Sorry, but the problem is so overwhelming complex.Bojanic wrote:Found it on your site, it is a bit difficult to navigate.
And my knowledge of website design was not sufficient to do have more than static html.
KataGo @ Igo Hatsuyoron 120
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
And my knowledge of website design was not sufficient to do have more than static html.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
sorin
- Lives in gote
- Posts: 389
- Joined: Wed Apr 21, 2010 9:14 pm
- Has thanked: 418 times
- Been thanked: 198 times
Re: KataGo @ Igo Hatsuyoron 120
There was an update on this posted yesterday, very fascinating read! https://blog.janestreet.com/deep-learni ... the-world/
Sorin - 361points.com
-
Gomoto
- Gosei
- Posts: 1733
- Joined: Sun Nov 06, 2016 6:56 am
- GD Posts: 0
- Location: Earth
- Has thanked: 621 times
- Been thanked: 310 times
Re: KataGo @ Igo Hatsuyoron 120
Can not go to sleep
My precious ...
Exploring two space high pincer of the 3-4 low approach with the new Kata-Igo network
This looks very exciting
Fascinationg to watch the new opinions on the variations
First there was a digital Go Seigen offering to teach me. Now it feels like the entire Honinbo lineup is visiting and discussing the variations in front of me
I am so grateful
My precious ...
Exploring two space high pincer of the 3-4 low approach with the new Kata-Igo network
This looks very exciting
Fascinationg to watch the new opinions on the variations
First there was a digital Go Seigen offering to teach me. Now it feels like the entire Honinbo lineup is visiting and discussing the variations in front of me
I am so grateful
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
Flying below the radar of area scoring
Lightvector highlighted (and simultaneously warned us all?) that KataGo – using area scoring – will be unable to distinguish between W+1(territory scoring) and W+2(territory scoring). This posting will show how significant his annotation was.
White's bad-shape move at 120 is gote locally, so Black – after inserting some forcing moves – might play tenuki with 125.
If Black plays away, White's wedge at 126 is best.
White did not only destroy all of Black's territory in the left centre.
She also keeps her liberty at a, and created an additional approach-move liberty at b.
White 136 and Black 137 are miai.
Finally, White answers at the top right, wedging in at 142.
Black 143 might be the reply that comes to mind first, but the subsequent sequence will not change the final score.
Black 147 is sente, ...
...but White will capture this stone (and connect) in the very end.
Black 149 continues with the Main Semeai.
The game will still end with W + 2.
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
However, Joachim suggested answering White's wedge (triangled) by giving atari from above, with the move at 143 here.
In principle, this move is a kind of Michael Redmond's kikashi, just played on the other side of the bamboo joint. Black gains an additional liberty at a, being worth one destroyed point of White territory.
(Funnily enough I had to convince Joachim from the validity of the above mentioned effect in this case. Usually, it is Joachim, who has to remind ME of findings that we had a decade ago.)
Black will capture at c, only after White connected at b.
+ + + + +
Let's compare the final board positions:
In Joachim's variation (at the left), White has two points of territory at the top, at a and at b.
In Joachim's variation, Black has captured two White stones, at c and at o.
The net effect is zero.
In KataGo's variation (at the right), White has one point of territory at e, which is compensated for by Black's point of territory at f.
Eventually, Black's additional point of territory inside his eye at c makes the difference.
The game – following Joachim's tesuji – now ends with W + 1!
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
There can only be one
Above we found that White's bad-shape move at a is no longer effective, as it is miai with the point of b.
My guzumi (triangled) remains the only effective bad-shape move in this problem
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
For your convenience follow the SGFs of both lines:
Lightvector highlighted (and simultaneously warned us all?) that KataGo – using area scoring – will be unable to distinguish between W+1(territory scoring) and W+2(territory scoring). This posting will show how significant his annotation was.
White's bad-shape move at 120 is gote locally, so Black – after inserting some forcing moves – might play tenuki with 125.
