[Rules] proposals
Makrai Jozsef
More than a year ago there was a discussion about looking for well defined
rules suitable for both humans and machine gametrees. That time the
conclusion was that it may not be possible to get an area rules any better
than superko (with its numerous errors), quite a disappointment for me.
Now, after some decay in my go activity, I have some new ideas asking for
opinions.
Complementing the widely accepted go fundamentals (strings, liberties,
capture, ko), I think the most important details that need to be defined:
D1. The exact procedure for stopping-continuing-ending the game.
D2. The scoring method.
D3. Handling of cycles.
My primary expectations against the rules (mostly the same as before):
E1. Preserve simplicity of the fundamental rules, be easy to understand
and easy to apply (for man as well as machine), avoid the
introduction of artificial concepts such as "seki", "disturbing and
fighting kos", "wholeboard repetition", "superko", "superpass", etc.
E2. Honor traditional (e.g. send2-return1, moonshine) rulings as long as
possible.
E3. Produce no anomalies or erroneus/unnatural rulings.
E4. Allow the game to progress and eventually end and score even with a
noncooperative opponent (live stone remover, false disputer, etc.)
Basically, what I wish for is something similar to what chesss and othello
already has. ;o)
About D3: I have a strong preference for true cycles as draws. First, AFAIK
all three professional rulesets (Japanese, Chinese and Korean) practically
treat them as such. Second, common sense and the generic conceptions of
board games dictate this ruling as well: when neither side can achieve a
won terminal position, there is a draw.
Quick note about D1: The number of possible continuations must obviously be
limited. It is also important that this limitation do not have any
practical consequence, ie. no possible disputed dead stone capture should
ever be prevented. Without superko, it may not be possible to define any
special set of conditions for when should the end be forced. In this case,
the number of requestable continuations can simple be left to the
tournament rules to specify (not a perfect solution, though).
About D2: What I like in territory scoring is that dame can be left
unfilled (at least in friendly games). But it needs the special
confirmation phase, projecting the result back (and even that result is not
always unambigous, especially for computers).
To me, the strength of area scoring is that it has a strict theoretical
meaning on it's own: the number of controlled points of the board is one's
score. This makes it verifiable, understandable and undisputable. Territory
score, AFAIS, have no such intrinsic meaning, just a fortunate property of
being (in terms of difference) very close to area score. This theoretical
foundation causes several practical advantages: it may provide more
immunity against possible anomalies and special positions, it allows a
board position to be scored itself (without move history, prisoner count,
etc.). And most importantly, it makes dead stone removal by agreement a
simple and polite way to quicken the endgame, nothing more.
So, I would choose area score as reference. Two major problems that remain:
- The losing player can hold a game with an unbalanced cycle
(send2-return1, etc.), since prisoners don't count (putting it another
way: since the status of controlled points don't change by the fact
that the weak side have to give up an extra stone in each cycle to
force it on).
- Uncapturable but traditionally dead shapes, such as moonshine life or
bent four plus external threats.
The first one is an intrinsic property of area rules that needs an extra
rule to fix. My suggestion is a hackish rule saying players don't have
infinite number of stones, and whose supply run out can only pass. There
should also be an option for either player to claim an exchange of
prisoners (to avoid triple ko random exhaustion, and to make the actual
number of stones unsignificant (provided it is reasonable and roughly
equal, e.g. about 180 each on 19*19)). The advantages of this solution are
that it does not distort the actual game (unlike superko) because it only
has a theoretical effect; and that it is very close to OTB experience and
reality, therefore easy to understand and natural to apply. In fact, it
must already be applied in reality, regardless of rules. ;o) And as a nice
side effect it fixes single stone suicide as well, without requiring an
extra rule.
The second problem affects most rulesets. The reason is that those
positions are really not capturable in normal go, unless the ruleset
include some extra restriction of the players' moves. Two such restrictions
are the japanese pass for ko rule (as used in the confirmation phase) and
the superko rule. Unfortunately, the pass for ko rule in its traditional
form cannot achieve the desired effect (moonshine life is alive even in
confirmation). My suggestion here is another rule hack saying that when a
continuation of a stopped game is started either player can demand it to be
the last one with normal rules (last chance for threat elimination). The
rest of the game will then be played with modified (restricted) ko rules.
The advantage of this solution is (again) that it does not distort the game
itself (unlike superko), since it is only applied after the "real" part of
the game (and usually not applied at all).
Besides the restriction-based aspects (practical effects) as above, it may
also worth to look at these positions and ko rules from a viewpont of
theoretical background. Superko (which I particularly dislike from both
aspects) is based on the theory of wholeboard repetition, while the
Japanese rules with their pass for ko rule try to introduce the concept of
locality, ie. that at the end, all groups have to find local life
independently of kos and other places of the board.
I personally think that it is important to compare the results of any new
ko rules to normal go play. Any differences that are not especially called
for are signs of problems, IMHO. This is why I think this concept of
locality is also of a questionable value: It is not inherent in normal go,
but an artificial (though far less artificial than superko) invention. Why
should bent4 be killed unconditionally? Is there anything in the basics of
go that suggests that? The somewhat obscure human desire and tradition of
killing moonshine can probably be backed up by a different reasoning.
So for the modified ko rules my suggestion is as follows. When a player
could play a legal ko capture, he now also has a new option, to do a "PASS
FOR A KO LOCK" instead, specifying the ko capture he prepares. Later, if he
eventually succeeds in making the prepared ko capture (ie. the opponent
could not or did not use this extra opportunity to win the ko), the
opponent would not be allowed to recapture in that ko unless a regular
pass of either player intervened.
By this rule I hope to grab the concept of one sided kos, differentiate
(but not explicitly in the rules!) between a true, "fighting" ko and a
virtual ko such as moonshine that can not be won by one side only losing it
can be delayed infinitely. Most of my earlier attempts in this direction
failed because allowing an abuser to destabilize a double ko seki. This one
seems safer in this respect. ;o) I also hope for rulings that I personally
feel "correct", which are: true cycles, unstable multiple kos are to be
drawn in the game, stable multiple kos are either drawn or left alive
according the players choices, "fighting" kos to be played out just like in
normal go, and lopsided kos like moonshine be captured.
To summarize, my proposed ruleset is: area scoring, NO superko, limited
stone supply, and the modified ko option in continuations (as above).
Based on these ideas, a territory scoring ruleset also seem possible. This
would use a J89-like confirmation scheme, replacing the traditional
(flawed) pass-for-ko hack with my new ko hack. ;o) Although I'm more
interested in the area version now, some quick thoughts about the territory
ruleset: it would obviously be very different to (and incompatible with)
Japanese because of the more freedom in ko fights. Also, despite being a
territory ruleset, it could still need the "limited stone supply" rule as
well, to be safe (unlike J89) in a more formalized confirmation scheme
(based on claims of capturability, without enabling a new stone). And it
also seems advisable to drop "seki" and "no suicide".
Regards
József
dnc(AT)fw(DOT)hu
PS: Can someone confirm these:
A B C D E F G
.---------------.
4 | O O O O O O O |
3 | . . O # # O . |
2 | O O # . . # # |
1 | # # # # # # # |
`---------------'
This kind of position is classified as "double ko seki with double ko
stones" in chinese ruletext. It is also a variation of round-robin ko. It
seems to be stable (seki) in normal go, but also allows either player to
force a draw on repetition if he wants. It seems to collapse in superko
(both kinds).
A B C D E F G H I J
.---------------------.
