I think I slightly misread the idea though. I thought you would program the puzzle based on available info when a game finishes. I don’t think the ogs dev needs to do this right? So I still think it’s a great idea (and in bumping this same thread again for you!) And I hope you can create this tool.
Thank you! Yes i would program the puzzle based on available info when a game finishes. Similar to what I have already done and described in my collection: Play Go at online-go.com! | OGS
Maybe I can help with the development of the automatic generation of “Breakpoint → Puzzle”. In any case, I would be very happy to be able to work on the development of onlinego. Is that question in the right place here? Or where do I have to send my application? I would be very grateful for an answer or even an acceptance!
The idea sounded interesting to me at first, but on second thought I wonder:
What use will knowing such a “better” AI move have for me if I don’t know the right follow-up moves to play?
This seems to be a general problem when trying to imitate our new AI overlords.
Use for investigating. Not all AI moves are like mystery. AI is giving you follow up sequences too
Another possibility, which I would really like, would be if, while creating puzzles, you had a function or a button that linked you to the game situation you have built (and for which you of course also add a Komi value),
calculates the winchance and points to be expected for both players; calculated by the strongest AI and displayed while using the puzzle-editor.
So even beginners can develop really interesting puzzle-collections and tutorial-puzzle-series.
At the same time help in an efficient way to understand the steps of the new superhuman bots.
It’s time to use them to understand more what the game Go is about!
To be able to learn and to want to teach even better.
Example: Compare situations: Search for their causes as well as similar situations with the same cause for decision-making or generate them and sort them to a public puzzle-collection. →
Teaches in an efficient way to recognize this game situation or to bring it about and then to be able to make the right decisions or to be able to set a certain trap for the opponent.
And as already said: I could try to help with the development.
And I’m already super grateful to you, the developers of Onlinego, for still developing a Go puzzle editor for everyone to learn and teach, and making it available free of charge.
That will probably be the main reason why Go has become my favorite strategy board game (before Chess, Tak and Mill).
In another way it was “AlphaGo - The Movie” …
One more time I would like to try to describe and offer my hopes for a much more extensive, highly interesting and for the onlinego.com site challenging go puzzle editor, if necessary, to work on the programming (of course, free of charge and out of pure enthusiasm). I have already described my first two suggestions here:
Automatic generation of breakpoint puzzles.
Possibility to calculate the number of points, the chance of winning and suggestions for subsequent moves while editing your own puzzles.
The networking of puzzles:
For example to specifically improve the possibilities for the following structure:
Choose the best from all possible moves. If you choose a wrong move, you lose the game and should start over. If you choose correctly each time until the end, you win the game and you get the solution letters from a total of three variants in which this puzzle can lead you, by random selection. To be able to play the next level or puzzle: Search for a puzzle with the name Code1Code2Code3. Replace CodeX with the found solution letters from game variantX!
Everyone who wants to is hereby cordially invited to take part in the expansion of a maze of Go puzzles of all kinds that can be structured in this way. A little communication is of course required for a freely networked expansion. Otherwise you can also build your own puzzle, through which players can alternatively receive the codes that you have already found out!
- Copying, caching, dividing, pasting, cutting and relinking of puzzles, puzzle parts, collections and tutorial series.
I would be very happy about your answer as well as any suggestions, tips, links and discussions regarding this topic and I am grateful for everything and everyone who answers me and thus helps to realize my dreams.
Talking about puzzles, it seems it’s not possible to have multiple right answers, am i right? Quite a restriction.
Bei mir ist es so, dass ein Rätsel zu Ende ist sobald ich mich bis zu einem Punkt in den Verzweigungen gespielt habe, welchen ich als “richtige Antwort” gekennzeichnet habe → Rätsel gelöst. Erzeuge ich eine Spielsituation, welche ich entweder als “falsche Antwort” gekennzeichnet habe oder die auf Grund der Möglichkeit von “Freier Platzierung” nicht in den im Rätsel erstellten Verzweigungen enthalten ist, endet das Rätsel ebenfalls → “Falsche Antwort”.
For me it is the case that a puzzle is over as soon as I have played my way up to a point in the branches, which I have marked as the “correct answer” → puzzle solved. If I create a game situation which I have either marked as “wrong answer” or which is not included in the branches created in the puzzle due to the possibility of “free placement”, the puzzle also ends → “wrong answer”.
Biy what whem there are different ways to a right answer or even different right answers? It happens in some problems.
That’s correct! As I said before, I think it’s extremely good that onlinego has a free puzzle editor. I also see great potential to further develop it. At the moment you still have to create each path individually and mark each end as the “correct answer” (even if it is the same board position or an identical rotation).
Exactly how you can use the functions: “Estimate points”, “Copy branch” and “Insert branch” not (yet) in the puzzle editor but only in the game review. Still you have to open this in an additional tab for this.
Furthermore, in puzzles, “pass” cannot be selected as a possible move. This is a missing option and would also be needed in some problems.
Hello, firstly I would like to keep this suggestion up to date and secondly I would like to suggest that there is the “estimate points” button not only in the game review but also when editing and solving Go puzzles.
I’m not aware that KataGo (or any other AI?) has partial SE.
I don’t even get what you’re trying to do.
what do you think of my ideas? For your feedback or your own suggestion, what could be easily implemented and at the same time make online-go super interesting for yourself and other players…
Thank you for every comment that draws attention to this topic, which is very important to me, and keeps communication and discussion alive.
I think it would be super exciting if you could load every breakpoint from past games and play it to the end against Katago-micro. And that with the possibility of undoing moves and displaying tips (e.g.: Best move? from Katago-micro). Then the various automatically generated breakpoint puzzles should also have their level of difficulty automatically generated: depending on how often players had to undo their moves and had to be given tips. Their type can be generated by player voting. Likewise, these could be freely copyable and further developed by players. It would also be an idea of mine to add an Evaluate Situation button as an alternative tip to this new form of Go training.
Best regards Andreas Schneider
An idea of mine based on this would be an adaptive AI that distributes the breakpoint puzzles to the players in such a way that as many breakpoint puzzles as possible
to be played
be solved, evaluated, commented on, copied, further developed, optimize the learning effect in go, …
There is an Estimate Points button in the gamereview after each finished game you’ve played.
There is already an Estimate Points button in the gamereview after each finished game you’ve played. I would like to have this “estimate points” function in other places as well.
Of that I am aware.
Read again: I wrote that I am not aware that KataGo has PARTIAL SE.
Maybe you can try and inspire some of those who actually develop Go AIs if you do not intend to DIY?
I have trouble understanding what Andreas means by a “breakpoint” (to me it is a point in a programme where execution is stopped to allow the developer to investigate its behaviour). But anyway, Benjamin Teuber’s site AI Sensei does a a good job of creating problems from your games, which may help some people interested in Andreas’ idea, even if the approach is not what Andreas intends.