That was exactly my first view but when you consider white to move it seems to me better to play: That way white gets 5 points in sente (<=> 10 points gote)dfan wrote:Here are my calculations. No guarantee that they're correct, but they do match the flashcard.
That gives a total swing of 8 points in gote.
Cards or app for miai-value based endgame practice?
-
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: Cards or app for miai-value based endgame practice?
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Cards or app for miai-value based endgame practice?
Sente must be verified or rejected.Gérard TAILLE wrote:That way white gets 5 points in sente
-
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: Cards or app for miai-value based endgame practice?
White do you mean Robert? A flashcard is only made to train you to calculate a move value in a PRACTICAL game. I agree with you that verification would be appreciate in a theoritical approach but I guess such theoritical approach would need far too much time.RobertJasiek wrote:Sente must be verified or rejected.Gérard TAILLE wrote:That way white gets 5 points in sente
IOW, assuming you are white in a practical game: do you prefer a or b (assuming some kind of ideal environment)?
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Cards or app for miai-value based endgame practice?
I prefer the naive assumption that, if it is a local sente, it must be Black's...! White's region is wider open so Black should have the larger follow-up so any local sente would be Black's. Flashcard or not - the basics must be right! I.e., if a flashcard merely says 5 pt in [local] sente, the card creator must still verify a) it is a local sente, b) whether it is either Black's or White's local sente and c) write the right sente player on the flashcard.
-
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: Cards or app for miai-value based endgame practice?
My initial question was : "as WHITE, do you prefer to play at "a" or do you prefer to play at "b"? (as far as I am concerned I prefer white "b")RobertJasiek wrote:I prefer the naive assumption that, if it is a local sente, it must be Black's...! White's region is wider open so Black should have the larger follow-up so any local sente would be Black's. Flashcard or not - the basics must be right! I.e., if a flashcard merely says 5 pt in [local] sente, the card creator must still verify a) it is a local sente, b) whether it is either Black's or White's local sente and c) write the right sente player on the flashcard.
It is not that easy to understand your answer. I agree with you that a black move at "a" would be "more" sente that a white move at "b" but how can you justify your conclusion : "if it is a local sente, it must be Black's...!"
This could be true if the best black move would be the sente move black "a" but you never claimed that the sente black move at "a" was better than the gote black move at "b" did you?
In any case and in order to make some progress, can you please clarify your answer to my initial question?
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Cards or app for miai-value based endgame practice?
For local evaluation, one must always consider both Black's and White's start, like one does in CGT.
In a global context, one can consider one particular player's start and whether his usually correct local gote / ambiguous / sente play is also correct in the particular global context.
For the local verification of the options with follow-up as a local gote / ambiguous / local sente, for example, compare the tentative gote move value to Black's follow-up move value and White's follow-up move value. Here, we also have other first move options simply speaking without follow-up because they are (3-move) gote options. We also need to compare, if indeed, sente option to the gote option and the gote-sente-difference in points.
Thereby you identify the correct type and values. Except that you don't because you forgot Bill's theory and do not look up mine;) It depends on whether we are in the early or late endgame.
In a global context, one can consider one particular player's start and whether his usually correct local gote / ambiguous / sente play is also correct in the particular global context.
For the local verification of the options with follow-up as a local gote / ambiguous / local sente, for example, compare the tentative gote move value to Black's follow-up move value and White's follow-up move value. Here, we also have other first move options simply speaking without follow-up because they are (3-move) gote options. We also need to compare, if indeed, sente option to the gote option and the gote-sente-difference in points.
Thereby you identify the correct type and values. Except that you don't because you forgot Bill's theory and do not look up mine;) It depends on whether we are in the early or late endgame.
-
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: Cards or app for miai-value based endgame practice?
You may think that I do not agree with you but it is not true. Seeing your approach is purely theoritical (you very often remind the work made in CGT, Bill's theory or yours) I completly agree with you that a lot of work has to be done in order to verify and confirm if a black or white sente move is better than a black or white gote move, or if other sequences of moves have to be taken into account etc...RobertJasiek wrote:For local evaluation, one must always consider both Black's and White's start, like one does in CGT.
