It is currently Thu Oct 31, 2024 4:13 pm

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 10 posts ] 
Author Message
Offline
 Post subject: analyzing a dwyrin basic game with KataGo
Post #1 Posted: Sat Mar 27, 2021 1:09 pm 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00

Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #2 Posted: Sat Mar 27, 2021 3:20 pm 
Lives in gote

Posts: 495
Location: Netherlands
Liked others: 272
Was liked: 147
Rank: EGF 3d
Universal go server handle: gennan
TBH, I think it's not really fair to expect Dwyrin's play to be both basic and approved by KataGo at the same time.

I can imagine that 27 would make sense for a cautious 10k player. Defending your weakest group while taking some territory is good advice in general for weaker players.
I agree that in this case it's too passive for a dan player, because black's group isn't that weak. But that assessment may be hard to make by a 10k. So I think it qualifies as playing "basic".

Move 29 may be suboptimal by AI gospel, but I do think it's quite basic. I wouldn't be very surprised to see this move in an older pro game (ignoring the upper side). I can see that J16 is the best move for an AI, but that doesn't look very basic to me.

I agree that move 35 isn't really basic. That attachment is so advanced that humans needed AI to tell it's good. Or is this attachment already common at the "basic" player level?

I have no explanation for move 37 and 39. I don't see any merits of these moves, educational or otherwise. Perhaps move 37 might be OK for a 10k, but move 39 is just a very bad exchange (and thus giving a bad example).

Move 83 and the continuation looks too greedy. 10k players might be reluctant to sacrifice (or they might not even consider it), so I think this was a missed educational opportunity (sacrifice stones that serve no purpose).

Move 113 is a bit overaggressive, but that's not a big deal from an educational standpoint, I think.

Overall, I feel he did a decent job playing a "basic", educational game here. The only moves I really disapprove of are moves 39 and move 83. Those 2 moves were just bad, whichever way you look at it.


Last edited by gennan on Sat Mar 27, 2021 3:47 pm, edited 1 time in total.
Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #3 Posted: Sat Mar 27, 2021 3:46 pm 
Oza

Posts: 2356
Location: Ireland
Liked others: 662
Was liked: 442
Universal go server handle: Boidhre
:b27: is a very intuitive move for ddks. I think it's one of those "play here if you think the above group is weak, but approach on the bottom if you think you can fight with the group above" which is a nice teachable moment. Approaching is only correct if you're confident that the group above is strong. I would not mind this in a basic series in this context. It makes no sense alongside :b35: though. The player who will see :b35: is not the player who will play :b27:

I don't really think how Katago analyses the moves matters much. More whether there are solid concepts being taught and the level is even.

Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #4 Posted: Sat Mar 27, 2021 4:02 pm 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
gennan wrote:
TBH, I think it's not really fair to expect Dwyrin's play to be both basic and approved by KataGo at the same time.


It's easy when you have KataGo on your side. I don't expect him to be, I just wanted to check that wherever he was confident his basic moves where giving him a good game, he actually was having one. And in many cases he was.

gennan wrote:
I can imagine that 27 would make sense for a cautious 10k player. Defending your weakest group while taking some territory is good advice in general for weaker players.


Hmm, I don't see that group as weak though. Not superstrong either, just quite influential. But I'll give that to him, extending from influence is basic.

Quote:
I agree that in this case it's too passive for a dan player, because black group isn't that weak. But that assessment may be hard to make by a 10k. So I think it qualifies as playing "basic".


His target audience here was 2d, so I'm a little more demanding :)


Quote:
Move 29 may be suboptimal by AI gospel, but I do think it's quite basic. I wouldn't be very surprised to see this move in an older pro game. I can see that J16 is the best move for an AI, but that doesn't look very basic to me.


Fair enough. He teaches the equivalence of enclosure and extension as a kind of gospel and it really isn't (anymore). I think corner, sides, center is just as basic and it's also more often correct.

Quote:
I agree that move 35 isn't really basic. That attachment is so advanced that humans needed AI to tell us it's good. Or is this attachment already common at the "basic" player level?

I have no explanation for move 37 and 39. I don't see any merits of these moves, educational or otherwise. Perhaps move 37 might be OK for a 10k, but move 39 is just a very bad exchange.

Move 83 and the continuation looks too greedy. 10k players might be reluctant to sacrifice (or they might not even consider it), so I think this was a missed educational opportunity (sacrifice stones that serve no purpose).


Agree there and he actually did too, so not too unhappy with his message there.

Quote:
Move 113 is a bit overaggressive, but that's not a big deal from an educational standpoint, I think.


Tewari easily shows the connection was better but the opponent found a crafty move and that's not easy to respond to in a basic fashion.

Quote:
Overall, I feel he did a decent job playing a "basic", educational game here. The only moves I really disapprove of are moves 39 and move 83. Those 2 moves were just bad, whichever way you look at it.


Thanks for looking into it.

Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #5 Posted: Sat Mar 27, 2021 8:36 pm 
Lives in sente

Posts: 758
Liked others: 114
Was liked: 916
Rank: maybe 2d
I could probably make a "basics" KataGo to some degree, with a bit of fiddling.

