Practice game: Dead, unsettled or alive?

[jlt]

Suppose Black is a connected group attacked by White, without outside liberties, and has 2 internal liberties.

• If White can almost fill the eyeshape with a killing shape then Black is dead.
• Otherwise Black is alive.

Suppose Black is a connected group attacked by White, without outside liberties, and has 3 internal liberties.

• If White cannot almost fill the eyeshape with a killing shape then Black is alive.
• If White can almost fill the eyeshape with a killing shape, and if all shapes except possibly one consisting of adding just one white stone are killing shapes, then Black is dead.
• Otherwise the group is unsettled.

[shinuito]

@jlt

• If White can almost fill the eyeshape with a killing shape then Black is dead.
• Otherwise Black is alive.

Well

@shinuito

(Add a stone at the 1-2 point) and the bent four. Not one of these killing shapes, unless you add a special “in the corner” cases.

[jlt]

You quoted my paragraph with Black having 2 internal liberties but posted an example with 3 internal liberties.

P.S. I consider the rectangular 6 in the corner as a killing shape.

[shinuito]

I know, but I said, add the 1-2 point.

Practice game: Dead, unsettled or alive?

When analyzing a pure nakade shape (ignoring things like the special properties of the corner), the key question to ask is whether it’s possible to almost fill the eyeshape with a killing shape. [image] (It sounds like @david265 is giving an explanation for why these shapes in particular are killing - I’m just linking them for reference, and the term “almost fill” is useful to know about)

I would probably call these killing shapes as @antonTobi mention because they’re likely to work most places, while the six point one doesn’t work unless it’s in the corner with the outside liberties filled.

Similarly with the bent four in the corner vs everywhere else - it doesn’t seem to make sense to call it a killing shape

[david265]

@jlt

you seem to forget that a group can live by seki.

Yes, I’m not sure, but I think there is a good algorithm for noticing and avoiding or aiming for seki early, based on the fact that a group in seki always touches another group with exactly two separate liberties. This one needs more thinking.

[Groin]

@david265

exactly two separate liberties.

No.

@david265

This one needs more thinking

Seems so.

[Gia]

I’m not following the technical discussion because it’s above my playgrade, but I’ve done the most reading outside of an actual game since… probably ever.

This thing is fun.

[Glunkolin]

How about some more challenging shapes?

[antonTobi]

Great suggestions, thanks! I will add these in a future version.

If anyone else wants to make some suggestions, feel free to share them here in any format that is easiest for you to create.

I said earlier…

@antonTobi

Just give me some time to set up the problem creation tool so that you can export the file and send it to me when it’s done.

…but since I anyways need to fit everything together eventually we might as well skip that step. Entering some positions manually from images is pretty quick on my end.

[Groin]

What about a bit of “under the stones” + nakade positions?

[Feijoa]

@jlt

Suppose Black is a connected group attacked by White, without outside liberties, and has 3 internal liberties.

• If White cannot almost fill the eyeshape with a killing shape then Black is alive.
• If White can almost fill the eyeshape with a killing shape, and if all shapes except possibly one consisting of adding just one white stone are killing shapes, then Black is dead.
• Otherwise the group is unsettled.

I was hoping for something like this, and it works really well for me. It’s interesting how you don’t have to consider Black moves at all!

But I think it assumes that White and Black are free to play on the three liberties and there are no other points available. So unless I’m misunderstanding it fails in some cases:

In the first most of the shapes made by adding a white stone are not killing shapes, but Black cannot prevent the formation of one and is dead.

The second satisfies your conditions for “dead” but if Black plays the middle point White is not able to play to form a killing shape, so it’s unsettled.

[jlt]

Thanks for pointing that out. Some conditions need to be added for my criteria to work. Something like none of the players having enough space to form an eye.

[_KoBa]

So how about those “basic ko’s” for unsettled shapes?

Can prolly make a ton of those with small tweaks ^^

[antonTobi]

Updates:

• Some more problems, about 200 in total now (including suggestions from @shinuito @Glunkolin and @_KoBa, thanks!)
• Problems are sorted by difficulty (1-4), and you will get easier/harder problems depending on how quickly you answer. Harder problems give more points.
• New time control: 1 min + 1 sec increment.
• An incorrect answer immediately ends your game, so that you can focus on learning the position you got wrong, instead of rushing to solve the next one. (still on the todolist: show last position more clearly and what its correct status was)
• “Endless mode” is now “Practice mode”, where you can select a difficulty and do as many problems as you like of that difficulty.

Try it out and let me know what you think! How high can you score with the new point system?

status.antontobi.com

It’s quite hard for me to accurately grade the difficulty - if you want to help me make the ratings better, go through some problems in practice mode and let me know if you find any that are too easy or too hard compared to the others at that level.

I still want to add many more problems - if you have ideas for something that you’d like to see included, please share some diagrams here in the thread!

[ArsenLapin1]

Would you consider making the collection available? I’d like to make flashcards with these shapes.

(Any format that I can process programmatically would be great, pictures or sgf or other textual description of the position)

[yebellz]

The collection of positions are embedded in the code (as the constant “allBoards”, within this file:

antonTobi/status/blob/main/boards.js

and it seems to have been released under an MIT license. The surrounding code might give enough hints about how to programmatically process this into another convenient format, but I guess @antonTobi may have originally curated them in some other format and used some sort of tool to generate this encoding.

[ArsenLapin1]

Awesome, thanks!

[Jon_Ko]

My high score is now 42. I’m done.

[ArsenLapin1]

I made the flashcards.

There is an Anki plug-in to display sgf interactively on flashcards, but it’s only for the iPhone version, so I didn’t use it.

I couldn’t find a satisfying python package to edit go diagrams, so I made my own with matplotlib. This is my first time making go diagrams with python, criticism is welcome.

The file: antontobi_status_weiqi.apkg - Google Drive

Unfortunately the file is 5MB due to the size of the diagrams. This sounds like a waste of space for go diagrams. I’ll try to find a way to make them lighter in the future.

EDIT: A slightly smaller version antontobi_status_weiqi_lowdpi.apkg - Google Drive

[shinuito]

There was another person trying to make anki flashcards with python

ALPHA Release: Flash Goban - make flashcards of KaTrain positions

Visit the github page to download and install… Flash Goban Make Anki flashcards from (KaTrain) goban positions [image] How does it work? Open up Anki (after proper installation, see below) Open up Flash-Goban Open up a game in KaTrain Scroll through until you see a position where you could’ve done much better alt-tab to Flash-Goban and click “make flashcard” A flashcard is made where the front of the flashcard is the game position with no move analysis and the back of the flashcard is the …

could be some collab options