In a global context, one can consider one particular player's start and whether his usually correct local gote / ambiguous / sente play is also correct in the particular global context.
For the local verification of the options with follow-up as a local gote / ambiguous / local sente, for example, compare the tentative gote move value to Black's follow-up move value and White's follow-up move value. Here, we also have other first move options simply speaking without follow-up because they are (3-move) gote options. We also need to compare, if indeed, sente option to the gote option and the gote-sente-difference in points.
Thereby you identify the correct type and values. Except that you don't because you forgot Bill's theory and do not look up mine;) It depends on whether we are in the early or late endgame.
But here is the misunderstanding: providing a flashcard is not obvious the purely theoritical approach is clearly impossible.
As an example let's consider flashcard number 33: Can CGT, Bill's theory or your theory can help us to find the move value of such position? Surely not because it is far too difficult. I guess even a hundred pages of analyses will not be enough to calculate the correct move value.
this position is simplier but I guess that dozens of pages would be necessary to give the correct move value. Even without finding the correct move value it is already difficult to say if white move "a" is better than white move "b" and I clearly understand why you did not answer my question : the answer is simply too difficult to find in a pure theoritical approach.
BTW which go player will accept to read in detail such dozens pages of analysis?
As a consequence some other approaches are needed and discussions can take place to improve such approaches and to improve the first evaluation found.
But if you contest a result by saying it is not correct for a theoritical point of view then, though you are right, it is only a very negative approach because this theoritical approach is completly impractical and cannot bring any help.
Flashcards are for go players and not for theoriticians.
I am pretty sure you can bring nice ideas on some flashcards but as a go player not as a theoritician.
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Cards or app for miai-value based endgame practice?
In your previous message, your first diagram is more complicated than the second. I do not think that both are as complicated to need "dozens of pages". OTOH, they are also not so simple, from a POV of rather accurate evaluation, to allow for a naive "flashcard" guess. However, they are complicated enough to require somewhere between minutes and hours of thinking, and it is hard to predict how much time will be needed. As I have explained before, it takes only seconds to show a problem diagram but I do not have the time to solve each problem accurately whenever a new problem is shown. It is similar to intermediate to advanced life and death problems: not everybody has the time to solve all of them and explain the solution in detail just because the problems are shown.
However, one can make simplifying assumptions:
- Suppose we know which follow-ups are relevant and study only them.
- Suppose the options with follow-ups are sente for one starting player (here: Black) and gote for the opponent.
The following assumption is risky:
- For each starting player, suppose that one of the two options dominates and we know which it is.
However, one can make simplifying assumptions:
- Suppose we know which follow-ups are relevant and study only them.
- Suppose the options with follow-ups are sente for one starting player (here: Black) and gote for the opponent.
The following assumption is risky:
- For each starting player, suppose that one of the two options dominates and we know which it is.
-
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: Cards or app for miai-value based endgame practice?
Let's me show you how I analysed this position with my method (I am now not very far for being able to give you an article on the subject). First of all my analyse is based on area scoring (and not territory scoring)RobertJasiek wrote:In your previous message, your first diagram is more complicated than the second. I do not think that both are as complicated to need "dozens of pages". OTOH, they are also not so simple, from a POV of rather accurate evaluation, to allow for a naive "flashcard" guess. However, they are complicated enough to require somewhere between minutes and hours of thinking, and it is hard to predict how much time will be needed. As I have explained before, it takes only seconds to show a problem diagram but I do not have the time to solve each problem accurately whenever a new problem is shown. It is similar to intermediate to advanced life and death problems: not everybody has the time to solve all of them and explain the solution in detail just because the problems are shown.
However, one can make simplifying assumptions:
- Suppose we know which follow-ups are relevant and study only them.
- Suppose the options with follow-ups are sente for one starting player (here: Black) and gote for the opponent.
The following assumption is risky:
- For each starting player, suppose that one of the two options dominates and we know which it is.
At first sight I was convinced that the best black move as well as the best white move where the hane gote moves showed by dfan in viewtopic.php?p=278108#p278108
The result in area approach is the following: The five marked intersections represent the intersections each player can gain by playing first.
On average (I mean the count of the position) 2.5 of these intersections are for black and 2.5 are for white. What about the black sente move at "a" or the white sente move at "b" ?