4 | . O O O O # # # # . |
3 | O # O . O # . # O # |
2 | # # # O O # # O O O |
1 | . # . # O # O . O . |
`---------------------'
Two double ko sekis, with similar characteristics (stable, but forcing
repetition is also possible) in normal go. Under superko, the situation
seems to be more complex. There is an obvious and simple line of collapse,
but there are also some interesting sidelines. It is important whether the
sekis are independent or dependent like above, and also whether they are of
the same size. It seems to collapse under certain circumstances, and be
stable under other ones. Could any superko fan give a complete analysis?
Robert Jasiek
Makrai Jozsef wrote:
> artificial concepts [...] "wholeboard repetition"
If you call this artificial, then why is a whole-board basic ko
repetition not artificial? ;)
> E2. Honor traditional (e.g. send2-return1, moonshine) rulings as long as
> possible.
Do you want to become Mr. Ing-2? :)
> E4. Allow the game to progress and eventually end and score even with a
> noncooperative opponent
Interesting.
> A B C D E F G
> .---------------.
> 4 | O O O O O O O |
> 3 | . . O # # O . |
> 2 | O O # . . # # |
> 1 | # # # # # # # |
> `---------------'
> It seems to collapse in superko (both kinds).
See my webpages.
--
robert jasiek
Makrai Jozsef
> > artificial concepts [...] "wholeboard repetition"
>
> If you call this artificial, then why is a whole-board basic ko
> repetition not artificial? ;)
I believe we had a dispute over this a year ago. ;o) The other basics of ko
(strings, liberties, capture) are simple and LOCAL concepts, easy to apply
even for young children. Basic ko also has a valid LOCAL meaning (single
stone captures are hot).
1 For humans, superko obscures what should be the most clear: which moves
are legal.
2 For computers, it prevents proper analysis of the trees because two
identical board positions are not identical anymore (no transpositions).
> > E2. Honor traditional (e.g. send2-return1, moonshine) rulings as long
> > as possible.
> Do you want to become Mr. Ing-2? :)
=:) Well I don't know enough about his rules. But, if you look at my rules,
they consists of two simple and clear clauses (about limited stone supply
and ko-locking-passes). I think his rules may be a bit longer and more
obscure. ;o)
József
dnc AT fw DOT hu
Robert Jasiek
Makrai Jozsef wrote:
> if you look at my rules,
> they consists of two simple and clear clauses
What about full rules and their application to many examples?
--
robert jasiek
Planar
"Makrai Jozsef" <nos...@fw.hu> wrote:
> I believe we had a dispute over this a year ago. ;o) The other basics of ko
> (strings, liberties, capture) are simple and LOCAL concepts, easy to apply
> even for young children. Basic ko also has a valid LOCAL meaning (single
> stone captures are hot).
>
> 1 For humans, superko obscures what should be the most clear: which moves
> are legal.
I used to share that point of view. However, if you say that
whole board repetition is a drawn game, then the players still have
to be able to tell whether they are repeating the board position,
so you have the same problem as superko rules.
Moreover, as you said yourself, things are rarely purely local in
this game, players have to think global all the time.
> 2 For computers, it prevents proper analysis of the trees because two
> identical board positions are not identical anymore (no transpositions).
I don't think we should try to design rule sets that make it easier
for computers to play go: IMO the fact that it's hard for computers
is a good thing.
--
Planar
Makrai Jozsef
> > I believe we had a dispute over this a year ago. ;o) The other basics
> > of ko
> > (strings, liberties, capture) are simple and LOCAL concepts, easy to
> > apply
> > even for young children. Basic ko also has a valid LOCAL meaning
> > (single stone captures are hot).
> >
> > 1 For humans, superko obscures what should be the most clear: which
> > moves are legal.
>
> I used to share that point of view. However, if you say that
> whole board repetition is a drawn game, then the players still have
> to be able to tell whether they are repeating the board position,
> so you have the same problem as superko rules.
I think you are wrong. They don't have to be able to tell EXACTLY what is
the FIRST move that starts repetition, and they won't lose the game if they
misread that. It's enough if they can slowly recognize after a few cycles
that the game not really goes anywhere. Even then, they don't have to think
about wholeboard repetition: a rough understanding that none of them can
win anymore is sufficient for draw on agreement.
> Moreover, as you said yourself, things are rarely purely local in
> this game, players have to think global all the time.
Right! But it's the strategy that is global, not the rules IMHO.
> > 2 For computers, it prevents proper analysis of the trees because two
> > identical board positions are not identical anymore (no
> > transpositions).
>
> I don't think we should try to design rule sets that make it easier
> for computers to play go: IMO the fact that it's hard for computers
> is a good thing.
You won't lose this fact by superko-free rules. The factors that make it
hard are within the game itself: big board, many legal moves, etc. I would
never propose to take out anything from the game just to make it easier for
computers. But if we add something artificial, that better be comfortable,
at least for one of the two classes of players. ;o)
Robert Jasiek
Makrai Jozsef wrote:
> They don't have to be able to tell EXACTLY what is
> the FIRST move that starts repetition, and they won't lose the game if they
> misread that. It's enough if they can slowly recognize after a few cycles
> that the game not really goes anywhere. Even then, they don't have to think
> about wholeboard repetition: a rough understanding that none of them can
> win anymore is sufficient for draw on agreement.
How stupid do you suppose the players to be?
--
robert jasiek
Barry Phease
> About D3: I have a strong preference for true cycles as draws. First, AFAIK
> all three professional rulesets (Japanese, Chinese and Korean) practically
> treat them as such. Second, common sense and the generic conceptions of
> board games dictate this ruling as well: when neither side can achieve a
> won terminal position, there is a draw.
This is your preference. There are equally good reasons for avoiding
draws. Not enough is known about the game. Better players and new joseki
might make such positions more common and draws could be the norm for 12
dans (who might be able to achieve wins with superko). :)
> The rest of the game will then be
> played with modified (restricted) ko rules.
I personally don't think that having different rules for a resolutions
phase of the game is viable. The effects are too difficult to verify.
For instance the common idea is that the normal game can be played under
territory rules, but after 2 passes they can play on using pass stones to
resolve disputes. The practical upshot is that a player can delay
playing one-sided dame until the disputes phase to gain points.
> PS: Can someone confirm these:
>
> A B C D E F G
> .---------------.
> 4 | O O O O O O O |
> 3 | . . O # # O . |
> 2 | O O # . . # # |
> 1 | # # # # # # # |
> `---------------'
>
> This kind of position is classified as "double ko seki with double ko
> stones" in chinese ruletext.
No! The Chinese rules do not classify this. Ing rules call this a
disturbing ko seki. In Chinese rules it is probably a drawn game.
>It is also a variation of round-robin ko.
> It seems to be stable (seki) in normal go, but also allows either player
> to force a draw on repetition if he wants. It seems to collapse in
> superko (both kinds).
What do you mean "collapse"? Under superko it is the same as a double ko
seki. Either side can use it as ko threats.
--
Barry Phease
Makrai Jozsef
> > About D3: I have a strong preference for true cycles as draws. First,
> > AFAIK
> > all three professional rulesets (Japanese, Chinese and Korean)
> > practically
> > treat them as such. Second, common sense and the generic conceptions of
> > board games dictate this ruling as well: when neither side can achieve
> > a won terminal position, there is a draw.
>
> This is your preference. There are equally good reasons for avoiding
> draws. Not enough is known about the game. Better players and new
> joseki
> might make such positions more common and draws could be the norm for 12
> dans (who might be able to achieve wins with superko). :)
Well, if by 12 dans you mean perfect players, it seems obvious that - with
proper komi - the result of the normal game is either a draw on repetition
or jigo. A superko game is similar in this respect, no win possible. ;o)
Also I have no problems with someone preferring to avoid draws at any cost.
