Dear Cassandra,
No one could disagree with you when you show the difficulties for J89 to apply articles 7.1 and 7.2 and reach the expected result (I mean the result expected by japonese profesionnals). A lot of other examples have been documented, especially by Robert.
The conclusion is clear : article 7.1 and 7.2 has to be completly reworded.
My view is that Robert made a very great progress by his J2003 rule, by introducing a ko-pass (instead of the pass-for-ko in J89) and by introducing the local-2 concept to create independancies between regions.
With this J2003 a lot of positions (not correctly analysed by J89) are now given the expected result with J2003.
Does it remain some issues with this J2003?
Unfortunetly my answer is yes:
In the position above, found by Lightvector, we should declare all white stones alive and black marked stones dead while J2003 declares a seki
and it is the same in this position I found myself : we should declare all white stones alive and black marked stones dead while J2003 declares a seki
I am not aware of other positions which could be an issue for J2003. Maybe Robert can tell us if such other positions have been identified.
BTW I believe I have found a way to correct J2003 on this point but I have still to make some verifications.
Japonese counting
-
Gérard TAILLE
- Gosei
- Posts: 1346
- Joined: Sun Aug 23, 2020 2:47 am
- Rank: 1d
- GD Posts: 0
- Has thanked: 21 times
- Been thanked: 57 times
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
What is your opinion about the status of White's marked group under J89?Gérard TAILLE wrote:Dear Cassandra,
No one could disagree with you when you show the difficulties for J89 to apply articles 7.1 and 7.2 and reach the expected result (I mean the result expected by japonese profesionnals).
Addit "example-11-condition" (which turned the results of examples 16 to 18 "correct"):
Ko-bans can be only lifted in the order of their appearance.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
Quite apparently, even simple ko flew under the radar of the authors of a special rule to stop triple-ko positions with distributed ko shapes becoming "no result" during actual play (respectively "seki" in the status confirmation).Gérard TAILLE wrote: In the position above, found by Lightvector, we should declare all white stones alive and black marked stones dead while J2003 declares a seki.
+ + + + + + + + + + + +
By the way:
The position above is flawed, in that it belongs to the undefined crowd of unfinished positions.
It is a hanami-ko for Black, and so will not survive until the end of the game. White will have to add a move to terminate this ko fight.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
The same principle as in my posting above applies here:Gérard TAILLE wrote: and it is the same in this position I found myself : we should declare all white stones alive and black marked stones dead while J2003 declares a seki.
This position will not see the end of the game.
During status confirmation for the White stones, there will arise a double-ko, which is the reason for becoming this position a seki under J2003 (I suppose).
However, in a real game, this double-ko will become a source of unlimited ko threats for Black. It is very unlikely that no other ko will arise on the board.
Therefore, White would add a move in due time in a real game.
+ + + + + + + + + + + + + +
I think that this approach with unfinished and / or isolated positions has a similar problem as J89 has:
-- J89 tries to declare groups "dead", which could never be killed during actual play (especially triple-ko positions with distributed ko shapes).
(Would it be so disastrous to accept these (very rare) positions being "seki", similar to other cases that contain groups that are not of the "two-eyed-alive" type, but still cannot be captured? I.e. not using any special kind of ko-ban during status confirmation.)
-- The approach to generate "rule beasts" that contain unfinished and isolated positions tries to declare specific rulesets to be "flawed", despite we will never see such positions in the real world.
(Would it be so disastrous to accept these (very rare) positions being "abnormal during status confirmation", instead of trying to "correct" the corresponding ruleset?)