If black plays at the sente point "a" and white answers at "b" then, comparing to the previous diagram black takes only two of the five marked intersections => count is better for white => black "a" is wrong
On the other hand if white plays at the sente point "b" and black answers at "a" then white takes three of the five marked intersections => count is better for white => white "b" is correct providing it is sente. Obviously the potential sente ogeima after white "b" proves that white "b" is sente => white "b" is probably the best move.
=> the position is sente for white
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Cards or app for miai-value based endgame practice?
With rough analysis, you can prove everything incl. that the universe does not exist:)
-
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: Cards or app for miai-value based endgame practice?
I tried to show that, according to my analysis, the position is sente for white and your answer was the following general statement : no if the position is sente then it is necessarily a BLACK sente position. I tried then to bring some more information showing why I consider it is a white sente position and now you answered that my explanations are only nonsense.RobertJasiek wrote:With rough analysis, you can prove everything incl. that the universe does not exist:)
What do you bring yourself on this position? You never said what is for you the correct move value, you never said what is for you the correct count, you never said what is for you the best sequence for black (the sente or the gote one) and for white (the sente or the gote one).
Robert, it is very easy to be systematically negative by using general statement but it is far more difficult to bring positive criticism (like counter example) or to bring another analysis. I can easily understand you have no time enough to bring concrete information on this position but in that case please avoid negative (if not agressive) statement which are completly unuseful and instead try to encourage new participations to the discussion.
Why on earth do you avoid any cooperative work to have a better understanding of such position?
One certainty for me: you cannot be interested by my "rough" method.
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Cards or app for miai-value based endgame practice?
Approximations are not per se bad and I have not seen your final method so cannot comment on it. I have commented on your analysis of this example. More specifically:
What is the meaning of assigne 2.5 to each player so that the impact is 0 because 2.5 for Black minus 2.5 for White from Black's perspective is 0?
Since you make it 0, why may you call it an average?
However, you make the statement that, assuming sente, White b was correct. As before, this is a premature implication for analogue reasons.
Therefore, also your "Obviously" must be wrong.
Ok.Gérard TAILLE wrote: First of all my analyse is based on area scoring (and not territory scoring)
How do you derive these intersections from dfan's diagrams and 8 points territory difference?At first sight I was convinced that the best black move as well as the best white move where the hane gote moves showed by dfan in viewtopic.php?p=278108#p278108
Quoting dfan: That gives a total swing of 8 points in gote.
The result in area approach is the following: The five marked intersections represent the intersections each player can gain by playing first.
Why may you speak of a count of the [local] position when you consider 5 intersections but ignore the adjacent affected intersections?On average (I mean the count of the position)
Why may you form the average on these intersections in the context of the whole local endgame?2.5 of these intersections are for black and 2.5 are for white.
What is the meaning of assigne 2.5 to each player so that the impact is 0 because 2.5 for Black minus 2.5 for White from Black's perspective is 0?
Since you make it 0, why may you call it an average?
Why may / do you mean sente moves at this moment of analysis? (Also "sente point" twice below.)What about the black sente move at "a" or the white sente move at "b" ?
Ok.If black plays at the sente point "a" and white answers at "b" then, comparing to the previous diagram black takes only two of the five marked intersections
Better than what? (I can guess, but it is your analysis and must be clear from it.)=> count is better for white
This is an adventurous implication. You presume that a) White must answer, b) if White does answer, necessarily it would be bad for Black to maintain the initiative by playing the exchange in a local sente sequence, c) there cannot be a general condition comparing gote sequence option to sente sequence option sometimes determining one or sometimes the other to be correct. Your implication is premature without these considerations or justifying an approximative method with which they are ignored on purpose.=> black "a" is wrong
Ok.On the other hand if white plays at the sente point "b" and black answers at "a" then white takes three of the five marked intersections
As before.=> count is better for white
Now this is better: you consider the sente as a presupposition and case.=> white "b" is correct providing it is sente.
However, you make the statement that, assuming sente, White b was correct. As before, this is a premature implication for analogue reasons.