I just wanted to create a ruleset that is able to produce "correct"
rulings - by my personal preference. I don't know of any other ruleset
that would satisfy me: the best seem to be LJRG for normal territory (at
least until Robert fixes his flaws in J2003 ;o) ), Spight for superko
(fixes about one half of superko errors ;o), and ... what about normal
area? Chinese is far from logical. Also, Japanese (compatible) rules
are not really "normal" territory rules because of localization (bent4
etc.), so there seems enough space for my rules in territory mode as
well. ;o)
> > The rest of the game will then be
> > played with modified (restricted) ko rules.
>
> I personally don't think that having different rules for a resolutions
> phase of the game is viable. The effects are too difficult to verify.
Yes, I absolutely agree that they are potential problems (and represent
disharmony in themselves). However, I hope that my ko-locking-pass rules
are one of the safests in this respect. If you check them, they seem to
give the least possible advantage for either player, probably only
affecting lopsided kos. Any auto-locking scheme (J89 pass-for-ko for
example) has much more drastical effects.
Also, I hope that - similarly to the "limited stone supply" rule - this
rule is almost never actually applied in practice. Unless someone is forced
to pick up a moonshine life of course.
> > This kind of position is classified as "double ko seki with double ko
> > stones" in chinese ruletext.
>
> No! The Chinese rules do not classify this. Ing rules call this a
> disturbing ko seki. In Chinese rules it is probably a drawn game.
Hmmm. I got the impression in this version of the text:
http://www-2.cs.cmu.edu/~wjh/go/rules/Chinese.html
(Diagram 7 at the bottom - seems related)
> > It is also a variation of round-robin ko.
> > It seems to be stable (seki) in normal go, but also allows either
> > player
> > to force a draw on repetition if he wants. It seems to collapse in
> > superko (both kinds).
>
> What do you mean "collapse"? Under superko it is the same as a double ko
> seki. Either side can use it as ko threats.
By "collapse" I mean whoever plays first in it kills the other. Not like
regular double ko seki which is stable even in superko.
A B C D E F G
.---------------.
4 | O O O O O O O |
3 | . . O # # O . |
2 | O O # . . # # |
1 | # # # # # # # |
`---------------'
1.bA3 2.wE2 (2.wB3 3.bG3) 3.bB3 4.wD2 5.bD3 6.wB2 7.bE3, and 8.wA2
forbidden (both in SSK ad PSK) - am I wrong?
(Basically, most multiple kos seem to collapse in superko,
stable or not - which is no wonder, of course.)
BTW, I think generally neither side can use a double ko seki as ko threats
in superko, that only works in normal go, right?
Barry Phease
> Barry Phease wrote:
>> > This kind of position is classified as "double ko seki with double ko
>> > stones" in chinese ruletext.
>>
>> No! The Chinese rules do not classify this. Ing rules call this a
>> disturbing ko seki. In Chinese rules it is probably a drawn game.
>
> Hmmm. I got the impression in this version of the text:
> http://www-2.cs.cmu.edu/~wjh/go/rules/Chinese.html
> (Diagram 7 at the bottom - seems related)
Yes, but these are only examples, not rules. the Chinese rules are
confusing. In one clause they sound like superko rules, forbidding
repetition. In another clause they say that the referee can declare a
draw in case of repetition. This would probably happen here.
>> What do you mean "collapse"? Under superko it is the same as a double
>> ko seki. Either side can use it as ko threats.
>
> By "collapse" I mean whoever plays first in it kills the other. Not like
> regular double ko seki which is stable even in superko.
>
> A B C D E F G
> .---------------.
> 4 | O O O O O O O |
> 3 | . . O # # O . |
> 2 | O O # . . # # |
> 1 | # # # # # # # |
> `---------------'
>
> 1.bA3 2.wE2 (2.wB3 3.bG3) 3.bB3 4.wD2 5.bD3 6.wB2 7.bE3, and 8.wA2
> forbidden (both in SSK ad PSK) - am I wrong?
So w has to make a ko threat before continuing. If b keeps on going then
it reaches a point where b is forbidden. It can be fought like normal ko,
but if you start it you risk losing it.
>
> (Basically, most multiple kos seem to collapse in superko, stable or not
> - which is no wonder, of course.)
>
> BTW, I think generally neither side can use a double ko seki as ko
> threats in superko, that only works in normal go, right?
As we see from the above example there are often options. In general
though a ko complex can be fought like a normal ko (eg ko plus double ko
seki) or it looks like double ko with neither gaining from fighting it in
isolation.
Douglas Ridgway
> E4. Allow the game to progress and eventually end and score even with a
> noncooperative opponent (live stone remover, false disputer, etc.)
This requirement in particular seems quite tough. One possibility is
to use a variant of no-pass go, eg a move is a board play or returning
a prisoner, suicide prohibited, passing without returning a prisoner
prohibited, a player with no legal move loses. Game result is
equivalent to stone scoring. Good play by the winner can prevent the
board from filling more than about once. You also need to pay
attention to the time rules.
Filling the board is a little dull, though. Better to play with
cooperative opponents.
doug.
Makrai Jozsef
> > if you look at my rules,
> > they consists of two simple and clear clauses
>
> What about full rules and their application to many examples?
I intend to put up a page with the full text (area as well as territory)
soon. Examples will be made as circumstances and time makes it possible. ;)
Makrai Jozsef
If my opinion is wrong that's probably because I'm a bit stupid myself, and
judging others by this standard. ;o) Answering your question: it likely
depends on who the players are: programs on a pc, 6 year old kids, adult
beginners, average amateurs, or professional players. Speaking of
professionals, hereis a funny quote from their official rules text: ;o)
"Commentary on Article 12, clause 2:
In consideration of the difficulty of checking the repetition cycle, the
game ends without result if both players agree"
BTW, I didn't say that regular players can not usually spot the first
repeating move if they are careful enough, just that they don't have to.
Also, it could make sense to try a (long) cycle one more time, trying to
deviate or break it. Or if it can not be broken, maybe one can sacrifice?
Where would be the most appropriate point for that? etc.
Makrai Jozsef
> >> No! The Chinese rules do not classify this. Ing rules call this a
> >> disturbing ko seki. In Chinese rules it is probably a drawn game.
> >
> > Hmmm. I got the impression in this version of the text:
> > http://www-2.cs.cmu.edu/~wjh/go/rules/Chinese.html
> > (Diagram 7 at the bottom - seems related)
>
> Yes, but these are only examples, not rules. the Chinese rules are
> confusing. In one clause they sound like superko rules, forbidding
> repetition. In another clause they say that the referee can declare a
> draw in case of repetition. This would probably happen here.
From my limited knowledge of Chinese rules, I suppose that they use
exceptional rulings such as:
- true cycles like triple ko as draws
- false cycles like send2-return1 forbidden
- moonshine life is dead
I except similar rulings from my rules, but without relying on precedents.
> > A B C D E F G
> > .---------------.
> > 4 | O O O O O O O |
> > 3 | . . O # # O . |
> > 2 | O O # . . # # |
> > 1 | # # # # # # # |
> > `---------------'
> >
> > 1.bA3 2.wE2 (2.wB3 3.bG3) 3.bB3 4.wD2 5.bD3 6.wB2 7.bE3, and 8.wA2
> > forbidden (both in SSK ad PSK) - am I wrong?
