I like that suggestion
1 Like
No problem.
Now I got a clear picture of what you are aiming at.
I second this, though I think the categories should be based on TPK, DDK, SDK, and dan. If we want more specificity, use “low”, “mid”, and “high” as desired. So maybe, TPK, low DDK, high DDK, low SDK, high SDK, low dan, mid dan, high dan. Maybe add a “pro” category for stuff on level of the igo hatsuryuon.
What does this mean?
It’s a classic set of extremely difficult problems. Any decent pro should be able to solve them but they will be beyond most amateurs.
No. This misses the point. Firstly, too many categories makes it too hard to decide where a puzzle belongs. Secondly, judging the difficulty as a rank is prone to inaccuracy, whereas generic categories to indicate the relative difficulty of a puzzle or set is much easier to judge. Thirdly, the suggested categories of elementary, easy, medium, hard and very hard are already used in many problem sets (e.g. Cho Chikun encyclopedia of life and death) so assigning the difficulty becomes trivial as it is already decided. And finally, players (especially beginners) might not understand what DDK, SDK, etc mean (indeed, I am SDK and I have never come across TPK before - what does that mean?), but anyone can easily understand the labels elementary, easy, medium, hard and very hard.
1 Like
Single digit kyu is 1-9 kyu, double digit refers to 10-19 kyu and TDK probably to 20-30 kyu.
But if you take TDK literally it would mean 100-999 kyu.
I haven’t met such TDK’s either 
Is TDK perhaps a nonsensical term?
If it means triple digit kyu, then yes, it is nonsense. Even kids start at 50k
1 Like
Maybe, but I said TPK, not TDK. Twenty Plus Kyu.
And I’m not saying we should have all those subdivisions, that’s just an example of how we could build on the rank band distinction if we wanted more subdivisions. The core of my proposal is TPK, DDK, SDK, and dan. This is fewer categories than were proposed by others, so it is disingenuous to accuse me of having too many categories just because my suggestion is easily extensible if, and I repeat if, desired.
It also bears noting that I support the suggestion to not replace the kyu/dan rankings, but add labels with the coarser categories.
1 Like
I have four categories with room to naturally draw finer distinctions if it proves desirable.
Hence my four (within the subitizing range for what it’s worth) generic categories.
But those terms are more subjective than the commonly used rank bands. If we stuck only to tradition, OGS would mail you a certificate for completing a 1d puzzle collection in return for a generous donation, and if you need sure hints along the way, well, you did pay, so we’ll send you the certificate.
A beginner will not understand what life and death means, what seki means, what ko means, that unconditional life is preferable to seki and seki to ko and ko to death, which are all necessary for doing tsumego. These, like TPK (mostly used on OGS, but that’s where we are), DDK, SDK, and dan, will be taught by osmosis and/or individuals. The advantage of using rank bands like this is that they are rooted with a more solid meaning beyond the puzzle creator’s subjective notions of difficulty.
1 Like
My stance generally has been I prefer labels that don’t mention ranks, unless those ranks actually somehow connect to real ranks, imagining OGS puzzle ranks were indicative of the Rated game ranks.
Anyway I think the temporary proposal which I agree with is:
- don’t remove any existing labels 30kyu to 9dan that exist
- add new labels so puzzle creators can use them instead, like easy, medium difficult etc as mentioned above. One could also add in the labels sdk ddk Dan etc.
Everyone is slightly happier in the short term and then people can look forward to another solution in the future
1 Like
Correct.
Often TDK is used.
TPK makes sense 
2 Likes
Apologies, no accusations intended. Just trying to explain why easy, medium, hard is preferable over DDK, SDK, etc.
Thank you. Seems like we are all pretty much in agreement on this.
Ah, fair enough. Thanks.
Regardless of the number of levels, elementary, easy, medium and hard are still preferable because:
-
Difficulty categories separate from the notion of rank are easier to judge than DDK, SDK, etc that are inherently tied to rank
-
With respect to:
Subjective difficulty descriptors are good because author assigned difficulty, whether rank, rank band, or category, is inherently subjective. It is, therefore, contradictory to describe a subjective decision on difficulty with an objective measure like rank or rank band. Easy, medium, hard, etc are more consistent in this respect.
