# Tsumego: rating difficulty level

Recently I have been adding a lot of puzzles (tsumego) to OGS Puzzles.
One problem I encountered is that It is hard to determine the difficulty level of a go puzzle.
The “system” I use at the moment is that I take my strength (approximately 10 kyu) as a standard:

• no trouble solving a puzzle: 15 kyu
• no clue at all: 5 kyu
• anything in between the two: 10 kyu

Apart from it being not a very precise rating is that this is of course a rather subjective “system”. My rank is certainly not constant

So let’s aim fo a better system.
What are the factors that determine the difficulty of a go puzzle?

• length of sequence: how many moves are required to solve the problem?
• special moves: are there tesuji’s (snapback, throw in) involved in order to solve the problem?
• time: how much time is needed to solve a go puzzle?
• complexity: how many stones are required in setting up the go puzzle?

Maybe there are more factors that determine the difficulty level of a go puzzle.
Any suggestions?

3 Likes

Honestly i would be amazed if you reach to define a level of difficulty according to a grid of criteria like these.

My answer would be rather a bit elitist or simplistic but I think to give a “ranking” at the difficulty of a problem, you need the qualifications of a strong player (and perhaps good teacher too).

Complexity: the quantity of needed stones to lie down a problem is not correlated to the difficulty. Some problems with very few stones can be very difficult.

Length: if you consider a shicho, it can be like 20 moves to read… in a easy way. A under the stones tactic can be just 5 moves maybe but really hard to see.

Tesuji is a bit hard more when you study them at first but less later, even the knowledge of them will help you. More difficult are unusual working moves, like an empty triangle, a contact to attack…

Time: good criteria but need some community work test and keep memory after. Because it’s advised to not spend to much time on a problem, there will be some wasting time when trying to evaluate difficulties by the time passed to solve it.

6 Likes

1-5: easy
6-10: intermediate
11-25: hard

Straightforward solution: simple
Special moves: complicated

At a glance: easy
Thinking a while: intermediate
Thinking long: difficult

1-5: simple
6-10: intermediate
11-25: difficult

No: easy
Yes: difficult.

As @Groin has already pointed out, the reading depth is not necessarily what makes a problem difficult. I would add to that the number of possible alternatives one has to read out every move, i.e. the reading breadth. One could combine these two and take the size of the entire tree of variations as a metric. I believe this would also be a good measure that is proportional to how much time a computer program needs to solve a problem with a “brute force” approach.

4 Likes

Yes the wideness is key. On this side, we still should care about the branches you can easily cut (but the computer will not), when there is no trick on this aspect in the problem.

2 Likes

So a shicho through the board will always be a difficult problem?

where do you base this on?

I have 5 factors and I think that the combination of these determines the difficulty level of a go problem. Focussing on only one of the factors will cripple the system.

By the way: I do realise that this mind experiment may lead to nothing, but wouldn’t it be great if it did lead to an instrument to rate go problems level with?

It is easy to criticise but it is harder to construct something out of the blue
I wouldn’t mind a more constructive approach.

It’s not easy to criticize if you are convinced. In my opinion most of these criteria are not so pertinent and I did try to tell you why, with all the respect I can give.

I didn’t go very far because the difficulty of a problem is related to what a player has already assimilated. With a good training, some 8k can be very proficient with under the stones class of problems, while some other 2d may still have great difficulty.

So it could be more interesting to have a set of different kind of problems evaluated as a whole as a single one to mitigate these differences.

About “constructive” isn’t the wideness a new idea to include in your 5 factors, but which is still not considered as a basic criteria? Or did you eliminate already the “complexity” (quantity of stones to display)?

In my first answer I introduced the mang shou, the blind moves in Chinese, moves difficult to find because they are unusual, not in the common way of playing.
Instead you still focuse on your 5 own criteria. Ok just don’t tell me I’m not constructive.

1-5: * No: * Fast: * 1-5: * No" *
6-10: ** Yes: ** Normal: ** 6-10: ** Yes: **
11-25: *** Slow: *** 11-25: ***

5-8 stars: definitely weaker than your rank
9-12 stars:: equal strength
13-17 stars: definitely stronger

No, still trying to build a system.

Have a look at these. I think this is this is the concept @Groin mentioned.

1 Like

If we could just get this github request pushed through…

It kind of sounds like it should just be a case of adding a couple of new strings in in some place, and then editing the display of the puzzle band for the overall puzzles. I wouldn’t know where to look to see if it was front end or backend though. From here Make Tsumego Great Again!

1 Like

I carefully reread your first answer, but didn’t find any reference to either mang shou or blind moves.
Not a problem, I like the concept and are going to try to incorporate it.

Ok I didn’t name it, it’s part of the comment on tesujis

When I was making puzzles, I tried to think about a few things:

1. how many moves deep do you have to read?
2. how many reasonable first moves are there?
3. are there any uncommon moves in the solution?

But I had no idea how to translate those questions and answers into a rank… I just made some shit up and slowly people complained about it being “too hard” or “too easy” so I just adjusted it over time.

3 Likes

Tried to incorporate the feedback and also rephrase some aspects as to get the picture more clear.

What are the factors that determine the difficulty of a go puzzle?

• length of sequence: how many moves are required to solve the problem?
• number of reasonable first moves: how many reasonable first moves are there?
• special moves: are there tesuji’s (snapback, throw in, etc.) or uncommon moves (mang shou, blind moves) involved in order to solve the problem?
• deep reading: how many moves deep do you have to read?
• reading breadth (wideness): how many possible alternatives do you have to read out in order to solve the problem?
• time: how much time is needed to solve a go puzzle?
• scale of problem: how many stones are required in setting up the go puzzle?
• complexity: is the situation easy to grasp or does it take some time to understand the situation?

Disclaimer: I do realise that this is not exact science. Most of the factors have a subjective feeling. The only thing I am trying to achieve is getting a clear insight in the rating process. By describing it it also becomes controllable for others.
I aim at a guideline which can help people who do go problems, certainly not a hard set of rules.

2 Likes

15k: half of 15k players will solve this problem on their first try.

10k: half of 10k players will solve this problem on their first try.

5k: half of 5k players will solve this problem on their first try.

This should be the only aspect of the rating process. Everything else is subjective.

1 Like

We need an update to the puzzles part though for that.

For the moment we’re stuck with the user uploading the puzzles having to make a call on their difficulty

1 Like