>
> So w has to make a ko threat before continuing. ...
> It can be fought like normal ko
Sure, so the final ruling about who will die here is not completely random.
But remember, we have started from a stable seki at the first place....
> If b keeps on going then
> it reaches a point where b is forbidden. It can be fought like normal
> ko, but if you start it you risk losing it.
Sorry, I still think you are slightly wrong here. IMO there's no "if" you
start it, also B has no option but to keep on going. Remember, whoever
starts can kill unless external threats are found - but going first means
one threat advantage. If you don't start it the risk is bigger - you will
be the first needing a threat.
Another way to look at it: The players started a difficult fight. Both of
them played by his best, finally reached a seki like above - an even
result. All this so far happened in harmony with the laws of go -
strength, shape, liberties, etc. Then, suddenly those laws are no longer
valid, the players are in a completely different environment. Like if
gravity would stop suddenly. ;o)
Robert Jasiek
Makrai Jozsef wrote:
> "Commentary on Article 12, clause 2:
> In consideration of the difficulty of checking the repetition cycle
Length of a cycle in moves:
sending-2-returning-1: 4
eternal life: 4
triple ko: 6
double ko seki: 8
basic round-robin ko: 8
This shows another example of the official commentary's low quality.
--
robert jasiek
Barry Phease
> Barry Phease wrote:
>> So w has to make a ko threat before continuing. ...
>> It can be fought like normal ko
>
> Sure, so the final ruling about who will die here is not completely random.
> But remember, we have started from a stable seki at the first place....
Nothing is "stable" in go until the end. Think of this as being similar
to a mannen ko. In most cases it will be left as a seki, but either
player can choose to fight it.
If you consider Japanese rules this position will result in a no-result
game. However a (normal) double ko seki is not particularly stable
either. If you have a double ko seki, and one player gets a lead. The
other player only has to create a ko to threaten a triple ko which will
result in a no-result. This means that when you get a lead, you have to
fight with one hand tied and possibly accept unfavourable local results to
avoid ko.
>
>> If b keeps on going then
>> it reaches a point where b is forbidden. It can be fought like normal
>> ko, but if you start it you risk losing it.
>
> Sorry, I still think you are slightly wrong here. IMO there's no "if"
> you start it, also B has no option but to keep on going.
No1 You can always choose to play away (leaving it in the quasi-stable
position). Of course you have to keep an eye on ko the balance of ko
threats. This is true of all kos to a lesser extent.
>Remember,
> whoever starts can kill unless external threats are found - but going
> first means one threat advantage. If you don't start it the risk is
> bigger - you will be the first needing a threat.
This is an amateurish error. In ANY ko you can choose not to fight it if
the balance of ko threats is against you, or if the ambient temperature is
high enough.
>
> Another way to look at it: The players started a difficult fight. Both
> of them played by his best, finally reached a seki like above - an even
> result. All this so far happened in harmony with the laws of go -
> strength, shape, liberties, etc. Then, suddenly those laws are no longer
> valid, the players are in a completely different environment. Like if
> gravity would stop suddenly. ;o)
You expect go to be too easy. Perhaps we should ban snapback for you too.
Makrai Jozsef
> If you consider Japanese rules this position will result in a no-result
> game. However a (normal) double ko seki is not particularly stable
> either. If you have a double ko seki, and one player gets a lead. The
> other player only has to create a ko to threaten a triple ko which will
> result in a no-result. This means that when you get a lead, you have to
> fight with one hand tied and possibly accept unfavourable local results
> to avoid ko.
;o) It was me bringing up this argument last year, when we analyzed what
would have been the impact for allowig send2-return1 in area rules. ;o) But
I'm not sure it supports the claim that a true cycle must be forbidden at
all costs...
> > whoever starts can kill unless external threats are found - but going
> > first means one threat advantage. If you don't start it the risk is
> > bigger - you will be the first needing a threat.
>
> This is an amateurish error. In ANY ko you can choose not to fight it if
> the balance of ko threats is against you
I may of course be completely wrong, but in my experience it is often good
to take a ko first (if it can be taken freely) even if one does not intend
to fight it (provided there are no exceptionally big moves elsewhere, OC).
This at least makes the opponent waste a threat. Not fighting a ko would
mean (to me) more like a situation where the opponent took it already, and
I don't care to find a threat, leaving it as is.
> > Another way to look at it: The players started a difficult fight. Both
> > of them played by his best, finally reached a seki like above - an even
> > result. All this so far happened in harmony with the laws of go -
> > strength, shape, liberties, etc. Then, suddenly those laws are no
> > longer
> > valid, the players are in a completely different environment. Like if
> > gravity would stop suddenly. ;o)
>
> You expect go to be too easy. Perhaps we should ban snapback for you too
Why? Is it against the fundamental laws of go in any sense? ;o)
Planar
"Makrai Jozsef" <nos...@fw.hu> wrote:
> Barry Phease wrote:
> > You expect go to be too easy. Perhaps we should ban snapback for you too
I totally agree with Barry here, although I wouldn't have phrased
quite so tersely. I really don't understand why you are trying
to engineer the rules to mandate a particular result in the
(unusual) position that you are discussing. Is it because the
result of applying the rules (AGA rules for instance) is in
disagreement with your intuition ? That is not a good reason
to change the rules.
> Why? Is it against the fundamental laws of go in any sense? ;o)
What (besides the rules) do you call the fundamental laws of go ?
--
Planar
Makrai Jozsef
> I really don't understand why you are trying
> to engineer the rules to mandate a particular result in the
> (unusual) position that you are discussing. Is it because the
> result of applying the rules (AGA rules for instance) is in
> disagreement with your intuition ? That is not a good reason
> to change the rules.
Are you sure you are not quite unjust here? It's not ME proposing anything
particularly new (remember, my rules are more like a logical formulation of
the current Chinese rules than anything else in this sense). It's YOU
argueing for a radical new rule (superko) with its side effects and its
serious disagreement with both the Japanese and Chinese rules. Am I wrong?
Planar
"Makrai Jozsef" <nos...@fw.hu> wrote:
> Are you sure you are not quite unjust here? It's not ME proposing anything
> particularly new (remember, my rules are more like a logical formulation of
> the current Chinese rules than anything else in this sense). It's YOU
> argueing for a radical new rule (superko) with its side effects and its
> serious disagreement with both the Japanese and Chinese rules. Am I wrong?
Unless I am mistaken, Chinese rules have a superko rule and
also specify a draw in case of whole-board repetition. There's
no way you can find a logical formulation for that. So you get
to choose between superko and draws. But I have to ask, why
did they include a superko rule ? To me, it looks like "the
referee may declare a draw or a replay" is a leftover from some
draft of the rule set.
In other words, I think the spirit of Chinese rules is to include
a superko rule.
Anyway, if your only disagreement is with superko, I admit that
my point is much weaker. I was under the impression that you
wanted to ensure MSL is dead even with a double ko seki on board,
for example.
--
Planar
remove the dash from my address if you want to send me mail
Makrai Jozsef
>
> > Are you sure you are not quite unjust here? It's not ME proposing
> > anything
> > particularly new (remember, my rules are more like a logical
> > formulation of
> > the current Chinese rules than anything else in this sense). It's YOU
> > argueing for a radical new rule (superko) with its side effects and its
> > serious disagreement with both the Japanese and Chinese rules. Am I
> > wrong?
>
> Unless I am mistaken, Chinese rules have a superko rule and
> also specify a draw in case of whole-board repetition. There's
> no way you can find a logical formulation for that.