- And with respect to:
I mean a beginner who at least has a basic understanding of this idea. Before that, a player is still learning the rules really.
Again, I mean a beginner who at least has a basic understanding of seki and ko. These are part of the rules so are learned about (to a basic level) very early, probably long before terms like DDK, SDK, etc. Plus, many elementary tsumego only require knowledge of life and death (live or die only, no possibility of seki or ko).
Hence my previous point that the plain English labels of easy, medium, hard, etc are more easily understood than the abbreviations DDK, SDK, etc.
In conclusion:
Yes, this is exactly the point. In an ideal world, we would have an automated system that determines a puzzle’s rank relative to players’ ranks. But we don’t have that. Difficulty is assigned subjectively by the author. Therefore, until we have a robust automatic rank-based system, we should just use easily understood, commonly used and easy to decide subjective difficulty levels, the most obvious choice for which is easy, medium, hard, etc.
1 Like
Let the set of options {easy, medium, hard, very hard} be a and the universe in which this is the case be A.
Let the set of options {TPK, DDK, SDK, dan} be b and the universe in which this the case be B.
Consider all members of a and b to be themselves sets containing all tsumego placed within them with respect to difficulty.
Let a tsumego t taken from set T of all tsumego exist.
Let an observer O(A) place t into some x in a. O(A) has made a subjective decision, even if subconsciously, as to a rule which defines an x taken from a into which any given member of T belongs. They then make the further subjective decision of how to evaluate the placement of t, given that it may be reasonably asserted that the rule for placing all members of T into the correct x taken from a is computationally intractable.
Let an observer O(B) place t into some x in b. O(B) does not make a subjective decision as to the rule which defines into which x taken from b any given member of T belongs, because that rule is clear and well understood, if necessarily vague. They then make only one subjective decision: to evaluate the placement of t, given that, as before, it may be reasonably asserted that the rule for placing all members of T into the correct x taken from b is computationally intractable.
So my suggestion is less subjective.
No, they’re harder to understand because there is no common framework from which to render them mutually intelligible. Terms like DDK and dan are rooted in something solid, and this allows meaningful communication.
I agree, but subjective is good in the current system. Subjective wording of the difficulty category (easy, medium, hard) implies that the decision of the difficulty is subjective, which it is. The result is that the difficulty is taken as a guide, which is all it can be without an automated system. However, objective wording (DDK, SDK, etc) implies the decision of the difficulty is objective, which it is not. The result is that the difficulty may be taken as absolute, which it cannot be without an automated system.
Clearly, this is a matter of opinion. Maybe the pragmatic solution is to include both. Puzzle authors can then choose categories (easy, medium, hard), rank bands (DDK, SDK, etc), or individual ranks - whatever the author thinks will be most useful for their puzzle solvers. The difficulty drop-down menu would then look like this:
Elementary
Easy
Medium
Hard
Very hard
30-20 kyu
20-10 kyu
9-1 kyu
1-9 dan
30 kyu
29 kyu
28 kyu
(etc)
7 dan
8 dan
9 dan
Seems OK to me 
1 Like
101weiqi is the best Tsumego platform I know if by a fair margin. It’s all in Chinese, but someone posted an intro page on imgur: Short Introduction to 101Weiqi - Album on Imgur
1 Like
I think you’re underestimating how much variance there is over a 10k range. I’m currently a 5k on OGS; for puzzles on 101weiqi, 8k borders on too easy and 3k borders on too hard. I’m also right in the middle of a range band. For me, I’d only run into a few puzzles outside of my “range” for difficulty over the proposed range bands. For someone that’s a 10k, most puzzles in the 20k-10k range will border on trivial and most puzzles in the 1k-9k range will be too hard for effective training. Being able to steadily increase difficulty is essential for building pattern recognition skills, and 10k range bands won’t really allow people to do that.
