Gote move vs sente move in yose
-
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
Gote move vs sente move in yose
I saw this yose problem in a quite old endgame book (1976) and the author claimed the best yose move for white was the gote move at "a".
I am wondering if a move at "b" may be better, hoping for a sente sequence.
What is your view?
-
gennan
- Lives in gote
- Posts: 497
- Joined: Fri Sep 22, 2017 2:08 am
- Rank: EGF 3d
- GD Posts: 0
- Universal go server handle: gennan
- Location: Netherlands
- Has thanked: 273 times
- Been thanked: 147 times
Re: Gote move vs sente move in yose
Taking b as net score 2 points (4 + -2) [diagram A], I think a would be the middle ("average") between (black responds) net score 2 points (4 + -2) [diagram B] and (black tenuki) net score -5 points (0 + -5) [diagram C] = -1.5 points.
So a is 3.5 points better for white than b.
At least I think this is the endgame calculation method that InSeong Hwang teaches in his online go school (barring privilege / follow-up sequence mistakes that I may have made).
Edit: fixed an arithmetic error.
So a is 3.5 points better for white than b.
At least I think this is the endgame calculation method that InSeong Hwang teaches in his online go school (barring privilege / follow-up sequence mistakes that I may have made).
Edit: fixed an arithmetic error.
Last edited by gennan on Fri Apr 05, 2024 4:23 pm, edited 2 times in total.
-
skydyr
- Oza
- Posts: 2495
- Joined: Wed Aug 01, 2012 8:06 am
- GD Posts: 0
- Universal go server handle: skydyr
- Online playing schedule: When my wife is out.
- Location: DC
- Has thanked: 156 times
- Been thanked: 436 times
Re: Gote move vs sente move in yose
Without looking too closely, I feel like if black plays B and white tenukis, black's best follow-up is still A, so B is just a suboptimal gote move.
-
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: Gote move vs sente move in yose
Taking your diagA I am not sure your movegennan wrote:Taking b as net score 2 points (4 + -2) [diagram A], I think a would be the middle ("average") between (black responds) net score 2 points (4 + -2) [diagram B] and (black tenuki) net score -5 points (0 + -5) [diagram C] = -3 points.
So a is 5 points better for white than b.
At least I think this is the endgame calculation method that InSeong Hwang teaches in his online go school (barring privilege / follow-up sequence mistakes that I may have made).
is correct because of the following answer:I think you have to play the move
and 
in the other order:Let's take your result for diag A and C. I agree here with your count: the white gote sequence allows white to gain 7 suplementary points.
Now the basic question is: in this situation, with a difference of 7 points, do you prefer the sente sequence or the gote one?
Concerning your diag B isn't it better for black to play
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Gote move vs sente move in yose
In the position created by the previous move, the average between Black continues or White continues (because Black tenukis) is the count of that position if, at its own moment, it is a local gote.gennan wrote:Taking b the middle ("average") between (black responds) [...] and (black tenuki)
Therefore, to possibly use this gote count, one must verify whether that position is indeed a local gote. If it is, the use of this gote count for value calculations of the initial position requires justification. With what justification do you want to do so?
Uh, is it, or do you not know, what is the related method he teaches? If eventually you know, what exactly is it for traditional or modern endgame theory?At least I think this is the endgame calculation method that InSeong Hwang teaches in his online go school (barring privilege / follow-up sequence mistakes that I may have made).
-
gennan
- Lives in gote
- Posts: 497
- Joined: Fri Sep 22, 2017 2:08 am
- Rank: EGF 3d
- GD Posts: 0
- Universal go server handle: gennan
- Location: Netherlands
- Has thanked: 273 times
- Been thanked: 147 times
Re: Gote move vs sente move in yose
Oops. Fortunately my move order mistake doesn't change the net score evaluation for diagram A.Gérard TAILLE wrote: I think you have to play the move
I admit that my evaluation of diagram B was a bit hand-wavy. It was more an approximation that I might apply in an actual game with limited time available. A more accurate score calculation would involve more variations and take more time.Gérard TAILLE wrote: Let's take your result for diag A and C. I agree here with your count: the white gote sequence allows white to gain 7 suplementary points.
Now the basic question is: in this situation, with a difference of 7 points, do you prefer the sente sequence or the gote one?