> like "the
> referee may declare a draw or a replay" is a leftover from some
> draft of the rule set.
It is quite unfortune that I don't know of any detailed explanation of the
Chinese rules (except the short text I linked earlier). Maybe someone from
China can clarify them in this respect? But, until I know more, I still
think that they forbid send2-return1 (mentioned explicitly in the text),
draw on true cycles (mentioned explicitly in the text), and kill moonshine
(mentioned explicitly in the text). This also agrees with Japanese and
Korean rules, AFAIK. I doubt that this is anything new that could change
between two versions of a draft. To me, the text sounds more like it
include the "reappearance of the position" as an attempt to explain
the traditional rulings for the first and the last, and - seeing no better
choice - included the intended rulings for true cycles as well.
Please see my initial post for a logical formulation of these rulings,
without superko.
> Anyway, if your only disagreement is with superko, I admit that
> my point is much weaker. I was under the impression that you
> wanted to ensure MSL is dead even with a double ko seki on board,
Yes, my rules also include the pass-for-a-ko-lock rule in playing out
disputes, that achieves this too (although I'm not sure it's absolutely
important). But I was under the impression that you are referring to our
discussion with Barry, about the round-robin seki variant under superko,
and that you were wondering why I want to engineer rules that would fit my
intuition in this unusual position. ;o)
Erik
back. Did you see my rules?
(http://www.cs.unimaas.nl/~vanderwerf/5x5/migos_rules_v0.3.txt) I think
they accomplish what you like (Area scoring, msl=dead, bent-four=played
out, long cycle=draw unless one side captures more stones per cycle).
Best,
Erik
Planar
"Makrai Jozsef" <nos...@fw.hu> wrote:
> It is quite unfortune that I don't know of any detailed explanation of the
> Chinese rules (except the short text I linked earlier). Maybe someone from
> China can clarify them in this respect?
I'm not a specialist. All I know is what I read from Robert's
web page. It would be good to have more details indeed.
> Yes, my rules also include the pass-for-a-ko-lock rule in playing out
> disputes, that achieves this too (although I'm not sure it's absolutely
> important).
In my opinion, it's not worth adding even a comma to the rules in
order to make sure MSL is dead in all cases.
> But I was under the impression that you are referring to our
> discussion with Barry, about the round-robin seki variant under superko,
> and that you were wondering why I want to engineer rules that would fit my
> intuition in this unusual position. ;o)
That, too. I think it's strange that you would give this unusual
position as a justification for your rules. So what if my intuition
disagrees with the rules ? That's what it does all the time anyway:
some groups that should live (according to my intuition) end up dead,
invasions that should die will live, useless reinforcements turn
out to be necessary, etc, etc.
I don't think our intuition is a good guide of what the rules
should do in strange positions. If you use it that way, you'll
fall in the same trap as Japanese rules: you'll try to localize
everything, making the game simpler and the rules more complex,
which is clearly in disagreement with the spirit of the game.
Makrai Jozsef
> > But I was under the impression that you are referring to our
> > discussion with Barry, about the round-robin seki variant under
> > superko,
> > and that you were wondering why I want to engineer rules that would fit
> > my intuition in this unusual position. ;o)
>
> That, too. I think it's strange that you would give this unusual
> position as a justification for your rules. So what if my intuition
> disagrees with the rules ?
But please don't forget: it's not something just about my intuition, it's
how the game is played in China, Japan and Korea. Also, while Japanese
rules had several rulesets that try to formulate them logically (LJRG is a
good example, at least without the optional parts about cycles and initial
setup - or look ar Robert's recent work), Chinese rules had none so far, at
least that I would know of. Unless some serious flaw will be found in my
rules, they may be a good candidate for this task.
Makrai Jozsef
Thank you, and yes I saw your work and the rules. (Although I'm not sure
about the thread - could you give a Google link?) My initial goal was
to create a _reference_ ruleset, that is very simple, suitable for _both_
humans and computers. I sincerely hope that I fulfilled that goal. ;o)
It is certain that if I ever get to write the simple solver I thought about
more than a year ago, I will also need optimizations and sophisticated
knowledge-based internal rules similar to yours. Directly putting my rules
into code wouldn't do good for tree sizes. =:) But I'm not so keen on
writing it - what's the point now that you did 5*5? ;o)
Robert Jasiek
Makrai Jozsef wrote:
> a radical new rule (superko) with its side effects and its
> serious disagreement with both the Japanese and Chinese rules.
It is unknown whether superko is a 20th century invention or was
known long ago.
It is not radical since it leads to different games in less than
1 of more than 5,000 or maybe 20,000 games.
--
robert jasiek
Robert Jasiek
Makrai Jozsef wrote:
> Chinese rules had none so far
You have overlooked the thread Simplified Chinese Rules, haven't you?
--
robert jasiek
Makrai Jozsef
> > Chinese rules had none so far
>
> You have overlooked the thread Simplified Chinese Rules, haven't you?
No, and sorry Robert, I didn't really understand why you called them
Simplified Chinese rules. They are - as far as I remember - simplified area
rules with positional (yuck) superko (probably more similar to western
rules such as AGA or NZ).
Unless I'm mistaken, Chinese rules draw on true cycles. The text is
confusing of course, but in Chapter 3. Section 20. (about the intended
meaning of the "reappearance of the same position"), even in false cycles
such as send2-return1 it speaks of not absolutely prohibiting the first
repeating move, but "a player cannot refuse to end the game by reason of
the position in Diagram 8 or any other similar positions". (What a
definition! ;o) ) To me, this probably means that a few cycles may be
played if desired, only perpetual repetition and preventing game end is
prohibited.
It's so sad I only have that short text with it's contradicting
clauses - anybody knows more?
Robert Jasiek
Makrai Jozsef wrote:
> > You have overlooked the thread Simplified Chinese Rules, haven't you?
> No, and sorry Robert, I didn't really understand why you called them
> Simplified Chinese rules. They are - as far as I remember - simplified area
> rules with positional (yuck) superko (probably more similar to western
> rules such as AGA or NZ).
AGA or NZ use situational superko. Chinese use positional superko.
The main part of the Chinese 1988 Rules specify the positional
superko. Then later in the referees section this basic principle
is modified by exceptional rulings. So if we SIMPLIFY, then we
remove the exceptions and not the main rule, right?! Since I
have done so, I can also call the resulting simplification
Simplified.
If instead the purpose would be to model modern Chinese ko
tradition closely but generally, then presumably we should
steal the long cycle and the ko-pass rules from version 19 of
the Japanese 2003 Rules;)
--
robert jasiek
Makrai Jozsef
> have done so, I can also call the resulting simplification
> Simplified.
I had no problems with "Simplified". It is the word "Chinese" I found
confusing, especially in the recent context "a logical formulation of
current Chinese rules". ;o)
Barry Phease
>
> It's so sad I only have that short text with it's contradicting
> clauses - anybody knows more?
The Chinese text is no better. :)
Erik
> Erik wrote:
>
>
>>This thread seems quite similar to a discussion we had a few months
>>back. Did you see my rules?
>>(http://www.cs.unimaas.nl/~vanderwerf/5x5/migos_rules_v0.3.txt) I think
>>they accomplish what you like (Area scoring, msl=dead, bent-four=played
>>out, long cycle=draw unless one side captures more stones per cycle).
>
>
> Thank you, and yes I saw your work and the rules. (Although I'm not sure
> about the thread - could you give a Google link?)