If Black plays away, White's wedge at 126 is best.
White did not only destroy all of Black's territory in the left centre.
She also keeps her liberty at a, and created an additional approach-move liberty at b.
White 136 and Black 137 are miai.
Finally, White answers at the top right, wedging in at 142.
Black 143 might be the reply that comes to mind first, but the subsequent sequence will not change the final score.
Black 147 is sente, ...
...but White will capture this stone (and connect) in the very end.
Black 149 continues with the Main Semeai.
The game will still end with W + 2.
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
However, Joachim suggested answering White's wedge (triangled) by giving atari from above, with the move at 143 here.
In principle, this move is a kind of Michael Redmond's kikashi, just played on the other side of the bamboo joint. Black gains an additional liberty at a, being worth one destroyed point of White territory.
(Funnily enough I had to convince Joachim from the validity of the above mentioned effect in this case. Usually, it is Joachim, who has to remind ME of findings that we had a decade ago.)
Black will capture at c, only after White connected at b.
+ + + + +
Let's compare the final board positions:
In Joachim's variation (at the left), White has two points of territory at the top, at a and at b.
In Joachim's variation, Black has captured two White stones, at c and at o.
The net effect is zero.
In KataGo's variation (at the right), White has one point of territory at e, which is compensated for by Black's point of territory at f.
Eventually, Black's additional point of territory inside his eye at c makes the difference.
The game – following Joachim's tesuji – now ends with W + 1!
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
There can only be one
Above we found that White's bad-shape move at a is no longer effective, as it is miai with the point of b.
My guzumi (triangled) remains the only effective bad-shape move in this problem
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
For your convenience follow the SGFs of both lines:
- Attachments
-
- JoachimsO14AfterN13.sgf
- (3.81 KiB) Downloaded 1619 times
-
- KataGosF9.sgf
- (3.52 KiB) Downloaded 1604 times
Last edited by Cassandra on Fri Dec 13, 2019 1:53 am, edited 1 time in total.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Bill Spight
- Honinbo
- Posts: 10905
- Joined: Wed Apr 21, 2010 1:24 pm
- Has thanked: 3651 times
- Been thanked: 3373 times
Re: KataGo @ Igo Hatsuyoron 120
I think what y'all are doing is marvelous. 
I haven't studied it to speak of, but there is something here I don't understand. You say that Black147 is sente. OK. I gather that the reason is that White cannot allow Black to make two eyes.
Just in this little local area Black has captured the
stone, for one point, but White, unless there is something special about this ko, has ⅓ point, on average, towards capturing the Black stone. On average, then, Black has ⅔ pt. However, IIUC, a dame is worth one pt. for Black, and White148 has filled that dame. That means that Black really has, on average, -⅓ pt., while if White simply saves the stone, Black has 0 pts. and still has the dame. That is, Black147 is a losing sente and Black should not play it. (We're behind the looking glass, but there you go. 
)
Now, if Black does not play B147 here, White, later on, may fill the dame without saving the stone. In that case the play may go this way locally.
In this case Black takes the ko later rather than sooner, and, as we know, that may be important in a semeai.
So, from what little I understand, it looks like Black147 is a mistake.
I haven't studied it to speak of, but there is something here I don't understand. You say that Black147 is sente. OK. I gather that the reason is that White cannot allow Black to make two eyes.Just in this little local area Black has captured the
stone, for one point, but White, unless there is something special about this ko, has ⅓ point, on average, towards capturing the Black stone. On average, then, Black has ⅔ pt. However, IIUC, a dame is worth one pt. for Black, and White148 has filled that dame. That means that Black really has, on average, -⅓ pt., while if White simply saves the stone, Black has 0 pts. and still has the dame. That is, Black147 is a losing sente and Black should not play it. (We're behind the looking glass, but there you go. )Now, if Black does not play B147 here, White, later on, may fill the dame without saving the
stone. In that case the play may go this way locally.In this case Black takes the ko later rather than sooner, and, as we know, that may be important in a semeai.
