Make Tsumego Great Again!

As before, that’s a matter of opinion so probably best to just add both.

@Samraku - what about the abbreviations? Do you accept my point that 30-20 kyu, etc are preferable over TPK, etc.?

No; the fact that TPK, DDK, SDK, and dan are used in conversation far far more than the delineation of ranks they represent, is extremely strong evidence that they are more useful terms: natlangs naturally develop to communicate efficiently and effectively.

Just a technicality, 20 kyu belongs to 2 categories.
Shouldn’t it be

1 Like

The (minor) thing I don’t necessarily like about this, is that a puzzle set at the moment gets a rating which takes the min and max ratings from all the puzzles in the set (so it looks like a range like this, but with more freedom on the endpoints).

So having a range of difficulties for the puzzles slightly clashes with the way the ratings for the set work at the moment.

1 Like

They’re just rough bands so it doesn’t really matter if there’s some overlap, but I suppose 19-10 kyu rather than 20-10 kyu might help to avoid confusion so probably is slightly better.

Clearly, we have another difference of opinion here. Can some other people comment on this point please? What is preferable?

  1. Explicit, unambiguous bands with obvious meanings (30-20 kyu, etc); or

  2. Abbreviations that may or may not be understood (SDK, etc) that refer to ambiguous terms (since TPK is usually a subset of DDK but here DDK is to refer to the usual DDK but excluding the TPK part)

Yes, I agree that this is not ideal. But I think it can work. Consider a set with a mixture of difficulties at individual ranks (e.g. 12 kyu, 10 kyu, etc) and rank ranges (19-10 kyu, 9-1 kyu, etc.) and descriptors (easy, medium, etc). One would have to take the minimum of the minimum rank and minimum rank range lower bound (in this case, 19 kyu) and the maximum of the maximum rank and maximum rank range upper bound (in this case, 1 kyu), and the minimum descriptor (easy) and the maximum descriptor (medium) to label the set, in this case as “19-1 kyu, easy-medium”. Seems OK to me. Other sets might only have the 19-1 kyu bit or only the easy-medium bit. In fact, most sets will probably only have one or the other as I suspect that sets will generally just use one type of difficulty label.

This is another reason why abbreviations are less good than explicit rank ranges IMO, as the example set would be labelled as “12-10 kyu, DDK-SDK, easy-medium” - the abbreviations just introduce an extra set label that can otherwise just be absorbed into the set rank range.

1 Like

Why not both?
In an option where players can choose which categorisation they want to use.

By the way this discussion seems to become endless. When will the first steps be taken to bring all this into practice?

1 Like

I’ve submitted a feature request to Github Add extra 'Difficulty' labels for Puzzles · Issue #1239 · online-go/online-go.com · GitHub. I don’t really understand which parts of the site are open source and which aren’t (not that I necessarily have the ability to implement it), but that should increase its visibility, and if it’s easy to implement maybe it’ll happen soon (or even sooner if it ends up being front end).

Feel free to read/comment the suggestion there, and I can edit what I’ve written if you’d rather keep the suggestion there clean and keep the discussion here.

2 Likes

Good idea. I will take a look later and comment here. Thanks for posting it to GitHub

Too many options is too confusing. Plus, they would serve the same purpose.

Agreed. Let’s just agree something and get on with it!

1 Like

@shinuito - I read the thing on GitHub. Looks great pal. I can’t see anything that needs changing. Thanks for doing that :slightly_smiling_face:

2 Likes

Funny coincidence. At the moment I am building a puzzle collection based on Utaro Hashimoto’s collection. One of the problems is indeed difficulty level of the puzzles :grin:

1 Like

Actually a list of (small) things that I find would really improve the experience of making go puzzles in the puzzles section are:

  • Adding in the option to pass as a colour (Black/White, player or non-player controlled).
  • Adding in more stone marking features like that of analysis mode. Specifically, normally in analysis mode you can shift+click to make a custom label, or you can use the keyboard to manually change to a specific letter/number to label the board. At the moment the only was I can see to place two ‘A’ labels on the board is to click and drag from one place you want it to the other which is inconvenient to say the least :slight_smile: Apparently shift click does work once the letter label is selects.
  • To not delete all of the variations when you change the setup slightly. If you make a bunch of variations on a puzzle and then decide that maybe you should add one more stone outside, it clears the whole variation tree which is awful…

I feel like some of these shouldn’t be that big of a change to make, but I guess I have no idea what the code behind it is like.

1 Like

Related:


Others:

It’s good to actually collect some of the suggestions together definitely to make them visible again. Maybe a few things get updated at once.

Puzzles do actually have an individual label to them (now) like life and death, fuseki, Joseki, tesuji etc although I don’t know where it’s used. It doesn’t seem to be visible anyway?

I see. Tags are visible in puzzle editor but not in puzzles themselves. Maybe they are working on filters.

1 Like

I looked for it, but can’t find where the labels are used for.
So I never bothered to select a category for the puzzles.
Also the paragraph about creating puzzles in Documentation & FAQ doesn’t give a clue.

1 Like

Just to add to that, I also think those labels that are currently given, aren’t really sufficiently broad enough, with the problem collections I’ve made in the past, I would have trouble categorizing them with those current tags.

Which is why I still prefer the below alternative

Tactics, Tesuji, Strategy, Theory, Opening, Joseki, Fuseki, Attack, Defense,

Or even subgenres

Opening (Joseki, Fuseki)
Tactics (Honte, Proper, Common, Shape)
Technique (Tesuji, Haengma, Tsumego)
Strategy (Attacking, Defending, Invasion, Reduction, Theory)
Endgame (Counting, Calculation)
Miscellaneous (Assorted, Untagged)

Those should be robust enough to categorize most problems, I mean, the current tags of “Elementary” and “Best Move” could refer to literally almost anything, so if you wanted to learn a specific idea with those it’s going to be nightmare to sort through until you get anything nearing what you want.

The other side to things is that, the vast majority of problems right now are just tsumego related, and people tend to go outside the box when they are presented with clear examples, and to put a name on things. Until it becomes easy to navigate and show that a problem can be more than just tsumego, we won’t be getting many middlegame or specific problems. There are vastly more ideas and concepts in go than just tsumego, we should really be thinking of these problems as a (superior) interactive alternative to books when it comes to learning the game.

3 Likes

With sufficient progress in Katago, this should be discussed again.

I think one day we will be able to practice the classified (and elo rated) tactical problem & tsumego sets as Lichess-style on the OGS.

The following is a reference site created by Katago’s producer, lightvector.
https://neuralnetgoproblems.com/

1 Like

I wonder if comparing to Lichess might be apples and oranges.

In chess, the very term “problem” implies uniqueness of the solution. IOW, a “problem” with multiple different ways to reach the goal (sometimes allowing for things like forced repetition) is not a problem. While already the existence of a “Best Move” category for go problems seems to imply there are other kinds (and I can confirm this from personal experience, and sometimes disappointment, being a sort of refugee from chess).

As a variation of things tsumego sites do (loosely on topic of the OP), goproblems https://www.goproblems.com added a puzzle rush feature where you get three lives and a limited amount of time to solve as many puzzles of increasing difficulty as you can.

And of course goproblems was the example given at the start of the thread for using Elo for puzzles.

2 Likes

One could relax the requirement of a single unique solution. And you can also let users rate puzzles as “good” or “bad”, which can help filter out weird incomprehensible AI stuff