Why is the ogeima sente? By experience (Sensei's Library, Bill Spight, I) with careful endgame value analysis, a monkey into two rows of possible territory along the edge is a local gote move.Obviously the potential sente ogeima after white "b"
Therefore, also your "Obviously" must be wrong.
Wrong statements prove nothing.proves
See above.that white "b" is sente
Writing lots of implication symbols does not establish a proof by itself.=> white "b" is probably the best move.
You have made too many mistakes and left too many analysis gaps (more than zero) to make your final "implication".=> the position is sente for white
-
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: Cards or app for miai-value based endgame practice?
Your questions are valuable questions Robert but I do not want to simply answer all your criticisms without seeing you taking risks by giving your own feeling (analysis?) of the position.RobertJasiek wrote:Approximations are not per se bad and I have not seen your final method so cannot comment on it. I have commented on your analysis of this example. More specifically:
Ok.Gérard TAILLE wrote: First of all my analyse is based on area scoring (and not territory scoring)
How do you derive these intersections from dfan's diagrams and 8 points territory difference?At first sight I was convinced that the best black move as well as the best white move where the hane gote moves showed by dfan in viewtopic.php?p=278108#p278108
Quoting dfan: That gives a total swing of 8 points in gote.
The result in area approach is the following: The five marked intersections represent the intersections each player can gain by playing first.
Why may you speak of a count of the [local] position when you consider 5 intersections but ignore the adjacent affected intersections?On average (I mean the count of the position)
Why may you form the average on these intersections in the context of the whole local endgame?2.5 of these intersections are for black and 2.5 are for white.
What is the meaning of assigne 2.5 to each player so that the impact is 0 because 2.5 for Black minus 2.5 for White from Black's perspective is 0?
Since you make it 0, why may you call it an average?
Why may / do you mean sente moves at this moment of analysis? (Also "sente point" twice below.)What about the black sente move at "a" or the white sente move at "b" ?
Ok.If black plays at the sente point "a" and white answers at "b" then, comparing to the previous diagram black takes only two of the five marked intersections
Better than what? (I can guess, but it is your analysis and must be clear from it.)=> count is better for white
This is an adventurous implication. You presume that a) White must answer, b) if White does answer, necessarily it would be bad for Black to maintain the initiative by playing the exchange in a local sente sequence, c) there cannot be a general condition comparing gote sequence option to sente sequence option sometimes determining one or sometimes the other to be correct. Your implication is premature without these considerations or justifying an approximative method with which they are ignored on purpose.=> black "a" is wrong
Ok.On the other hand if white plays at the sente point "b" and black answers at "a" then white takes three of the five marked intersections
As before.=> count is better for white
Now this is better: you consider the sente as a presupposition and case.=> white "b" is correct providing it is sente.
However, you make the statement that, assuming sente, White b was correct. As before, this is a premature implication for analogue reasons.
Why is the ogeima sente? By experience (Sensei's Library, Bill Spight, I) with careful endgame value analysis, a monkey into two rows of possible territory along the edge is a local gote move.Obviously the potential sente ogeima after white "b"
Therefore, also your "Obviously" must be wrong.
Wrong statements prove nothing.proves
See above.that white "b" is sente
Writing lots of implication symbols does not establish a proof by itself.=> white "b" is probably the best move.
You have made too many mistakes and left too many analysis gaps (more than zero) to make your final "implication".=> the position is sente for white
Let me remind you that you never said what is for you the correct move value, you never said what is for you the correct count, you never said what is for you the best sequence for black (the sente or the gote one) and for white (the sente or the gote one).
I decided clearly to wait for more information before answering to your questions. That way I will be also able to discuss your method. Isn't it more fair Robert?
BTW what is for you a proof?
Assume my own move value calculation is different from yours. Does that mean that one of the two move values is wrong? It's not sure because our defintion may be different. To my knowledge you never gave your own move value definition on this forum (I guess this defintion can only be found in your book but I fear a great majority of readers has not got this book).
What do you propose to decide witch approach will be the more correct?
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Cards or app for miai-value based endgame practice?
I do not evaluate local endgames by guessing but by exploring the game tree and applying theory. Without, I do not know the correct values and sente / gote types.