* Add a bonus for reducing the uncertainty in the ownership prediction (so, prefer sequences leading to both sides having settled shapes quickly)
* Add a pointwise penalty to any moves that are much lower than the top move KataGo's own policy or that fluctuated in score or value a lot as it searched (if KataGo itself required a bunch of search to change its mind, it definitely wasn't basic).
* Also do it within the search tree (so also prefer moves that lead to sequences where nothing surprising will need to happen for that move to turn out well)
* Add a small penalty for playing far from the opponent's move (so, prefer to respond locally and avoid tenuki)
* Add a small per-stone penalty for killing opponent's stones (so, prefer peaceful lines and disprefer trades)

It will probably still play some exotic moves that are bot-intuitive but not human-intuitive, but overall I bet with some tuning you could end up with a fairly peaceful style. Backed up, of course, by inhumanly good global judgment that lets it pick the "right" peaceful variation that always gives a good result and only gives up a slight amount of points.

There's some chance this is very hard, but there's also a chance of it just working after enough fiddling with a mix of the above. Would need to try it out to know.


This post by lightvector was liked by 4 people: ez4u, gennan, Knotwilg, MikeKyle
Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #6 Posted: Sun Mar 28, 2021 2:03 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
For 27 extension, what words of explanation accompanied it? If it was along the lines of "For 24 white normally plays e13 d12 exchange in sente which helps grow white's top but also helps make black stronger on left, and then tenukis. Black's group is safe, has d17 cut to capture the corner stone if it needs eyes. But because he blocked at b17 I don't have d17 insurance and didn't turn e13 white is aiming at c12 peep and attacking my group. So I am going to extend to make it safe." then it's fine. That's the lesson here, extend to ensure safety because white's choice of moves reduced your eyespace. It is rather tame for 2d basics (didn't know he did basics aimed at dans) but fine for 10ks.


This post by Uberdude was liked by: Boidhre
Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #7 Posted: Sun Mar 28, 2021 2:12 am 
Gosei
User avatar

Posts: 1769
Liked others: 180
Was liked: 492
I note that no one defined what "basic" means. Here is an attempt:

1) Follow simple ideas like proverbs, no complicated strategies.
2) Use standard shapes.
3) Do not play moves that require deep reading.

Is it possible to play at dan level just by playing in a "basic" way in the above sense?

Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #8 Posted: Sun Mar 28, 2021 2:41 am 
Lives in gote

Posts: 495
Location: Netherlands
Liked others: 272
Was liked: 147
Rank: EGF 3d
Universal go server handle: gennan
jlt wrote:
I note that no one defined what "basic" means. Here is an attempt:

1) Follow simple ideas like proverbs, no complicated strategies.
2) Use standard shapes.
3) Do not play moves that require deep reading.

Is it possible to play at dan level just by playing in a "basic" way in the above sense?

I think it is. To play at dan level, I think it's more than enough fi you just avoid big mistakes and really bad moves, while playing good endgame.
Moves like move 27 that are passive, but just lose 1-2 points, are fine, as long as you don't do it every move. You might even play at 3-5d EGF level or better like this.
Even before AI existed, I was told that if you're able to play the "normal move" in every situation, you're already pro level. An example may be Lee ChangHo at his peak. He played kind of boring and never played anything fancy and he was world champion for a decade. He just had phenomenal positional judgement and great endgame, so he could confidently play simple moves and beat arguably more "genius" (and more extravagant) players, such as Cho Hunhyun.

Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #9 Posted: Sun Mar 28, 2021 8:26 am 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Uberdude wrote:
For 27 extension, what words of explanation accompanied it? If it was along the lines of "For 24 white normally plays e13 d12 exchange in sente which helps grow white's top but also helps make black stronger on left, and then tenukis. Black's group is safe, has d17 cut to capture the corner stone if it needs eyes. But because he blocked at b17 I don't have d17 insurance and didn't turn e13 white is aiming at c12 peep and attacking my group. So I am going to extend to make it safe." then it's fine. That's the lesson here, extend to ensure safety because white's choice of moves reduced your eyespace. It is rather tame for 2d basics (didn't know he did basics aimed at dans) but fine for 10ks.


Yes it was.

Top
 Profile  
 
Offline
 Post subject: Re: analyzing a dwyrin basic game with KataGo
Post #10 Posted: Sun Mar 28, 2021 8:27 am 
Oza
User avatar

Posts: 2420
Location: Ghent, Belgium
Liked others: 359
Was liked: 1020
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
jlt wrote:
I note that no one defined what "basic" means. Here is an attempt:

1) Follow simple ideas like proverbs, no complicated strategies.
2) Use standard shapes.
3) Do not play moves that require deep reading.

Is it possible to play at dan level just by playing in a "basic" way in the above sense?


Here are dwyrin's basics. And his answer to your question is yes.

https://senseis.xmp.net/?Dwyrin

Dwyrin's basics are, non-exhaustively

first corners, then sides, then center
territory rather than influence - influence is harder to play with
protect your weak groups first
when "behind enemy lines", defend - usually by jumping across the sector line or making a base
weakness depends on proximity of own/enemy stones, clear weak points or cutting points

make shape and poke at your opponent's shape
aim for sente - always ask yourself: "do I have to respond here?"
find the biggest area - an easy heuristic here is the number of empty points between two stones (usually corner positions)
an enclosure and its extension are miai: if your opponent takes one, you take the other
when invading, look for multiple options (jumping out, making a base)
there is no reason to fight. Usually the thing (a stone, a piece of territory) is not worth fighting over. Just take larger points
the outcome of a fight is unpredictable; in particular you might lose sente and it's the opponent who takes a large point
you can't always back off though: when your opponent does something unreasonable, he shouldn't get away with that easily

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 10 posts ] 

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: Bing [Bot] and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group