black. The count is 29/4 = 7 1/4 Furthermore, W has the option of not playing ![]() ![]() Under NTE, when T is the gain of the first and smallest ko involved, an n-step (your favour) ko in opp's hand has the same count as an (n-1)-step (your favour) ko in your hand. I don't know how this varies with temperature. But as T goes down, smaller ko threats are required to maintain the ko since they no longer need to balance the value of sente. Local profit matters more than sente and hence winning the ko is more important. Is there any situation where the 2 step ko is better? last check This is like a golden chicken standing on one leg shape and doesn't look good for W initially. However, upon further inspection, B has gotten a 2 step ko (at ![]() ![]() I almost completely concede this one. The one-step ko is better for B in the large majority of cases. One last try: when could the 2 step ko possibly be better? W has to expect more than 2pts locally. Also, the temperature has to be low enough to change the influence of sente-gote and prioritise local variations. However, I can't see any possible way my 2 step ko could be better. Since it seems worse whichever side is to play. So the 1 step ko dominate in this case? |
Author: | Gérard TAILLE [ Thu Jan 06, 2022 4:44 am ] |
Post subject: | Re: What means "approach ko" ? |
dhu163 wrote: Thanks for the exercise. Did I get it wrong? Well B plays first with one more liberty so I expected them to win the shape unconditionally rather than getting to a ko even if it is 1 step ko. However, you are right that deciding between them is difficult. This is a 1 step ko in B's hand in gote, but B's forcing moves at ![]() ![]() Note that if ![]() ![]() ![]() ![]() I think that ![]() ![]() ![]() ![]() ![]() ![]() The average is before ![]() ![]() B gets a 2 step (B's favour) ko in W's hand in gote. Oops I need to go a little more slowly ![]() In this last diagramm I do not see the two step ko. IOW how black can play two tenuki moves before playing a direct ko? |
Author: | kvasir [ Thu Jan 06, 2022 6:08 am ] |
Post subject: | Re: What means "approach ko" ? |
I think it is redundant to refer to approach ko as n-step ko because it is a capturing race and 'n' is not the number of liberties. In fact a group in an n-step ko will often have n+1 liberties and n-1 if there is opportunity to connect the ko early to convert into a direct capturing race. It is not very useful. I have often seen 'step ko' used to mean what some people want to call a stage ko. I think that is because of the reason that less useful terms don't tend to stick. Myself, I tend to count 'n' like this: "zero", "few" and "too many" approach moves. I know it is called a "shameless ko" in Chinese when when 'n' is rather big. |
Author: | Gérard TAILLE [ Thu Jan 06, 2022 10:48 am ] |
Post subject: | Re: What means "approach ko" ? |
My initial question in this thread was : How do you call the result of this semeai ? To simplify my question let's take the position after the first 5 moves I proposed: Position 1 and now the same question : How do you call the result of this semeai ? If I understand correctly, dhu163 and Schachus call this a "one-move approach ko". Where is the point? According to the SL definition if this position is a "one-move approach ko" then, after a black tenuki move we should reach a direct ko. Position after black tenuki: As you can see black can follow by a move at "a" for a direct ko but after ![]() Strictly speaking, if you consider position 1 being a "one-move approach ko" then don't you have to modify the definition of "one-move approach ko" given in SL ? Position 2 BTW with this new position I consider it is really a "one-move approach ko", according to SL definition |
Author: | Gérard TAILLE [ Thu Jan 06, 2022 3:59 pm ] |
Post subject: | Re: What means "approach ko" ? |
dhu163 wrote: Thanks for the exercise. Did I get it wrong? Well B plays first with one more liberty so I expected them to win the shape unconditionally rather than getting to a ko even if it is 1 step ko. However, you are right that deciding between them is difficult. This is a 1 step ko in B's hand in gote, but B's forcing moves at ![]() ![]() Note that if ![]() ![]() ![]() ![]() I think that ![]() ![]() ![]() ![]() ![]() ![]() The average is before ![]() ![]() B gets a 2 step (B's favour) ko in W's hand in gote. The swing is now increased to 28+1 = 29, but the zero point hasn't shifted. At temperatures above the ko gain 29/8 = 3 5/8, this is close to but slightly worse for B than a 1 step ko in B's hand since 2K/8 > K/5. This is 2 factors, both worse for Black. The count is 29/4 = 7 1/4 Furthermore, W has the option of not playing ![]() ![]() Under NTE, when T is the gain of the first and smallest ko involved, an n-step (your favour) ko in opp's hand has the same count as an (n-1)-step (your favour) ko in your hand. I don't know how this varies with temperature. But as T goes down, smaller ko threats are required to maintain the ko since they no longer need to balance the value of sente. Local profit matters more than sente and hence winning the ko is more important. Is there any situation where the 2 step ko is better? last check This is like a golden chicken standing on one leg shape and doesn't look good for W initially. However, upon further inspection, B has gotten a 2 step ko (at ![]() ![]() I almost completely concede this one. The one-step ko is better for B in the large majority of cases. One last try: when could the 2 step ko possibly be better? W has to expect more than 2pts locally. Also, the temperature has to be low enough to change the influence of sente-gote and prioritise local variations. However, I can't see any possible way my 2 step ko could be better. Since it seems worse whichever side is to play. So the 1 step ko dominate in this case? I did some complementary works to compare a black move at "a" or "b". The simpliest positions for these compraraisons seems not under NTE but when white has one ko threat available: In this diagram with a white ko threat at "c" the correct move for black is a move at "a" In this second diagram with a unremovable white ko threat the correct move for black is now a move at "b" |
Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |