Forum for Ko master

Path: <= CGT path =>

Definition question [#2187]

Back to forum     Back to page

New reply

 
reply
145.120.14.118: Definition question (2010-02-09 00:22) [#7185]

In the article, it says:

komaster, for the person who can win a ko (or superko) fight without having to ignore a ko threat

However, in both figures the komaster is ignoring (potential) ko threats.

First figure (black komaster), white 7 is ignored.

Second figure (white komaster), black 2 and 8 are ignored.

So what gives?

X
tapir: Re: Definition question (2010-02-09 01:10) [#7186]

threat is obviously assumed to be big enough to be answered. that is a threat which isn't big enough but gains some points as compensation isn't considered as a threat to be answered. the komaster really is only the person who would usually win the ko (according to overall threat situation). the explanation at komonster makes it kind of clear, where the komonster has further threats to decrease the potential compensation the koloser can gain by his ignored "threat" (which isn't termed threat here)

tapir: SL/Bill and Berlekamp treatment do not fit, or is the whole concept inconsistent? (2011-07-29 19:36) [#8644]

in fact, on a twentieth reading this doesn't ring true anymore. or i totally misunderstand.

in berlekamps terms:

"For each region of the game, a speci ed player is designated as the local Komaster." (Contradicts this page.)

"Komaster makes two moves consecutively, to and from the critical ko position, and the Komaster pays two taxes (at the same tax rate)" (w/ tax = value of play elsewhere, fits this page.)

That is, he assumes a komaster as the player who only pays the tax twice (with tax meaning the value of a move elsewhere, or not?). That is, there is no compensation for a player losing the ko other than two average (gote) moves elsewhere. In this meaning, a komaster seems to be a rare occurrence in real games (you need many ko threats or maybe infinite ko threats by means of double ko to achieve this), usually you get some compensation - your threat has some followup, not big enough to win you the ko, but enough to reduce the loss. However, at the same time it is assumed there always is a komaster for potential ko. But this very page contradicts, by saying "It is possible, of course, that a ko has neither komaster nor koloser." (Already included in version 10, probably by Bill himself.) In the original Berlekamp text this isn't possible. Go figure!

Thus, if go players try to understand ko in CGT terms, this results in a big mess. Because, as I understand it, there is no concept, term to describe the ordinary bread-and-butter ko, where one side wins the ko but has to give a bigger compensation than two independent gote moves. If anything, this compensation is hidden in the "auction" to decide the regional komaster (see Berlekamp), which may result in an higher tax during the ko than for moves elsewhere. However, this is a pretty unpleasing and well-hidden solution and will be usually lost if you read the derivation of this ideas on SL.

However, despite this lack of clarity, we have lines like "Position is hyperactive, value depends on who is komaster" all over SL. Thus, despite claiming on this page, that you don't necessarily have a komaster, we write lines taken straight from Berlekamp, who assumes there always is. This leads to a lot of confusion and I would be surprised if 1% of the readers understand them. I surely do not.

68.124.184.165: Re: SL/Bill and Berlekamp treatment do not fit, or is the whole concept inconsistent? (2011-07-30 03:44) [#8648]

tapir: However, at the same time it is assumed there always is a komaster for potential ko. But this very page contradicts, by saying "It is possible, of course, that a ko has neither komaster nor koloser." (Already included in version 10, probably by Bill himself.) In the original Berlekamp text this isn't possible. Go figure!

Bill: Berlekamp was concerned with evaluating non-placid kos. He had a breakthrough with the concept of komaster, which allowed him to do so. He never claimed that any given go position must have a komaster, only that if it did, he could evaluate the ko.

Later on I extended Berlekamp's work and Berlekamp came up with the idea of a neutral threat environment, so that now we can evaluate any ko or superko, whether there is a komaster or not.

68.124.184.165: Re: SL/Bill and Berlekamp treatment do not fit, or is the whole concept inconsistent? (2011-07-30 03:48) [#8649]

tapir: in berlekamps terms:

"For each region of the game, a speci ed player is designated as the local Komaster." (Contradicts this page.)

Bill: Berlekamp's original komaster idea was in terms of the whole board. However, he recognized that there might be whole board go positions where a different player would be komaster for different kos. That kind of situation is what he has in mind with that statement.

68.124.184.165: Re: SL/Bill and Berlekamp treatment do not fit, or is the whole concept inconsistent? (2011-07-30 03:52) [#8650]

tapir: as I understand it, there is no concept, term to describe the ordinary bread-and-butter ko, where one side wins the ko but has to give a bigger compensation than two independent gote moves.