It's not so important, but what I had in mind popped up in some threads
about (super) ko rules, somewhere around may/june this year. (look fo
threads like: "reason for complicated ko-rule", "and again the ko rule",
and "Migos rules")
> My initial goal was
> to create a _reference_ ruleset, that is very simple, suitable for _both_
> humans and computers. I sincerely hope that I fulfilled that goal. ;o)
I took a brief look at it. It seems promising. Still some
comment/questions, though.
1) Suicide should be illegal to stay closer to the Chinese rules (and
prevent abuse).
2) Ko
"Capturing a single stone in a way that the new stone doesn't
immediately become part of a bigger string is a ko capture. The opponent
cannot immediately recapture, placing a stone back to where his original
stone were captured from."
In the position below this would make a move at "a" a ko-capture!
. # . .
# O a .
. # . .
To me this seems to be a bit strange (though quite harmless).
In this case however...
. # # O .
# O O a O
. # # O .
3) I'm not sure but I think there may be a problem if you allow players
to agree on a draw at anytime in a tournament. If I remember correctly
some tournament rules require approval by a referee (to prevent abuse).
4) Could you give an analysis of what happens if the following three 5x5
positions are played/scored by your rules?
1:
. # # # #
# # # # #
# # # # #
# # # # #
# # # # #
2:
# # # # #
# . # # #
# # # # #
# # # # #
# # # # #
3:
# # # # #
# # # # #
# # . # #
# # # # #
# # # # #
(Hint: in all cases the black stones can be obviously captured, so if
White just passes he could claim the whole board in all 3 cases.
However, if play would continue (by an actual capture) 1 would become
completely black, 2 would become mixed, and 3 would become completely
white.)
5) An analysis of hane-seki would also be interesting.
> ...what's the point now that you did 5*5? ;o)
Well, at least one person still has doubts about my solution so it would
be nice if others could confirm it :-) Furthermore, 6x6 and 7x7 are
still unsolved, and tsume-go is always an option...
Best,
Erik
Makrai Jozsef
> It's not so important, but what I had in mind popped up in some threads
> about (super) ko rules, somewhere around may/june this year. (look fo
> threads like: "reason for complicated ko-rule", "and again the ko rule",
> and "Migos rules")
Aha, thank you. Superko looks surprisingly popular. ;o)
> 1) Suicide should be illegal to stay closer to the Chinese rules (and
> prevent abuse).
For compatibility, you are obviously right. For abuse, I'm not sure.
Suicide seems to be covered nicely by finite stone supply. And maybe taking
away this option opens some doors for abuse in itself, such as in some
positions similar to yours. Also, my intuition advices strongly against
forbidding suicide. OTOH, it's devastating for a solver. ;o) Maybe no
suicide could be an optional part of the rules. But I'm not sure of its
consequences...
> 2) Ko
> "Capturing a single stone in a way that the new stone doesn't
> immediately become part of a bigger string is a ko capture. The opponent
> cannot immediately recapture, placing a stone back to where his original
> stone were captured from."
>
> In the position below this would make a move at "a" a ko-capture!
>
> . # . .
> # O a .
> . # . .
>
> To me this seems to be a bit strange (though quite harmless).
>
> In this case however...
>
> . # # O .
> # O O a O
> . # # O .
Here <a> is not a ko capture since it did not capture a single stone. W
recapturing at the left would be, though.
This is a very interesting question, I have pondered over it myself before
deciding on the definition. What is a ko capture?
A B C D E F
.-------------.
5 | . . . # . # |
4 | O # # O # # |
3 | O O O . O O |
2 | . . O . O . |
1 | . . O . O . |
`-------------'
Black plays bD3. Is this a ko capture? Normally this is not a question
since W can not recapture anyway. But it can be interesting in a Japanese89
confirmation sequence played with pass-for-kos:
A B C D E F
.-------------.
5 | . . . # . # |
4 | O # # O # # |
3 | O O O . O O |
2 | . . O . O O |
1 | . . O # # # |
`-------------'
1.bD3 2.wD2 3.bB5. Can white play 4.wD4 or he must pass for a "ko" before?
In rule texts there is surprisingly little definition of what a ko is. The
text of Chinese rules is as generous as ever, maybe didn't even attempt a
definition. Western rules (with superko) don't deal with "ko captures" as
such. Japanese ruletext say a ko is a "shape in which the players can
alternately capture and recapture one opposing stone". By this definition
1.bD3 is not a ko capture (and 4.wD4 is legal), which also seems more
reasonable to me. OTOH, there are also some other texts (not just mine)
which define ko captures as certain kinds of single stone captures... OC,
this is again with no practical consequences, since they don't use
pass-for-ko. But I'm still curious. What is a ko capture, after all? =:)
> 3) I'm not sure but I think there may be a problem if you allow players
> to agree on a draw at anytime in a tournament. If I remember correctly
> some tournament rules require approval by a referee (to prevent abuse).
Yes, the problem is similar in chess too. But I think any attempt for
"objective" checking or "referee" approval is doubtful, since a jigo or
repetition can be constructed if necessary (there was even a joseki, that
leads to chosei). It seems inavoidable part of board game(s). When two
persons play, they can fake a draw if they want to... Even in Othello! ;o)
> 4) Could you give an analysis of what happens if the following three 5x5
> positions are played/scored by your rules?
You didn't mention whether in area or territory mode. Also, it seems that
these positions are not directly related to my rules since neither finite
stone supply nor virtual kos have any role. So I try to look at them from a
more generic normal go with area / territory viewpoint. Let MJA be area and
MJT be the territory mode.
> 1:
> . # # # #
> # # # # #
> # # # # #
> # # # # #
> # # # # #
There seem to be several cases here.
1. If B agrees his stones are dead, this (and the other positions too) is a
draw in MJA since no players surrounded any territory by alive stones, and
there are no alive stones of either player. This is also true for MJT since
B dead stones are of no use if they are not in W territory. If there is an
alive W group somewhere (case B), the whole board area score would favor
white, and dead stones in territory would increase the MJT score as well.
2. B doesn't agree on his stones being dead (main line).
2.1. If W accepts them being alive, B+25 in area mode and B+1 in territory
mode (all the three positions).
But he will not:
2.2. MJA: W captures and lose by area (in case B above this is not
certain); B+25.
2.2. MJT:
1. W can try a claim and confirmation. This doesn't work since capturing B
enables an alive B stone in the center.
2. W can resume and capture. B plays the center. Game stops. Result depends
on history (prisoner count). If there are no extra prisoners, W has 24
prisoners and no territory for a score of 24. B has 1 prisoner (the dead W
stone) and 24 pts territory. B+1.
> 2:
> # # # # #
> # . # # #
> # # # # #
> # # # # #
> # # # # #
2.2. MJA: W needs to capture and play goes on to an even position.
2.2. MJT:
1. Claim still don't work since B can live after capture.
2. W resumes, captures, and achieves an even position. But W wins because
of prisoners.
> 3:
> # # # # #
> # # # # #
> # # . # #
> # # # # #
> # # # # #
2.2. MJA: W needs to capture for the area score W+25.
2.2. MJT:
1. Claim works since B can not live after capture. But it's useless
since W did not surround anything with alive stones.
2. So W continues, captures and achieves a full board win. Score is W+48.
> (Hint: in all cases the black stones can be obviously captured, so if
Well, not "obviously", which is why you raise the question. ;o) Maybe a
more clear wording of the text would be appropriate. My intention was to
define dead stones in a way that they are dead only if both players agree.
> White just passes he could claim the whole board in all 3 cases.
No, since he has not surrounded it (with alive stones).