Concerning your diag B isn't it better for black to play
To compare black's atari in your diagram with diagram B (i.e. black takes gote), I'll assume black connects if white connects [diagram D]. Without expanding the move tree further, I estimate the net score as about 2 points (5 + -3), which is the same as the net score for diagram B. This is still not 100% accurate, but I think that expanding the move tree further will result in a net score that is within 0.5 points of that estimate.
Or we can consider your diagram an alternative for diagram C (i.e. black will take sente if white connects), as in diagram E.
If we assume that white 5 in diagram E will be sente, the net score is about -4 points (2 + -6) [diagram F].
If we assume that white 5 in diagram E will be gote, the net score is -9 (-2 + -7) [diagram G].
So the net score for diagram E is the middle of the net scores of diagram F and diagram G = -6.5 points, which I included in the label of diagram E.
From this, my conclusion is that your proposed black atari loses about 1.5 points for black compared to diagram C, unless black still takes gote by continueing as in diagram D, which is more or less neutral compared to diagram B.
I'll readily admit that my score evaluations may be off by fractions of a point when analysed exhaustively. But who has time for that in a real game, and does it matter when the game is not a half-pointer?
Local gote or not may depend on the rest of the board. The provided position is not the whole board, so taking the average is an approximation. It's all I can do with the information provided.RobertJasiek wrote: Therefore, to possibly use this gote count, one must verify whether that position is indeed a local gote. If it is, the use of this gote count for value calculations of the initial position requires justification. With what justification do you want to do so?
I don't know. This is what I understood from his lessons. I don't claim to be an endgame expert, but what I descibed above seems practical and good enough for me to apply in my own games.RobertJasiek wrote: Uh, is it, or do you not know, what is the related method he teaches? If eventually you know, what exactly is it for traditional or modern endgame theory?
But I don't think an exhaustive evaluation will change the answer to the original question if b is better than a. No it isn't, but if black responds to a as in diagram B or D, the net score will be the same as in diagram A (net score 2 points).
So in a sense a is also sente (when the ambient temperature is below 7 points).
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Gote move vs sente move in yose
gennan,
the following things you make right:
- You use negative numbers for White's points as I do, IIRC Bill Spight has done and, by convention, one would do in maths.
- You approximate by ignoring detailed follow-up variations and dissecting sequences into their individual moves, what usually is acceptable for pragmatic reasons.
- You do not describe clearly that you are doing so but you do it at least implicitly: a gote sequence continued by a sente sequence follow-up (privilege) inherits the count after to before the sente sequence follow-up.
You invent the phrase 'net score' for what I call the 'difference value'. Both have disadvantages. Yours is that 'score' has a very specific meaning and one should not create confusion by a too similar name of a different term.
You make, in particular, the following conceptual(!) mistakes:
- Writing, say, (2 + -5) is not proper maths annotation but (2 + (-5)) is, as you should have learnt at school.
- You do not distinguish and determine the sente and gote types of a local endgame in every particular position. You use the words sente and gote in an informal sense related to assuming whether an answer is made [to create an alternating sequence of even or odd length].
- Consider your position created in E. Your position F is Black's gote sequence continuation to the count -4. Your position G is White's sente sequences continuation to the count -9. You derive the count of the position created in E as the average -6.5. This alleged count calculated as the average would be right if the position created in E was a local gote endgame. However, a local gote endgame is one in which Black has a gote sequence and White has a gote sequence. However, G has a sente sequence. Therefore, the position created in E must inherit the sente count -9.
Does Inseong Hwang's go school teach you all those conceptual mistakes, are they all yours or which are whose?
EDIT 2: adapted your corrected values and better interpreted your calculations of positions E F G.
the following things you make right:
- You use negative numbers for White's points as I do, IIRC Bill Spight has done and, by convention, one would do in maths.
- You approximate by ignoring detailed follow-up variations and dissecting sequences into their individual moves, what usually is acceptable for pragmatic reasons.
- You do not describe clearly that you are doing so but you do it at least implicitly: a gote sequence continued by a sente sequence follow-up (privilege) inherits the count after to before the sente sequence follow-up.
You invent the phrase 'net score' for what I call the 'difference value'. Both have disadvantages. Yours is that 'score' has a very specific meaning and one should not create confusion by a too similar name of a different term.
You make, in particular, the following conceptual(!) mistakes:
- Writing, say, (2 + -5) is not proper maths annotation but (2 + (-5)) is, as you should have learnt at school.