Bill: That is the no komaster situation. :) We can evaluate it, but must take specific ko threats into account.

tapir: Re: SL/Bill and Berlekamp treatment do not fit, or is the whole concept inconsistent? (2011-07-30 14:19) [#8651]

Well. Thank you for your answers. While we are at it, I just add some more questions :)

1) komaster-dependent positions when there is no komaster?

But how do I understand sentences like "the count depends on who is komaster", which to me clearly suggest there always is a komaster? Or explicitly: What is the count in those (hyperactive) positions when there is no komaster (no compensation meaning)?

2) why determining a komaster by auction - when there is no compensation?

"For each region of the game, a specified player is designated as the local Komaster. In principle, the privilege of being Komaster might be determined by local auctions, held before the play begins. If the global game contains several regions with potential kos, then it is quite possible for Left to be Komaster of some regions, and Right to be Komaster of others. ...

The Komaster, who just moved to the critical ko position, is compelled to make another local move immediately. | This rule for resolving ko overrides the referee’s normal preference for alternating moves. Instead, the Komaster makes two moves consecutively, to and from the critical ko position, and the Komaster pays two taxes (at the same tax rate), one tax for each move."

To me paying two taxes first sounded like "no compensation but the value of a move elsewhere", however "local auction" suggests the tax for being komaster can be very different from that value, that is it might include compensation. Right? (Honestly, without having your background knowledge this lines read very much like there always is a komaster, who simply is the player winning the ko, the compensation is the local tax as determined by auction, which implies it can be significantly higher than the tax during the previous moves. The difference between the tax during the ko and the tax before/after the ko would then be the genuine compensation due to non-winning ko-threats. However, the komaster would then be just the person who wins the ko, with no deeper meaning.)

If you use komaster in the strict no compensation meaning, is it then even possible to have different komasters for different kos? (Or even if possible wouldn't it be wrong play in most cases.) Say you have a 10pt ko and a 20pt ko. If A is komaster for the 10pt ko and B for the 20pt ko, that means A has exactly as much >10pt threats as B has total threats, while B has exactly as much >20pt threats as A has total threats. If there is a solution, then it is again a very exceptional case (= nobody has <10pt threats, right?). In all other cases we end with no komaster situations, however the text suggests it applies in general not only in some highly unlikely situations.

3) how often there is a komaster (no compensation meaning) that isn't komonster?

It seems komastership is very rare. If you have enough ko-size value threats to answer all the smaller threats of your opponent until he only gains two consecutive gote moves elsewhere, you likely have even more turning you into a spightian komonster, right? At least that is what I took from this, it should be much more likely to either have less or more ko threats than the exactly - say - 20 threats to answer all threats and still win the ko.

4) what do we do with the cases where there is no komaster (no compensation meaning)?

Likely the most common and relevant case in ko fights is a ko, where the losing player gets some compensation and the ko winner can't answer all threats. Currently, there is a big confusion in SL some people use komaster simply as "player who wins the ko", some as "player who wins the ko, without giving any compensation". Terms like koloser contribute to this, because the word itself makes you believe, a koloser is someone losing a ko. Neutral threat environment seems like just another exceptional case to me, useful for mathematical study, but pretty impractical to the ordinary go playing guy/gal, right?

5) how useful is cgt ko terminology to ordinary go playing guy/gal?

If komaster is used strictly in the no compensation meaning, it looks like the ordinary ko is an area uncovered by cgt ko terminology. If this is so, this should be noted somewhere (at least I tried to make sense of kos by this terminology for quite a while and wasn't very successful). The most important insight to me (I started on SL as a bloody beginner.) was that it is useful to postpone resolving a ko at times. I took that lesson from misunderstanding the term komonster, when in fact you don't even have to be komaster (in the no compensation meaning) to reduce the compensation by just playing along while the secondary ko threats of the opponent are becoming smaller. However, the question when exactly to stop and to resolve the ko isn't really answered by the terminology. In fact, even if you are komaster in the no compensation meaning, it might be usually better to allow some compensation and keep your larger threats for another ko.

Best, Tapir

68.124.184.165: Re: SL/Bill and Berlekamp treatment do not fit, or is the whole concept inconsistent? (2011-07-31 17:05) [#8654]

tapir wrote:

Well. Thank you for your answers. While we are at it, I just add some more questions :)

1) komaster-dependent positions when there is no komaster?