> 5) An analysis of hane-seki would also be interesting.
Yes. I plan to analyse all official Japanese examples, hane-seki amongst
them...
> > ...what's the point now that you did 5*5? ;o)
>
> Well, at least one person still has doubts about my solution so it would
> be nice if others could confirm it :-) Furthermore, 6x6 and 7x7 are
> still unsolved, and tsume-go is always an option...
Maybe I'll start with a solver which is based directly on my rules (just
for fun). ;o) I doubt it will be fast enough for even 4x4, though. ;) Quite
long lines seem to be possible, if everything is to be played out... Even
without suicide and with knowledge about unconditional life, sure eyes and
no playing inside of them by either player...
Robert Jasiek
Makrai Jozsef wrote:
> What is a ko capture?
During the previous years I have found about half a dozen valid
and dozens of wrong definitions:) Here is a valid one:
>>
A ko consists of two adjacent intersections so that a play by one
player on one of the intersections followed by a play of the
opponent on the other intersection would recreate the position.
A ko-capture is a play in a ko so that it is still a ko.
<<
Shamelessly cited from the Japanese 2003 Rules:) Of course, the
definition is not needed for the basic ko rule, which relies on
"position". Rather it is convenient for ko-pass rules...
> In rule texts there is surprisingly little definition of what a ko is.
Read proper rule texts! :(
> But I'm still curious. What is a ko capture, after all? =:)
See above.
> Yes. I plan to analyse all official Japanese examples, hane-seki amongst
> them...
Better wait for my example pages that contain more standard
test examples:) (On request, I send you the drafts by email.)
--
robert jasiek
Makrai Jozsef
> > What is a ko capture?
>
> During the previous years I have found about half a dozen valid
> and dozens of wrong definitions:) Here is a valid one:
Verifying the validity of a definition is only possible if we know what we
are trying to define. ;o) But this is where the my doubts are: is a
non-recapturable single stone capture (with the new stone remaining a
single stone) a ko capture or not? Probably not, but I'm still not sure. :)
> >>
> A ko consists of two adjacent intersections so that a play by one
> player on one of the intersections followed by a play of the
> opponent on the other intersection would recreate the position.
>
> A ko-capture is a play in a ko so that it is still a ko.
> <<
You can guess that no stubborn enemy of superko (like myself) will be happy
with a definition of basic ko referencing "recreate the position". ;o)
> > Yes. I plan to analyse all official Japanese examples, hane-seki
> > amongst them...
>
> Better wait for my example pages that contain more standard
> test examples:) (On request, I send you the drafts by email.)
Thank you, I would be grateful.
Although I don't know if I will have enough time for all the work... :-|
Actually, I also have some notes and examples, beasts, flaws in existing
ruletexts etc. collected which also needs to be HTML-ized... I planned
another page from it (leaving the current one containing only the rule
text).
Robert Jasiek
Makrai Jozsef wrote:
> But this is where the my doubts are: is a
> non-recapturable single stone capture (with the new stone remaining a
> single stone) a ko capture or not? Probably not, but I'm still not sure. :)
You problem is not not to know what a ko-capture is - there are
(can be) legal and illegal ko-captures!
> You can guess that no stubborn enemy of superko (like myself) will be happy
> with a definition of basic ko referencing "recreate the position". ;o)
For that reason I refer to position ;(
--
robert jasiek
Erik
> Erik wrote:
>>prevent abuse).
>
>
> For compatibility, you are obviously right. For abuse, I'm not sure.
> Suicide seems to be covered nicely by finite stone supply.
Yes, the finite stone supply helps in principle. However, with this type
of abuse it may still take quite some time before the stone difference
is large enough. Especially under time constraints this might be a problem.
And maybe taking
> away this option opens some doors for abuse in itself, such as in some
> positions similar to yours. Also, my intuition advices strongly against
> forbidding suicide.
Your intuition is at odds with both the Chinese and Japanese rules.
OTOH, it's devastating for a solver. ;o)
Well, it's not that bad. Benson's algorithm can be modified to allow
suicide. The efficiency would decrease quite significantly but I think I
could still solve 5x5 without too much problems.
>>2) Ko
>>"Capturing a single stone in a way that the new stone doesn't
>>immediately become part of a bigger string is a ko capture. The opponent
>>cannot immediately recapture, placing a stone back to where his original
>>stone were captured from."
>>
>>In the position below this would make a move at "a" a ko-capture!
>>
>>. # . .
>># O a .
>>. # . .
>>
>>To me this seems to be a bit strange (though quite harmless).
>>
>>In this case however...
>>
>>. # # O .
>># O O a O
>>. # # O .
>
>
> Here <a> is not a ko capture since it did not capture a single stone. W
> recapturing at the left would be, though.
Ah yes, ofcourse, though this also seems to indicate that the original
text is somewhat confusing (at least to me :-)).
>
> This is a very interesting question, I have pondered over it myself before
> deciding on the definition. What is a ko capture?
Why do you focuss on defining ko capture? Is it not enough to use a rule
for dealing with repetition? I think if you really want to define ko
capture you should just stick to basic ko.
>...
>>(Hint: in all cases the black stones can be obviously captured, so if
>
>
> Well, not "obviously", which is why you raise the question. ;o) Maybe a
> more clear wording of the text would be appropriate. My intention was to
> define dead stones in a way that they are dead only if both players agree.
I had the impression that your rules where also meant to be 'computer
friendly'. If this is indeed the case relying on agreement is not a good
idea.
Good luck with your work.
Erik
Erik
> You can guess that no stubborn enemy of superko (like myself) will be happy
> with a definition of basic ko referencing "recreate the position". ;o)
Ok, then how about:
Capturing a single stone in a way that the new stone gets only one
liberty and does not immediately become part of a bigger string is a ko
capture.
Simon Goss
>A ko consists of two adjacent intersections so that a play by one player
>on one of the intersections followed by a play of the opponent on the
>other intersection would recreate the position.
>
>A ko-capture is a play in a ko so that it is still a ko.
Neat. But ...
>Shamelessly cited from the Japanese 2003 Rules:)
I think it's misleading and inappropriate to call your rules Japanese.
They aren't Japanese. Japanese-style maybe.
--
Simon
Makrai Jozsef
> >>1) Suicide should be illegal to stay closer to the Chinese rules (and
> >>prevent abuse).
> >
> > For compatibility, you are obviously right. For abuse, I'm not sure.
> > Suicide seems to be covered nicely by finite stone supply.
>
> Yes, the finite stone supply helps in principle. However, with this type
> of abuse it may still take quite some time before the stone difference
> is large enough. Especially under time constraints this might be a
I think an abuser can always waste a similar amount of time just by playing
dead stones in territory, wihout suicide. ;o)
>> OTOH, it's devastating for a solver. ;o)
>
> Well, it's not that bad. Benson's algorithm can be modified to allow
> suicide. The efficiency would decrease quite significantly but I think I
> could still solve 5x5 without too much problems.
In my nightmares I thought about sequences like:
.---------.
| a c e g |
| i k m o |
| p n l j |
| h f d b |
`---------'
At least with a dull solver.
> >>In this case however...
> >>
> >>. # # O .
> >># O O a O
> >>. # # O .
> >
> >
> > Here <a> is not a ko capture since it did not capture a single stone. W
> > recapturing at the left would be, though.
>
> Ah yes, ofcourse, though this also seems to indicate that the original
> text is somewhat confusing (at least to me :-)).