- You do not distinguish and determine the sente and gote types of a local endgame in every particular position. You use the words sente and gote in an informal sense related to assuming whether an answer is made [to create an alternating sequence of even or odd length].
- Consider your position created in E. Your position F is Black's gote sequence continuation to the count -4. Your position G is White's sente sequences continuation to the count -9. You derive the count of the position created in E as the average -6.5. This alleged count calculated as the average would be right if the position created in E was a local gote endgame. However, a local gote endgame is one in which Black has a gote sequence and White has a gote sequence. However, G has a sente sequence. Therefore, the position created in E must inherit the sente count -9.
Does Inseong Hwang's go school teach you all those conceptual mistakes, are they all yours or which are whose?
EDIT 2: adapted your corrected values and better interpreted your calculations of positions E F G.
Last edited by RobertJasiek on Fri Apr 05, 2024 6:09 pm, edited 2 times in total.
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Gote move vs sente move in yose
In your EDIT you write: "Local gote or not may depend on the rest of the board."
Local evaluation of counts, move value, gains and (if there are local gote versus sente options) the gote-sente-difference does not(!) depend on the rest of the board. The decision of how (what options) and how long to play locally can (and often does) depend on the global environment.
See also EDIT 2 of my previous message.
Local evaluation of counts, move value, gains and (if there are local gote versus sente options) the gote-sente-difference does not(!) depend on the rest of the board. The decision of how (what options) and how long to play locally can (and often does) depend on the global environment.
See also EDIT 2 of my previous message.
-
gennan
- Lives in gote
- Posts: 497
- Joined: Fri Sep 22, 2017 2:08 am
- Rank: EGF 3d
- GD Posts: 0
- Universal go server handle: gennan
- Location: Netherlands
- Has thanked: 273 times
- Been thanked: 147 times
Re: Gote move vs sente move in yose
Yes, I'm not too well informed about the terminology. I don't remember InSeong being very specific on terminology, but I don't think he uses terminology like 'net score' or 'difference value'. Perhaps his teachers weren't very specific on terminology either, or there may be a language barrier.RobertJasiek wrote: You invent the phrase 'net score' for what I call the 'difference value'. Both have disadvantages. Yours is that 'score' has a very specific meaning and one should not create confusion by a too similar name of a different term.
I'll keep using 'net score' in my posts in this thread to maintain consistency between my posts.
I think Dutch school maths notation is not as strict. I think that the spacing in the expression "2 + -5" would be accepted by most (including most maths school teachers) as clear enough that the " + " is a binary operator (addition) and the "-" is a unary operator (negation), which takes precedence over binary operators even without parentheses.RobertJasiek wrote: - Writing, say, (2 + -5) is not proper maths annotation but (2 + (-5)) is, as you should have learnt at school.
Also in computer programming the expression would be evaluated like that, with or without parentheses. So I would also leave those parentheses around -5 out when programming, to avoid clutter.
Ofcourse I could also just have written "2 - 5" instead of "2 + -5", but I thought that notation might be less clear about the reasoning behind it.
You are right. InSeong didn't teach me wrong, it's just my mistake in executing what he teaches (I tried to provide a caveat in my first post that any mistakes are most likely mine, not his).RobertJasiek wrote: - You do not distinguish and determine the sente and gote types of a local endgame in every particular position. You use the words sente and gote in an informal sense related to assuming whether an answer is made [to create an alternating sequence of even or odd length].
- Consider your position created in E. Your position F is Black's gote sequence continuation to the count -4. Your position G is White's sente sequences continuation to the count -9. You derive the count of the position created in E as the average -6.5. This alleged count calculated as the average would be right if the position created in E was a local gote endgame. However, a local gote endgame is one in which Black has a gote sequence and White has a gote sequence. However, G has a sente sequence. Therefore, the position created in E must inherit the sente count -9.
Does Inseong Hwang's go school teach you all those conceptual mistakes, are they all yours or which are whose?
Diagram G is indeed white's privilege after reaching Diagram E, so the net score of Diagram E should be -9 points, as you said.
So we have this:
...versus this:
From diagram E and D, I think we can take the middle between -9 points and 2 points = -3.5 points for diagram H?