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Gérard TAILLE
- Gosei
- Posts: 1346
- Joined: Sun Aug 23, 2020 2:47 am
- Rank: 1d
- GD Posts: 0
- Has thanked: 21 times
- Been thanked: 57 times
Re: Japonese counting
Oops, the stones marked in your diagram form really three groups and you have to analyse them one by one To me such nested ko acts like the common double ko. In J89 this type of ko prevents killing a string of stones with two liberties. In the diagram above all white stones are alive except the two marked stones which are dead:Cassandra wrote:What is your opinion about the status of White's marked group under J89?Gérard TAILLE wrote:Dear Cassandra,
No one could disagree with you when you show the difficulties for J89 to apply articles 7.1 and 7.2 and reach the expected result (I mean the result expected by japonese profesionnals).
Now if
pass-for-ko created by 
then we have a loop and only one single white stone have been captured.If
is played in the upper left corner then
and white will be able connect the 
and create new uncapturable stones.-
Gérard TAILLE
- Gosei
- Posts: 1346
- Joined: Sun Aug 23, 2020 2:47 am
- Rank: 1d
- GD Posts: 0
- Has thanked: 21 times
- Been thanked: 57 times
Re: Japonese counting
You are right Cassandra this position can considered unfinished. But Lightvector proposed also a small change: The marked stone has been added to the diagram. Do you still consider white has to add a move?Cassandra wrote:Quite apparently, even simple ko flew under the radar of the authors of a special rule to stop triple-ko positions with distributed ko shapes becoming "no result" during actual play (respectively "seki" in the status confirmation).Gérard TAILLE wrote: In the position above, found by Lightvector, we should declare all white stones alive and black marked stones dead while J2003 declares a seki.
+ + + + + + + + + + + +
By the way:
The position above is flawed, in that it belongs to the undefined crowd of unfinished positions.
It is a hanami-ko for Black, and so will not survive until the end of the game. White will have to add a move to terminate this ko fight.
What is the difference? Now, in normal play, black is dead even if she starts to play and even if black has some unremovable ko threats at her disposal!
In J89 or J2008 it is still seki isn't it?
-
Gérard TAILLE
- Gosei
- Posts: 1346
- Joined: Sun Aug 23, 2020 2:47 am
- Rank: 1d
- GD Posts: 0
- Has thanked: 21 times
- Been thanked: 57 times
Re: Japonese counting
This position is quite interesting Cassandra.Cassandra wrote: The life-and-death examples in J89 -- treating a compound of bent-four and double-ko -- do NOT cover ALL possible variations of shared double-ko formations. J89 suppresses an example like the following one -- with NESTED ko shapes.
Let's put J89 and J2003 aside for a moment. What is for you the expected result for japonese professionnals about the marked black stones?
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
I am afraid that I am unable to have a look into the mind of Japanese professional players.Gérard TAILLE wrote:This position is quite interesting Cassandra.
Let's put J89 and J2003 aside for a moment. What is for you the expected result for japonese professionnals about the marked black stones?
But interpolating the given results of the corresponding J89 examples -- compound of bent-four and double-ko -- the "expected" result would claim all the marked stones being "dead".
Robert's J2003 will not have any problems to agree with that "expectation" (assumed that I applied these right).
However, J89 is the much more interesting environment.
I am very sure that it had a reason that J89 displays relative trivial sequences, but never a "solution" to the status assessment of the much more difficult examples.
Everyone loves "inside out" when it comes to sushi, but in J89 this nested double-ko shape (of White's envolved group) is used only on the "outside".
Try to apply J89 on this position (or the corresponding variation of example 16), and you will realise why.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
RobertJasiek
- Judan
- Posts: 6273
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Japonese counting
Unlike Cassandra, I think that we cannot just dismiss these positions as unsettled and therefore immaterial as status precedents because the position might have the colour-reversed copy on the board so that a jigo is possible. Then we still need to determine statuses.Gérard TAILLE wrote: In the position above, found by Lightvector, we should declare all white stones alive and black marked stones dead while J2003 declares a seki
However, I need to be convinced whether / why Japanese pros would have an unambiguous common sense on the statuses in this position. IMO, they do not. Therefore, J2003 is fine for this position.
