Ask Go players about Go (come play Go with us)

The Something Awful Forums > Discussion > Ask/Tell > Ask Go players about Go (come play Go with us)

«‹›8 »

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

GoQuest maybe? I haven't tried it but my friends were talking about it.

# ¿ Feb 17, 2019 21:57

Adbot: ADBOT LOVES YOU

# ¿ May 13, 2024 18:51

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

Huh, sounds like a game changer for learning Go. Compare this post from 2011:

Under 15 posted:

Anyone who did any good at Go probably played a lot of ranked 19x19 games early on. [�] There's nothing special in particular about playing ranked 19x19 games. If you really want to get strong at go, you'll do a lot of problems and 9x9 before even thinking about the big board, but problems are boring and it's hard to find 9x9 games at all, let alone quality ones.

# ¿ Feb 21, 2019 14:37

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

To be fair, for this sort of discussion, between posters who aren't online all the time, the chatroom would be like playing in the street, just asking to be made roadkill by any random fast-moving off-topic chat.

# ¿ Feb 26, 2019 02:19

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

derp posted:

discordd link in the OP no longer works. is there a new one?

https://discordapp.com/invite/aWTrnv4

# ¿ Jul 27, 2019 01:39

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

David Wu posted:

https://blog.janestreet.com/deep-learning-the-hardest-go-problem-in-the-world/

As far as we are aware, all prior bots including those developed since AlphaZero, almost completely fail to understand this problem. The highly unusual shapes cause them to completely miss many of the key moves and ideas, and the incredible precision of the fights prevents the generic knowledge that they have learned from being effective. In fact, even in ordinary games, long-distance/large-scale fights and blind spots of specific unusual shapes are known weaknesses of current otherwise superhuman bots, and sometimes humans can outdo them in those situations�and this problem hits both such weaknesses hard.

Could a bot be trained to understand this problem?

Normally, training consists of self-playing hundreds of thousands of ordinary games. We learned from AlphaZero that this makes the bot very, very good at ordinary games.

So what if self-play consisted instead of exploration of this specific problem? Would the bot then master this problem?

# ¿ Jan 5, 2020 06:16

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

TheCog posted:

I made an account on KGS with username thecog, can I get an invite to the room?

We no longer have our own KGS room but we do have a Discord: https://discordapp.com/invite/aWTrnv4

# ¿ Feb 22, 2020 18:11

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

Quinn posted:

Every step I took into this world I felt was taking me away from my friends and family
I would have an amazing evening with Go and I didn't even know how to express that to my friends
What I like about the hobby of boardgames is that it makes me feel closer to my friends
Go made me feel farther apart from my friends
I would need new friends
I would need Go friends

powerful

# ¿ Jul 31, 2020 03:21

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

# ¿ Aug 5, 2020 01:28

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

https://brantondemoss.com/writing/kata/

Good article on KataGo, the latest improvement on the AlphaGo method*.

*which was independently discovered and presented with the more descriptive name "expert iteration" in a paper on Hex also published in 2017.

# ¿ Oct 11, 2020 16:57

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

Wikipedia posted:

In February 2022, after training KataGo for 11 months, [Igo Hatsuyōron 120] appears solved: the main variation has 205 moves and finishes with a White win by one point, assuming White begins the game and there are no prisoners initially. The score does not depend on the rule set, with the same result with Japanese, Chinese, Tromp-Taylor or New Zealand scoring. Though it cannot be certain that the solution is perfect, it is an improvement over previous solutions. KataGo took less than 4 months to figure out the best sequence of moves, and despite seven more months of training and multiple actions to encourage it to explore new lines of play, the best line of play remained unchanged, KataGo only consolidating its view of the problem. The expert Thomas Redecker contributed thousands of Igo's position diagrams to train KataGo and shares this view that the problem now appears to be solved.

Apparently it was thought to be a win for black until the AI analysis, though some discoveries had changed Black's winning score. Fujisawa Hideyuki (Shuko), who handled the manuscript discovered in 1982, had added the caption "Black to play and win".

(In the quotation, "assuming White begins the game and there are no prisoners initially" refers to the initial position's White stones outnumbering Black by 1, implying Black to play.)

Redecker's website: https://igohatsuyoron120.de/

# ¿ Jul 31, 2022 03:17

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

I found a group via MeetUp before the pandemic.

# ¿ Dec 13, 2022 22:02

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

# ¿ Feb 14, 2023 10:07

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

Long ago when I was 3-kyū, I posted a guide for newbies in this thread opining on what to learn first, and I can't believe I never thought of posting it directly on Sensei's Library until now.

There, now it's easier to link non-goons to: https://senseis.xmp.net/?Xom#toc1

# ¿ Feb 27, 2023 11:35

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

I just realized I never posted this yet: https://archive.is/tDYsY

FT posted:

Man beats machine at Go in human victory over AI

Amateur Kellin Pelrine exploited weakness in systems that have otherwise dominated board game�s grandmasters

KataGo and other Go AI's trouble counting the eyes of cyclic groups (groups surrounding an opposing group) was a known issue. In January, Pelrine became the first to cheese wins out of the weakness.

More recently, a challenge to capture all of KataGo's stones:
https://forums.online-go.com/t/eat-katagos-everything/47860
https://online-go.com/game/53421988

KataGo developer is still making progress teaching KataGo to evaluate cyclic groups by hand-curating examples.

# ¿ May 18, 2023 18:36

Adbot: ADBOT LOVES YOU

# ¿ May 13, 2024 18:51

Xom: Sep 2, 2008; 文化英雄; Fan of Britches

Hikaru no Go stones

Only registered members can see post attachments!

# ¿ Jul 22, 2023 23:18

The Something Awful Forums > Discussion > Ask/Tell > Ask Go players about Go (come play Go with us)

«‹›8 »