So, from what little I understand, it looks like Black147 is a mistake.
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
At some point, doesn't thinking have to go on?
— Winona Adkins
Visualize whirled peas.
Everything with love. Stay safe.
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
Many thanks, Bill, for clarifying this technical issue.
"Ko" is one of my very blind spots.
"Ko" is one of my very blind spots.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
Too much noise for KataGo?
Joachim's tesuji in the bamboo-joint seemed to be a quite unusual move, because KataGo did not even consider it, see "original.png" (green = KataGo's prefered move; red = other moves with displayed playouts).
Because the other considered moves were mainly part of the liberty-occupying part of the Main Semeai, so I wondered what would happen, if the other liberties are occupied first, and the bamboo-joint area is left untouched until shortly before the moment when the tail is captured.
The result of KataGo's analysis at this moment can be seen in "delayed.png".
Now, Joachim's tesuji is also KataGo's prefered move!!!
It seems to me that you must be very well aware of what you are doing when using KataGo as a very powerful tool for analysis.
Joachim's tesuji in the bamboo-joint seemed to be a quite unusual move, because KataGo did not even consider it, see "original.png" (green = KataGo's prefered move; red = other moves with displayed playouts).
Because the other considered moves were mainly part of the liberty-occupying part of the Main Semeai, so I wondered what would happen, if the other liberties are occupied first, and the bamboo-joint area is left untouched until shortly before the moment when the tail is captured.
The result of KataGo's analysis at this moment can be seen in "delayed.png".
Now, Joachim's tesuji is also KataGo's prefered move!!!
It seems to me that you must be very well aware of what you are doing when using KataGo as a very powerful tool for analysis.
- Attachments
-
- original.png (11.04 KiB) Viewed 16747 times
-
- delayed.png (10.88 KiB) Viewed 16747 times
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
Warning: The Semeai Variation is near!
White will choose the Capture Variation to realise her territorial gains at the left, due to KataGo's new atari / capture / connection.
However, if you want to evaluate variations, you have to be aware that the Semeai Variation might be nearer than expected. I had to learn this painfully ...
In the diagrams above, we compare the final position of the Semeai Variations after KataGo's new moves (at the left) with that of Our Solution (at the right).
Legend:
Triangle = plus one point for Black
Square = plus two points for Black
Circle = minus one point for Black
Cross = minus two points for Black
There are 13 plus points for Black, but 16 minus points, so KataGo's variation is three points better than Our Solution.
Our Solution ended with three points for Black, so KataGo's Semeai Variation results in a drawn game. This is only ONE point worse than KataGo's Capture Variation.
+ + + + + + + + + +
An application case:
Let us assume that -- by some magic -- Black was able to reduce White's territory in the upper right by an additional point (without compensation), so that the Capture Variation would end in jigo.
However, Black invested an additional stone, so White will gain an additional prisoner = point when choosing the Semeai Variation, which then will give her the same one point win that she had before.
White will choose the Capture Variation to realise her territorial gains at the left, due to KataGo's new atari / capture / connection.
However, if you want to evaluate variations, you have to be aware that the Semeai Variation might be nearer than expected. I had to learn this painfully ...
In the diagrams above, we compare the final position of the Semeai Variations after KataGo's new moves (at the left) with that of Our Solution (at the right).
Legend:
Triangle = plus one point for Black
Square = plus two points for Black
Circle = minus one point for Black
Cross = minus two points for Black
There are 13 plus points for Black, but 16 minus points, so KataGo's variation is three points better than Our Solution.
Our Solution ended with three points for Black, so KataGo's Semeai Variation results in a drawn game. This is only ONE point worse than KataGo's Capture Variation.
+ + + + + + + + + +
An application case:
Let us assume that -- by some magic -- Black was able to reduce White's territory in the upper right by an additional point (without compensation), so that the Capture Variation would end in jigo.