Then we can compare diagram H to diagram J with net score 7 points (8 + -1) :
From diagram H and J, I think we can take the middle between -3.5 points and 7 points = 1.75 points for diagram K?
Then we can compare diagram K with diagram B:
So in the end, diagram K seems slightly worse for black than diagram B (by 0.25 points), so black's atari may not be optimal endgame. But for a really accurate evaluation, I suppose diagram B needs further analysis of follow-ups, which may lead to a fractional adjustment of its net score, though I don't expect the result to be worse for black than net score 2 points.
Edit: I made multiple edits to this post, but I think I'm done now.
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
Re: Gote move vs sente move in yose
Absolutely not! A count of either player's follower could be positive, zero or negative. Therefore, we must not combine signs prematurely. Doing so nevertheless is suggested only by lazy people unaware of the general possibility, or worse, actively deceiving readers about fake simplicity.gennan wrote:I could also just have written "2 - 5" instead of "2 + -5"
No. An average is not for White's one sequence comparing White's other sequence but, once more,
...versus this:
From diagram E and D, I think we can take the middle between -9 points and 2 points = -3.5 points for diagram H?
A COUNT CALCULATED AS THE AVERAGE IS FOR A LOCAL GOTE ENDGAME HAVING BLACK'S GOTE SEQUENCE VERSUS WHITE'S GOTE SEQUENCE.
You have to start AFTER White's initial move!
Now, for the position after White's first move there:
If all our earlier assumptions were right, the gote count of H' would be (2 + (-9)) / 2 = -3.5.
However, in particular, we still do not know if White's sente option might be better:
Black a - White b is a privilege.
***
We need to decide between gote and sente options. Afterwards, we need to back-track White's first move and compare with J.
We do not have enough analysis yet to proceed to answering this question.From diagram H and J, I think we can take the middle between -3.5 points and 7 points = 1.75 points for diagram K?
-
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: Gote move vs sente move in yose
The subject of this topic is "Gote move vs sente move in yose". This issue is of major importance in yose and in a game betweeen a professionnal and an amateur I guess the professionnal will often gain more than 10 points by a far better handling of this notion.
I agree with Robert : you cannot take the average value of a net score after a sente sequence and a gote one.
As an amateur I agree with you Gennan I always avoid an accurate count for two reasons: firstly an accurate evaluation is too time consuming for me, and secondly it is very easy to to be wrong somewhere when a lot of positions are involved.
The best I can suggest is to consider only the logical follow-up sequences assuming the result will be very near from the ideal calculation.
Let me show you how I answer my question concerning the choice between black a or black b in the above diagram.
After black
I simply consider that black will be later able to play 
in sente with white answering
Here after I consider white will be later able to play in sente with black answering
My conclusion is that black "a" and black "b" in the first diagram of this post are equivalent (net value +2)
It remains my initial question:
what is the best move for white ?
I agree with Robert : you cannot take the average value of a net score after a sente sequence and a gote one.
As an amateur I agree with you Gennan I always avoid an accurate count for two reasons: firstly an accurate evaluation is too time consuming for me, and secondly it is very easy to to be wrong somewhere when a lot of positions are involved.
The best I can suggest is to consider only the logical follow-up sequences assuming the result will be very near from the ideal calculation.
Let me show you how I answer my question concerning the choice between black a or black b in the above diagram.
After black
I simply consider that black will be later able to play in sente with white answering Here after
I consider white will be later able to play in sente with black answering My conclusion is that black "a" and black "b" in the first diagram of this post are equivalent (net value +2)
It remains my initial question:
what is the best move for white ?
-
gennan
- Lives in gote
- Posts: 497
- Joined: Fri Sep 22, 2017 2:08 am
- Rank: EGF 3d
- GD Posts: 0
- Universal go server handle: gennan
- Location: Netherlands
- Has thanked: 273 times
- Been thanked: 147 times
Re: Gote move vs sente move in yose
I did not consider that option for white. Indeed it should be included in the analysis.RobertJasiek wrote: However, in particular, we still do not know if White's sente option might be better:
Black a - White b is a privilege.
OK, but shouldn't we then also consider a different move for white than connecting withGérard TAILLE wrote:
Here after
My conclusion is that black "a" and black "b" in the first diagram of this post are equivalent (net value +2)
, like so?If white takes sente after
, a will be black's privilege, reverting to a net score of 2 points (more or less the same as diagram A as it was corrected by you, where white started with the first move at 
).If white takes gote by playing at a, it will revert to Robert's diagram above with net score 0 points (2 + -2).