Also it's probably in poor English. ;o) For the future, polishing the text
will surely be a must. :)
> > This is a very interesting question, I have pondered over it myself
> > before deciding on the definition. What is a ko capture?
>
> Why do you focuss on defining ko capture? Is it not enough to use a rule
> for dealing with repetition? I think if you really want to define ko
> capture you should just stick to basic ko.
This is what I try. ;o) I guess the reason for going after "ko captures"
rather than "ko shapes" or repetitive positions is that I try to focus on
the rules directly generating legal moves.
> > Well, not "obviously", which is why you raise the question. ;o) Maybe a
> > more clear wording of the text would be appropriate. My intention was
> > to
> > define dead stones in a way that they are dead only if both players
> > agree.
>
> I had the impression that your rules where also meant to be 'computer
> friendly'. If this is indeed the case relying on agreement is not a good
> idea.
Being able to enumerate legal moves (and special actions like demanding
continuation) in all positions seems to be the first layer of
computer-friendliness. ;o) I don't think this is trivial with a
Chinese-style ruleset. :)
Then, I think a computer player should be able to perform similar actions
like a human. An example from chess would be offering draw or evaluating
draw offers. In go, "agreement" would probably mean that a solver treats
all strings alive, except dead stones in sure eyes of unconditionally alive
strings. Many other optimizations seem to be possible - as I mentioned my
rules was not meant to be put directly into code, but more like to be used
as a reference.
Makrai Jozsef
Nice! But as I tried to say to Robert, it's not the definition that seems
hard, but deciding whether the recapturable state is really a requirement
for a capture to be ko capture. Is it? It seems hard to create a test where
the two concepts behave differently, allowing objective judgement. Even my
earlier example (j89 confirmation) is useless because a ko pass would never
be required there: no matter if the earlier capture is a ko capture or not,
it's obviously not in the SAME ko (surrounding stones changed).
Robert Jasiek
Simon Goss wrote:
> I think it's misleading and inappropriate to call your rules Japanese.
> They aren't Japanese.
They are more Japanese rules than any other ruleset before AFA
contents is concerned. OC, they have not originated in Japan.
Instead one of their intentions is their future use in Japan
instead of the Japanese 1989 Rules (and if the Japanese go
associations really want to maintain all those exceptions).
To avoid the title being misleading, in my current draft on a
commentary the first information is as follows:
>>
The Japanese 2003 Rules are meant to explain the spirit of
Japanese style rules completely and logically.
The Japanese 2003 Rules are not official rules. They are
invented by Robert Jasiek.
The Japanese 2003 Rules have well defined terms, are logical,
and are complete. They preserve the spirit of the 1989 rules
of the Japanese Go Associations. The latter have badly defined
terms, are incomplete, and are illogical. Therefore it is
strongly recommended that they are replaced by the Japanese
2003 Rules.
<<
If it is not appropriate to call the best explanation of
Japanese style rules ever "Japanese", then no attribute is
appropriate.
--
robert jasiek
Ben Finney
> If it is not appropriate to call the best explanation of Japanese
> style rules ever "Japanese", then no attribute is appropriate.
You can justifiably be proud of this re-formulation of the Japanese
rules, but your definition of "best explanation of Japanese style rules"
may not match the definition of others -- e.g. the very people you wish
to convince. It may be best to avoid using unqualified descriptions
like "best" for this reason.
--
\ "I like to fill my bathtub up with water, then turn the shower |
`\ on and pretend I'm in a submarine that's been hit." -- Steven |
_o__) Wright |
Ben Finney <http://bignose.squidly.org/>
Robert Jasiek
Ben Finney wrote:
> but your definition of "best explanation of Japanese style rules"
> may not match the definition of others -- e.g. the very people you wish
> to convince. It may be best to avoid using unqualified descriptions
> like "best" for this reason.
Therefore "best" is not in the title:) Since you do not have a chance
yet to judge about the full commentary and the commentaries on
examples, you do not know yet why I can use that word in the draft of
the commentary:) Wait and be impressed later, only to criticise then
that I use ascii diagrams so that it will still be possible to read
or download the webpages at all;)
--
robert jasiek
Erik van der Werf
>>>OTOH, it's devastating for a solver. ;o)
>>
>>Well, it's not that bad. Benson's algorithm can be modified to allow
>>suicide. The efficiency would decrease quite significantly but I think I
>>could still solve 5x5 without too much problems.
>
>
> In my nightmares I thought about sequences like:
>
> .---------.
> | a c e g |
> | i k m o |
> | p n l j |
> | h f d b |
> `---------'
>
> At least with a dull solver.
Such lines are not a problem with simple alpha-beta pruning.
>>I had the impression that your rules where also meant to be 'computer
>>friendly'. If this is indeed the case relying on agreement is not a good
>>idea.
>
>
> Being able to enumerate legal moves (and special actions like demanding
> continuation) in all positions seems to be the first layer of
> computer-friendliness. ;o) I don't think this is trivial with a
> Chinese-style ruleset. :)
>
> Then, I think a computer player should be able to perform similar actions
> like a human.
I don't think this will happen in the near future. :-(
> An example from chess would be offering draw or evaluating
> draw offers. In go, "agreement" would probably mean that a solver treats
> all strings alive, except dead stones in sure eyes of unconditionally alive
> strings. Many other optimizations seem to be possible - as I mentioned my
> rules was not meant to be put directly into code, but more like to be used
> as a reference.
Sure, proposing something about life or death of stones in
human-computer games is always possible. However, in the case of
disagreement computers have a huge problem. For example, at least as far
as I know, there is no computer that could participate in the rules
discussions like the ones that take place in this forum. Furthermore,
agreement is not an option for a solver.
Best,
Erik
Makrai Jozsef
> > draw offers. In go, "agreement" would probably mean that a solver
> > treats
> > all strings alive, except dead stones in sure eyes of unconditionally
> > alive
> > strings. Many other optimizations seem to be possible - as I mentioned
> > my
> > rules was not meant to be put directly into code, but more like to be
> > used as a reference.
>
> Sure, proposing something about life or death of stones in
> human-computer games is always possible. However, in the case of
> disagreement computers have a huge problem. For example, at least as far
> as I know, there is no computer that could participate in the rules
> discussions like the ones that take place in this forum. Furthermore,
> agreement is not an option for a solver.
You surely know much more than me about this, but I see no need for
agreement with simple reductions like dead stones in sure eyes. (And in
case of more advanced reductions and human-computer disagreement the string
in question may simply be marked to be treated alive as long as it's on the
board, no?) I think the big problem with using my rules for a solver is
handling of moonshine and likes. If they are to be played out, that's a
huge loss and extra task for a solver. This is why I wrote that putting
them directly into code only seems possible for fun, and very small boards
(3x3 and maybe 4x4).
Bernd Gramlich
>> 1) Suicide should be illegal to stay closer to the Chinese rules
>> (and prevent abuse).
>
> For compatibility, you are obviously right. For abuse, I'm not sure.
> Suicide seems to be covered nicely by finite stone supply.
Usually, one does not collect prisoners in Chinese-style rules.
Therefore, 360 stones of each colour are enough even in absurd games.
--
Bernd Gramlich [bE6nd "gRamlIC]
Makrai Jozsef
> > For compatibility, you are obviously right. For abuse, I'm not sure.
> > Suicide seems to be covered nicely by finite stone supply.
>
> Usually, one does not collect prisoners in Chinese-style rules.
> Therefore, 360 stones of each colour are enough even in absurd games.
I referred to my rules posted here recently, which - despite claiming to be
Chinese-style - do collect prisoners. ;o)