However, Black invested an additional stone, so White will gain an additional prisoner = point when choosing the Semeai Variation, which then will give her the same one point win that she had before.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
lightvector
- Lives in sente
- Posts: 759
- Joined: Sat Jun 19, 2010 10:11 pm
- Rank: maybe 2d
- GD Posts: 0
- Has thanked: 114 times
- Been thanked: 916 times
Re: KataGo @ Igo Hatsuyoron 120
Yeah, I had noticed in some of my explorations with KataGo that actually sometimes it chose the semeai variation (as winning for white) if you sufficiently much had black deviate in a few spots.
The thing is, some of white's gains on the left side costs white some turns or some liberties, but don't actually lose very much in the case that ultimately the semeai variation is chosen. If black spends those extra liberties pushing in the upper right, black improves the capture variation, but those moves *lose* points on the semeai variation.
So given that most of black's ability to balance white's gains comes at the cost of making one variation worse to improve the other, whereas some of white's gains just purely make one variation better without costing in the other, black starts to get squeezed in both variations.
The thing is, some of white's gains on the left side costs white some turns or some liberties, but don't actually lose very much in the case that ultimately the semeai variation is chosen. If black spends those extra liberties pushing in the upper right, black improves the capture variation, but those moves *lose* points on the semeai variation.
So given that most of black's ability to balance white's gains comes at the cost of making one variation worse to improve the other, whereas some of white's gains just purely make one variation better without costing in the other, black starts to get squeezed in both variations.
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
A 50 per cent approach-move liberty
Did you ever wonder why the score difference between the Capture Variation and the Semeai Variation with KataGo's new ( & amateurish enhanced) solution is odd (W + 1 vs. Jigo), while this difference was even with Our Solution (B + 5 / B + 3), as well as decades ago?
Well, the cause is the marked cutting point in the center, which is only valid to a limited extent. It will vanish, if White chooses the Capture Variation with a move at "a", but it will turn into an approach-move liberty, if White follows the Semeai Variation by continuing with a move at e.g. "b".
With the Capture Variation, this intersection becomes one point of White territory.
With the Semeai Variation, however, due to Black's eye at O13 (and his connection at L6, and White's vanished eye in the centre at M12), White is able to approach at c AND d, only after connecting at the crossed board point before. This connection costs her one point of territory, but in the Semeai Variation only.
Did you ever wonder why the score difference between the Capture Variation and the Semeai Variation with KataGo's new ( & amateurish enhanced) solution is odd (W + 1 vs. Jigo), while this difference was even with Our Solution (B + 5 / B + 3), as well as decades ago?
Well, the cause is the marked cutting point in the center, which is only valid to a limited extent. It will vanish, if White chooses the Capture Variation with a move at "a", but it will turn into an approach-move liberty, if White follows the Semeai Variation by continuing with a move at e.g. "b".
With the Capture Variation, this intersection becomes one point of White territory.
With the Semeai Variation, however, due to Black's eye at O13 (and his connection at L6, and White's vanished eye in the centre at M12), White is able to approach at c AND d, only after connecting at the crossed board point before. This connection costs her one point of territory, but in the Semeai Variation only.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
Sorry, but this is not how the problem works!lightvector wrote:The thing is, some of white's gains on the left side costs white some turns or some liberties, but don't actually lose very much in the case that ultimately the semeai variation is chosen. If black spends those extra liberties pushing in the upper right, black improves the capture variation, but those moves *lose* points on the semeai variation.
So given that most of black's ability to balance white's gains comes at the cost of making one variation worse to improve the other, whereas some of white's gains just purely make one variation better without costing in the other, black starts to get squeezed in both variations.
Only territorial gains and losses that are located OUTSITE the Main Semeai area (i.e. not in the immediate vicinity of White's large group on the left nor of Black's large group in the upper right) have an identical effect on both Main Variations (Capture and Semeai Variation).