I lack time to check this position and the position with a three groups seki with asymmentric captuable strings carefully. ATM, I presume your analysis was correct. If so:and it is the same in this position I found myself : we should declare all white stones alive and black marked stones dead while J2003 declares a seki
- Congratulations on finding these important positions!
- The position above is the only counter-example to J2003/35a with its intention to model all positions (except the flawed official examples 16-18) according to common J pro understanding if implicitly it exists for a position.
- Your seki position is more important because more basic / probably more frequent than hane-seki.
None.I am not aware of other positions which could be an issue for J2003. Maybe Robert can tell us if such other positions have been identified.
Earlier positions (only one?) for earlier J2003 versions until Mai 2004 with version 35/35a led to its creation. So it took 17 years for the first counter-example in 2021-08.
One serious alternative J2003 rules design was proposed but with essentially the same effect.
If J2003 ko rules shall be redesigned, ko-pass should be avoided. I do not know if it is possible while maintaining most traditional beasts.
-
Gérard TAILLE
- Gosei
- Posts: 1346
- Joined: Sun Aug 23, 2020 2:47 am
- Rank: 1d
- GD Posts: 0
- Has thanked: 21 times
- Been thanked: 57 times
Re: Japonese counting
My position above (using ko-pass concept) is certainly an issue for J2003 but let me say that my seki here under is not an issue for this rule; it is just a curiosity:RobertJasiek wrote:Unlike Cassandra, I think that we cannot just dismiss these positions as unsettled and therefore immaterial as status precedents because the position might have the colour-reversed copy on the board so that a jigo is possible. Then we still need to determine statuses.Gérard TAILLE wrote: In the position above, found by Lightvector, we should declare all white stones alive and black marked stones dead while J2003 declares a seki
However, I need to be convinced whether / why Japanese pros would have an unambiguous common sense on the statuses in this position. IMO, they do not. Therefore, J2003 is fine for this position.
I lack time to check this position and the position with a three groups seki with asymmentric captuable strings carefully. ATM, I presume your analysis was correct. If so:and it is the same in this position I found myself : we should declare all white stones alive and black marked stones dead while J2003 declares a seki
- Congratulations on finding these important positions!
- The position above is the only counter-example to J2003/35a with its intention to model all positions (except the flawed official examples 16-18) according to common J pro understanding if implicitly it exists for a position.
- Your seki position is more important because more basic / probably more frequent than hane-seki.
None.I am not aware of other positions which could be an issue for J2003. Maybe Robert can tell us if such other positions have been identified.
Earlier positions (only one?) for earlier J2003 versions until Mai 2004 with version 35/35a led to its creation. So it took 17 years for the first counter-example in 2021-08.
One serious alternative J2003 rules design was proposed but with essentially the same effect.
If J2003 ko rules shall be redesigned, ko-pass should be avoided. I do not know if it is possible while maintaining most traditional beasts.
The marked black group on the left is dead while the marked black group on the right is alive but there are no consequences. Why? Because the big black group and the two white groups are alive with dame and that means that all the position is seki. IOW you just do not need to look for the status of the marked black groups of stones. Dead or alive will lead to the same seki result which is fine.
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
I am quite sure that the concrete result of a status assessment would be completely irrelevant, if such an "unsettled" position was mirrored elsewhere on the board.RobertJasiek wrote:Unlike Cassandra, I think that we cannot just dismiss these positions as unsettled and therefore immaterial as status precedents because the position might have the colour-reversed copy on the board so that a jigo is possible.
Last edited by Cassandra on Thu Aug 19, 2021 9:54 am, edited 1 time in total.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
J2003 is fine for examples 16 to 18, too, according to my understanding.RobertJasiek wrote:- The position above is the only counter-example to J2003/35a with its intention to model all positions (except the flawed official examples 16-18) according to common J pro understanding if implicitly it exists for a position.
Your ko-pass rule lifts BOTH J89 ko-bans in the double-ko AT ONCE, what is the diciding feature.