Endgame evaluation is not a matter of competition. Instead, it is application of theory. Depending on the quality of the theory, we can or cannot trust application results.
In a strict sense, a proof is what mathematics does: transformations of truths. For informal theory, proof might have an informal meaning and be weak.
Actually, I do not recall whether there have other mathematical definitions of move value; I would need to look it up. Berlekamp? Previous definitions focussed on CGT and used other terms with different intentions, especially related to infinitesimals. I have defined move value for some standard cases. For practical purposes (let us say: modulo infinitesimal), the different definitions are the same.
If you or anybody uses different calculations, the resulting values should be the same nevertheless. (There are extra considerations if options occur.)
IIRC, You do not seem to intend to redefine move value but you only intend to determine it differently. Therefore, you should produce the same values. Or their approximations.
I have stated my move value definition several times. The definition is two steps more formal than you currently need. Without options, compare move value to each follow-up move values. With options, it is more complicated.
An approach is correct if it conforms to mathematical theory. An approach not conforming to mathematical theory is outside evaluation for correct or wrong - it is so weak that it needs translation to mathematics or admitting of invalidity due to informality.
Endgame evaluation is not a matter of competition. Instead, it is application of theory. Depending on the quality of the theory, we can or cannot trust application results.
In a strict sense, a proof is what mathematics does: transformations of truths. For informal theory, proof might have an informal meaning and be weak.
Actually, I do not recall whether there have other mathematical definitions of move value; I would need to look it up. Berlekamp? Previous definitions focussed on CGT and used other terms with different intentions, especially related to infinitesimals. I have defined move value for some standard cases. For practical purposes (let us say: modulo infinitesimal), the different definitions are the same.
If you or anybody uses different calculations, the resulting values should be the same nevertheless. (There are extra considerations if options occur.)
IIRC, You do not seem to intend to redefine move value but you only intend to determine it differently. Therefore, you should produce the same values. Or their approximations.
I have stated my move value definition several times. The definition is two steps more formal than you currently need. Without options, compare move value to each follow-up move values. With options, it is more complicated.
An approach is correct if it conforms to mathematical theory. An approach not conforming to mathematical theory is outside evaluation for correct or wrong - it is so weak that it needs translation to mathematics or admitting of invalidity due to informality.
-
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: Cards or app for miai-value based endgame practice?
I see: you are the God of go theory, the only correct move value is NECESSARILY YOURS and any different value is necessarily wrong => I have to proof that my own values are the same as yours. Do you really think I can agree? BTW how can I know if my result are correct if you are unable to give us the result of your magic method?RobertJasiek wrote:I do not evaluate local endgames by guessing but by exploring the game tree and applying theory. Without, I do not know the correct values and sente / gote types.
Endgame evaluation is not a matter of competition. Instead, it is application of theory. Depending on the quality of the theory, we can or cannot trust application results.
In a strict sense, a proof is what mathematics does: transformations of truths. For informal theory, proof might have an informal meaning and be weak.
Actually, I do not recall whether there have other mathematical definitions of move value; I would need to look it up. Berlekamp? Previous definitions focussed on CGT and used other terms with different intentions, especially related to infinitesimals. I have defined move value for some standard cases. For practical purposes (let us say: modulo infinitesimal), the different definitions are the same.
If you or anybody uses different calculations, the resulting values should be the same nevertheless. (There are extra considerations if options occur.)
IIRC, You do not seem to intend to redefine move value but you only intend to determine it differently. Therefore, you should produce the same values. Or their approximations.
I have stated my move value definition several times. The definition is two steps more formal than you currently need. Without options, compare move value to each follow-up move values. With options, it is more complicated.
An approach is correct if it conforms to mathematical theory. An approach not conforming to mathematical theory is outside evaluation for correct or wrong - it is so weak that it needs translation to mathematics or admitting of invalidity due to informality.
Be serious your method is unable to give the result of the 48 flashcards in the file mentionned earlier within an acceptable time (let's say one or two days) is it?
Sure your method is fine for a theoritical approach but what is its value for practical game (I mean positions like the 48 flashcards) ? Note that the flashcard 24 we discussed is far to be the most difficult, is it? Can you tell us if a player uses really your method in actual game. Do you use yourself your method in actual game?