I don't know yet if the evaluation is the same for black's atari and black's descent.
Perhaps it doesn't really matter, except that black b gives white more viable options than black a?
Maybe also for your original question, the only difference is that white a gives white more viable options than white b?
-
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: Gote move vs sente move in yose
I agree with you gennan.gennan wrote:
In this diagram white can choose to take sente by playing either white "a" or white "b" with the same result (net score = +2).
For me this confirm that, in the following diagram, the black move at "a" and the black move at "b" are both gote and lead to the same result (net score = +2). Don't you agree?
-
gennan
- Lives in gote
- Posts: 497
- Joined: Fri Sep 22, 2017 2:08 am
- Rank: EGF 3d
- GD Posts: 0
- Universal go server handle: gennan
- Location: Netherlands
- Has thanked: 273 times
- Been thanked: 147 times
Re: Gote move vs sente move in yose
To humour myself, I tried to peek KataGo's thoughts on the position.
In this position, my KataGo seems to think that black c is slightly better than both black a and black b. I hadn't thought of black c yet, but it's an interesting move.
Black c does make sense, because white c is also a good move in response to black a or b, suggesting that c is a vital point.
From black c we can again evaluate a net score of 2 points (4 + -2)
And I suppose this variation is also the same result:
In the diagram above, black has 5 points, while white has either 1 or 5 points (so 3 points on average), depending on who plays d double gote first: net score is 2 points (5 + -3).
With all those moves having the same value, the only advantage I can see for one move over the other is perhaps how many good options you are giving your opponent to reach an optimal result, and how easy or difficult it is to find those good options. And perhaps there may be a difference in ko threats between different variations?
Other than that, I now agree with your very first post in this thread, where you disputed the author's claim.
Although I wonder if you'd still call the best move according to the author a gote move, because we seem to converge on it being a sente move.
In this position, my KataGo seems to think that black c is slightly better than both black a and black b. I hadn't thought of black c yet, but it's an interesting move.
Black c does make sense, because white c is also a good move in response to black a or b, suggesting that c is a vital point.
From black c we can again evaluate a net score of 2 points (4 + -2)
And I suppose this variation is also the same result:
In the diagram above, black has 5 points, while white has either 1 or 5 points (so 3 points on average), depending on who plays d double gote first: net score is 2 points (5 + -3).
Yes, I agree. Even though KataGo may have slight preferences for one over the other, I think that black a, b and c all give the same result of net score +2.Gérard TAILLE wrote: For me this confirm that, in the following diagram, the black move at "a" and the black move at "b" are both gote and lead to the same result (net score = +2). Don't you agree?
With all those moves having the same value, the only advantage I can see for one move over the other is perhaps how many good options you are giving your opponent to reach an optimal result, and how easy or difficult it is to find those good options. And perhaps there may be a difference in ko threats between different variations?
Other than that, I now agree with your very first post in this thread, where you disputed the author's claim.
Although I wonder if you'd still call the best move according to the author a gote move, because we seem to converge on it being a sente move.
-
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: Gote move vs sente move in yose
I agree with you Gennan, the three black moves a, b, and c give the same result (net score +2).gennan wrote:To humour myself, I tried to peek KataGo's thoughts on the position.
In this position, my KataGo seems to think that black c is slightly better than both black a and black b. I hadn't thought of black c yet, but it's an interesting move.
Black c does make sense, because white c is also a good move in response to black a or b, suggesting that c is a vital point.
From black c we can again evaluate a net score of 2 points (4 + -2)
Having seen the work of Gennan on this position above I was OC happy to have a discussion looking for the best black move in such position and it is quite satisfactory to see that we reached a common understanding.
Now let's return to my initial question:
What is the best white move a or b?
Considering this initial question, I consider the previous analysis as only another exercice on the topic on this thread.
Let me give you only the beginning of my analysis:
For me the white move
above is a big move made in a position where the temperature is probably equal to 10 points if not more!After this move
my intuition (or my go experience) tells me that the temperature has now decreased and as a consequence black will play tenuki. That means that white will then be able to play the following sequence in sente:As you see my analyse does not take into account a black move after white
which is a major simplification.How do you continue the analysis to find what is the best first white move?