But how do I understand sentences like "the count depends on who is komaster", which to me clearly suggest there always is a komaster? Or explicitly: What is the count in those (hyperactive) positions when there is no komaster (no compensation meaning)?

Bill: As of the time of Berlekamp's original paper, the only known way to evaluate hyperactive positions in a way similar to non-ko positions was through the concept of komaster. If, in an actual game, neither player is komaster, well, there you are. Now, such positions may be evaluated via the concept of a neutral threat environment, but such an evaluation is difficult and unrealistic.

What it bolls down to is this. A hyperactive position does not have a unique count without specifying the ko threat situation. The concept of komaster is an abstract way of specifying the ko threat situation. If it does not apply to a specific game, then you need to specify the ko threat situation in another way.

2) why determining a komaster by auction - when there is no compensation?

An auction is a means of evaluation.

"For each region of the game, a specified player is designated as the local Komaster. In principle, the privilege of being Komaster might be determined by local auctions, held before the play begins. If the global game contains several regions with potential kos, then it is quite possible for Left to be Komaster of some regions, and Right to be Komaster of others. ...

The Komaster, who just moved to the critical ko position, is compelled to make another local move immediately. | This rule for resolving ko overrides the referee’s normal preference for alternating moves. Instead, the Komaster makes two moves consecutively, to and from the critical ko position, and the Komaster pays two taxes (at the same tax rate), one tax for each move."

To me paying two taxes first sounded like "no compensation but the value of a move elsewhere", however "local auction" suggests the tax for being komaster can be very different from that value, that is it might include compensation. Right? (Honestly, without having your background knowledge this lines read very much like there always is a komaster, who simply is the player winning the ko, the compensation is the local tax as determined by auction, which implies it can be significantly higher than the tax during the previous moves. The difference between the tax during the ko and the tax before/after the ko would then be the genuine compensation due to non-winning ko-threats. However, the komaster would then be just the person who wins the ko, with no deeper meaning.)

If you use komaster in the strict no compensation meaning, is it then even possible to have different komasters for different kos? (Or even if possible wouldn't it be wrong play in most cases.) Say you have a 10pt ko and a 20pt ko. If A is komaster for the 10pt ko and B for the 20pt ko, that means A has exactly as much >10pt threats as B has total threats, while B has exactly as much >20pt threats as A has total threats. If there is a solution, then it is again a very exceptional case (= nobody has <10pt threats, right?). In all other cases we end with no komaster situations, however the text suggests it applies in general not only in some highly unlikely situations.

At the time of this paper, the way Berlekamp evaluated kos was via the concept of komaster. That is why he gives the impression that there always is one. If a ko positions is hyperactive, then who is komaster makes a difference in the count. A proper auction for being komaster would reflect that difference.

3) how often there is a komaster (no compensation meaning) that isn't komonster?

I suppose that it is not often.

It seems komastership is very rare. If you have enough ko-size value threats to answer all the smaller threats of your opponent until he only gains two consecutive gote moves elsewhere, you likely have even more turning you into a spightian komonster, right? At least that is what I took from this, it should be much more likely to either have less or more ko threats than the exactly - say - 20 threats to answer all threats and still win the ko.

However, being komonster may not be worth much. (Remember, the komonster is a kind of komaster. Being komaster is not rare.) For one thing, it may be better to save your ko threats for later (potential) kos rather than using them to make a small gain in the exchange for the current ko. For another, you probably need a number of extra ko threats to make a difference in the final score, even if you delay winning the ko.

68.124.184.165: Re: SL/Bill and Berlekamp treatment do not fit, or is the whole concept inconsistent? (2011-07-31 16:02) [#8653]

Bill: I see that I answered too quickly. We are not talking about bread-and-butter kos. We are talking about hyperactive kos and ko positions.

68.124.184.165: Re: Definition question (2011-07-30 03:37) [#8647]

tapir: that is a threat which isn't big enough but gains some points as compensation isn't considered as a threat to be answered.

Bill: Nope. If the player who loses the ko makes a ko threat and the player who wins the ko does not answer it, the ko winner is not komaster. The komaster is able to win the ko even after answering all of the koloser's threats.

68.124.184.165: Re: Definition question (2011-07-30 03:33) [#8646]

Bill: In the first figure, W7 is not a ko threat. In the second figure, B2 and B8 are not ko threats.

 
Back to forum     Back to page

New reply


Path: <= CGT path =>
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library