What we do not know concerning examples 16 to 18 being potentially "flawed", because it cannot be concluded from the "known" English translation of J89:
Does the Japanese understanding of lifting ko-bans during status confirmation demand that this is done in order?
(You will realise that ALL the ko-ban-liftings, which are specifically stated with the examples, follow this principle.)
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
Quick "solution" for Robert:RobertJasiek wrote:I lack time to check ... the position with a three groups seki with asymmentric captuable strings carefully. ATM, I presume your analysis was correct.
. Please note that it would be the ATTACKER, who utilised this means.)
would capture at "a", and capture White's stones at the top a bit later.))The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: Japonese counting
This is IMPOSSIBLE, Robert.RobertJasiek wrote:If J2003 ko rules shall be redesigned, ko-pass should be avoided. I do not know if it is possible while maintaining most traditional beasts.
But may I suggest writing a ko-pass rule that is as SPECIFIC as possible, and which is designed ONLY for those cases that cannot be solved without?
I might be mistaken, but I am sure that the sole and only purpose for the J89 ko-ban was to enable the killing of TRIPLE-ko groups.
See e.g. life-and-death example 8 "triple ko with an eye on one side".
However, the authors used a SINGLE-ko ban for this purpose, and -- as we now know -- this what not the very best idea, as they threw out the baby with the bathwater.
Even your J2003 ko-pass rule, which of course solves the TRIPLE-ko issues, has unseen / unwanted side effects on SINGLE-ko (e.g. lightvectors finding) and DOUBLE-ko (e.g. Gérard's example).
Therefore, write a ko-pass rule for TRIPLE-ko, and undesirable side effects on SINGLE-ko and DOUBLE-ko are excluded from the outset.
To give you an idea of what I have in mind (no idea to what extend this has to do with the "solution"):
"An uninterrupted sequence of TRIPLE-ko captures must not repeat its starting position."
I.e. latest move 6 of the TRIPLE-ko cycle is forbidden.
At least, this would be sufficient to solve example 8.
To emphasise again: I do not have any idea about potential side effects
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Gérard TAILLE
- Gosei
- Posts: 1346
- Joined: Sun Aug 23, 2020 2:47 am
- Rank: 1d
- GD Posts: 0
- Has thanked: 21 times
- Been thanked: 57 times
Re: Japonese counting
Don't be so sure Cassandra. I think I have a solution without ko-pass. I continue my verificationsCassandra wrote:This is IMPOSSIBLE, Robert.RobertJasiek wrote:If J2003 ko rules shall be redesigned, ko-pass should be avoided. I do not know if it is possible while maintaining most traditional beasts.
Your proposal does not solve the problem Cassandra because you act only on triple ko while the identified issues remaining on J2003 concerns only double ko.Cassandra wrote: But may I suggest writing a ko-pass rule that is as SPECIFIC as possible, and which is designed ONLY for those cases that cannot be solved without?
I might be mistaken, but I am sure that the sole and only purpose for the J89 ko-ban was to enable the killing of TRIPLE-ko groups.
See e.g. life-and-death example 8 "triple ko with an eye on one side".
However, the authors used a SINGLE-ko ban for this purpose, and -- as we now know -- this what not the very best idea, as they threw out the baby with the bathwater.
Even your J2003 ko-pass rule, which of course solves the TRIPLE-ko issues, has unseen / unwanted side effects on SINGLE-ko (e.g. lightvectors finding) and DOUBLE-ko (e.g. Gérard's example).
Therefore, write a ko-pass rule for TRIPLE-ko, and undesirable side effects on SINGLE-ko and DOUBLE-ko are excluded from the outset.
To give you an idea of what I have in mind (no idea to what extend this has to do with the "solution"):
"An uninterrupted sequence of TRIPLE-ko captures must not repeat its starting position."
I.e. latest move 6 of the TRIPLE-ko cycle is forbidden.
At least, this would be sufficient to solve example 8.
To emphasise again: I do not have any idea about potential side effects