3 Likes
It seems to me that there’s an assumption that it’s desirable to work on problems that are the right difficulty for one’s level. As has been well hashed over through this thread, there are numerous difficulties with that assumption, especially with attempts at defining the difficulty of a problem or problem set. But besides all that, I think part of the mistake with the assumption is that a player wouldn’t want to work on problems that are too easy. Actually, I think it’s desirable to train by working through lots and lots, repeatedly, quickly, and accurately, with occasional attempts at challenging ones (one that takes more than 2 minutes to solve, or cannot be solved correctly).
I’m SDK, and I’ve solved about 4000 tsumego this year. I think it’s really helped me a lot (I don’t have any hard evidence of rank increase, but I was able to beat a friend 9-1 in a jubango; we started at 6 handicap, and I beat him down to 9 handicap, at which level we won one game each. I could feel my reading was faster and stronger, as if playing a sport after a lot of strength training. Felt good!) I was using mostly Badukpop earlier in the year, and have more recently been using Tsumego Pro on mobile and Tsumego-Hero.com on my laptop. I still do a lot of “easy” and “elementary” problems, even though I’ve been playing 10 years, because I want to increase my reading speed and do a wide variety of problem types.
So, this is all to say that I think it’s not so important to try to work on the “right” “difficulty” of problems. If you’re setting out to solve hundreds and hundreds to improve, what’s needed is a variety of high quality problems so that you’re building your reading speed and accuracy.
Badukpop (free user only):
Tsumego Pro (paid for collections):
Tsumego Hero:
EDIT: Added my progress on tsumego-hero.com.
Also of note: when working through hundreds of hundreds of problems it’s okay and good to solve the same problem repeatedly over time. Badukpop’s puzzle mode does this automatically by giving random problems within a difficulty level. When considering that the same or similar problems occur on different apps and sites, I’ve probably solved the same ones dozens of times. This is good. It builds reinforcement and speed. I’m actually planning on wiping my progress meters from Tsumego Pro soon, and maybe reworking through collections on tsumego-hero again.
As you can see, I skipped the easiest collections in Badukpop and tsumego-hero.com, because they were so easy as to be boring. I think that’s a good enough metric for choosing a problem set to work on. As long as a problem set is not too easy as to be boring, and not too difficult as to be taxing and unsolvable, then I think you’re within the right range for you. This should result in a very wide range of appropriately challenging problems, and each user can easily make their own subjective assessment of that. Aiding that assessment should be the goal for classifying the difficulty of the problems.
4 Likes
I agree with this completely. I read an article by a professional Go player who recommended, for the best improvement of Go skill, that 90% of your Go time should be spent doing tsumego (the other 10% playing games presumably) and that the tsumego time should be split evenly between problems that are easy, medium and hard. I mention this as it is some professional support for the idea that a broad range of difficulty is probably a good thing.
But all that is more a discussion of how puzzle solvers should approach tsumego, which is getting a little off topic. The main thing here is to decide on how the difficulty drop-down menu in the tsumego creator should look. How about this:
Elementary
Easy
Medium
Hard
Very hard
30-20 kyu
20-10 kyu
9-1 kyu
1-9 dan
30 kyu
29 kyu
28 kyu
(etc)
7 dan
8 dan
9 dan
I think this is the solution that would please most people as the options are there for individual ranks, wide rank bands or general descriptors (easy, medium, hard) as people prefer.
Can we at least agree on that?
1 Like
I agree with this, but this is difficulty relative to the solver, which cannot be known by the puzzle creator. “easy” for a dan is different than “easy” for a sdk.
These should be replaced with TPK, DDK, SDK, and dan. If that is done, then your suggestion is a fine compromise.
Still another option is to use levels, like “level 1”, “level 2”… “level 5”, or a star rating, as some other sites and apps do. This avoids the difficulties and misconceptions that may come from associating them with a rank, as well as the relative challenge of what a “medium” puzzle is like.
An Elo rating for the puzzle that is not associated with rank could work, too. Then a user could say, “give me anything between 2000 and 4000 difficulty.”