White gained an absolute advantage of three points (B + 5 / B + 3 ==> B + 2 / Jigo).
Opponent's territory can be also destroyed in an identical manner in both Main Variations IN the Main Semeai area, but these destructions are costly.
In principle, the destructor has to pay in LIBERTIES in the CAPTURE Variation, but in POINTS of territory in the SEMEAI Variation.
White gained a temporary advantage of another five points (B + 2 / Jigo ==> W + 3 / W + 5).
In the CAPTURE Variation, White's connection of her single stone at the left gained her two points.
However, she also lost liberties in the process, so Black is able to connect his single stone in the centre, as well as creating an eye (with White prisoner) in the top right, spending this won liberties for his own territorial gain.
In total, White lost two points from her temporary advantage (W + 3 ==> W + 1).
In the SEMEAI Variation, White spent four additional stones for destroying Black territory at the left, but Black added only three stones to his large group in the upper right.
Additionally, due to Black's eye, White has to occupy all cutting points around Black's large group, as well as Black's eye point, so losing another three points in the process.
In total, White lost five points from her temporary advantage (W + 5 ==> Jigo).
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
dany
- Dies with sente
- Posts: 121
- Joined: Fri Dec 20, 2019 2:44 am
- Rank: IGS 4 dan
- GD Posts: 0
- Has thanked: 36 times
- Been thanked: 12 times
Re: KataGo @ Igo Hatsuyoron 120
144 White's move should be in M11 instead of M12. its one point better for White in Semeai Variation.Cassandra wrote:A 50 per cent approach-move liberty
Did you ever wonder why the score difference between the Capture Variation and the Semeai Variation with KataGo's new ( & amateurish enhanced) solution is odd (W + 1 vs. Jigo), while this difference was even with Our Solution (B + 5 / B + 3), as well as decades ago?
...
With the Semeai Variation, however, due to Black's eye at O13 (and his connection at L6, and White's vanished eye in the centre at M12), White is able to approach at c AND d, only after connecting at the crossed board point before. This connection costs her one point of territory, but in the Semeai Variation only.
so in both variations the result will be W + 1
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
Thanks a lot, dany. When thinking about the consequences of Joachim's tesuji of O14 (instead of O12), the Semeai Variation was not yet on our radar ...dany wrote:144 White's move should be in M11 instead of M12. its one point better for White in Semeai Variation.
so in both variations the result will be W + 1.
Joachim will be very happy now, as he found different scores in both Main Variations a bit disappointing.
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)
-
alreadydone
- Beginner
- Posts: 4
- Joined: Fri Aug 16, 2013 1:24 pm
- GD Posts: 0
- Been thanked: 5 times
Re: KataGo @ Igo Hatsuyoron 120
This is maybe OT, but has anyone tested DNN-based bots (Leela Zero, KataGo, SAI, etc.) on the whole-board endgame problems (attached) in Appendix C of [BW94], including the 9-dan stumping problem [0] and many 9x9 problems? (LZ may be less appropriate due to fixed 7.5 komi.) Additional problems are found in:
http://library.msri.org/books/Book29/files/kim.pdf
http://library.msri.org/books/Book56/fi ... kamura.pdf (4.2. Whole board problems)
https://webdocs.cs.ualberta.ca/~mmuelle ... board.html
https://webdocs.cs.ualberta.ca/~mmuelle ... hesis.html (compressed pdf, A.3, modifications of [BW94] problems; Chapter 7 describes the program Explorer)
https://www.dumbo.ai.kyutech.ac.jp/~tei ... 1slide.pdf (Problems 3 and 4)
I wonder how these problems designed with mathematical theory differ from those from traditional Go practice. I think they likely also require specific training (like what lightvector did for Hatsuyoron 120) to solve. The search would need to be extremely wide and the number of variations/branches may be enormous due to lack of forced moves.
The attachments contain solutions to the [BW94] problems, but only the first moves. It's said in the paper [1] that "Our program Explorer uses this tool to play full board endgames. It plays a nontrivial class of endgame positions perfectly"; Explorer can be downloaded at [2] but only runs in Mac. Also, I can't find SGF files for the "Perfect Play" and "Node Rooms" collections (see [3]) in Mueller's Computer Go Test Collection [4]. Any idea?
[BW94] Berlekamp--Wolfe 1994, Mathematical Go: Chilling Gets the Last Point
[0] https://zhuanlan.zhihu.com/p/33860825 (in Chinese)
[1] https://webdocs.cs.ualberta.ca/~mmuelle ... dgames.pdf (Optimizations as described in 4. Local Move Generation are potentially necessary to solve the problems besides the AlphaZero/KataGo algorithms.)
[2] http://www.computer-go.info/db/oprog.php?a=Explorer or https://www.applefritter.com/node/13972
[3] https://webdocs.cs.ualberta.ca/~mmuelle ... endix.html
[4] https://webdocs.cs.ualberta.ca/~games/go/cgtc/
http://library.msri.org/books/Book29/files/kim.pdf
http://library.msri.org/books/Book56/fi ... kamura.pdf (4.2. Whole board problems)
https://webdocs.cs.ualberta.ca/~mmuelle ... board.html
https://webdocs.cs.ualberta.ca/~mmuelle ... hesis.html (compressed pdf, A.3, modifications of [BW94] problems; Chapter 7 describes the program Explorer)
https://www.dumbo.ai.kyutech.ac.jp/~tei ... 1slide.pdf (Problems 3 and 4)
I wonder how these problems designed with mathematical theory differ from those from traditional Go practice. I think they likely also require specific training (like what lightvector did for Hatsuyoron 120) to solve. The search would need to be extremely wide and the number of variations/branches may be enormous due to lack of forced moves.
The attachments contain solutions to the [BW94] problems, but only the first moves. It's said in the paper [1] that "Our program Explorer uses this tool to play full board endgames. It plays a nontrivial class of endgame positions perfectly"; Explorer can be downloaded at [2] but only runs in Mac. Also, I can't find SGF files for the "Perfect Play" and "Node Rooms" collections (see [3]) in Mueller's Computer Go Test Collection [4]. Any idea?
[BW94] Berlekamp--Wolfe 1994, Mathematical Go: Chilling Gets the Last Point
[0] https://zhuanlan.zhihu.com/p/33860825 (in Chinese)
[1] https://webdocs.cs.ualberta.ca/~mmuelle ... dgames.pdf (Optimizations as described in 4. Local Move Generation are potentially necessary to solve the problems besides the AlphaZero/KataGo algorithms.)
[2] http://www.computer-go.info/db/oprog.php?a=Explorer or https://www.applefritter.com/node/13972
[3] https://webdocs.cs.ualberta.ca/~mmuelle ... endix.html
[4] https://webdocs.cs.ualberta.ca/~games/go/cgtc/
- Attachments
-
- Mathematical Go solutions.pdf
- (215.82 KiB) Downloaded 693 times
-
- mathematical Go problems.pdf
- (1.2 MiB) Downloaded 690 times
-
Cassandra
- Lives in sente
- Posts: 1326
- Joined: Wed Apr 28, 2010 11:33 am
- Rank: German 1 Kyu
- GD Posts: 0
- Has thanked: 14 times
- Been thanked: 153 times
Re: KataGo @ Igo Hatsuyoron 120
The Variation Tree of our website has been updated:
KataGo's New Move Suggestions
Please let me know if you find any issues ...
KataGo's New Move Suggestions
Please let me know if you find any issues ...
The really most difficult Go problem ever: https://igohatsuyoron120.de/index.htm
Igo Hatsuyōron #120 (really solved by KataGo)
Igo Hatsuyōron #120 (really solved by KataGo)