The New Sudoku Players' Forum

[saul]

Thank you for the warm welcome, good to be here!

Thank you for the countless hours of pleasure your game has given me.

[denis_berthier]

saul wrote: Does the name "kakuroatk" mean that you have something to do with producing the ATK kakuro puzzles?

yes, it's my game... here is a nice small hard puzzle for anyone to try...H23152 ... good luck!

Hi, kakuroatk. Congratulations for this game. Your Kakuro puzzles are my preferred ones. And I also like the presentation (the blue cells with thin white diagonal lines are very pleasant to the eye).

You may have seen we all have some questions about your classification of puzzles in 3 levels: E, M, H. Without disclosing any industrial secret, could you say a word about it ?

[denis_berthier]

here is a nice small hard puzzle for anyone to try...H23152 ... good luck!

yep, indeed good luck we need !

This puzzle is rated W14 by KakuRules.
I got more interested when I tried g-whips. I found lots of g-whips, leading to a resolution path very different from that without them.
But finally the gW rating is gW14 (with the same largest whip[14]).
I could only get shorter chains by using braids: the B rating is B10.

Does anyone have any special step allowing a shortcut leading to a simpler solution?

[kakuroatk]

You may have seen we all have some questions about your classification of puzzles in 3 levels: E, M, H. Without disclosing any industrial secret, could you say a word about it ?

yes, you will find (as you already have) that some may overlap...

there are harder mediums and easier hards... again it depends on your solving technique. if you use the surface area technique then it can greatly simplify some puzzles like H41001 for example... you may find it a little harder if you don't use that technique ... you may also find that mediums with a high concentration of white squares may seem a bit harder because it may discourage this and possibly other techniques ...

[saul]

Hi, everybody. Nobody has posted for a while. I don't have anything of great interest (except perhaps to me) to report. For the first time in a very long time I tried doing some hard puzzles. I had never succeeded in doing a hard puzzle before, except for one very easy one that Henri called to our attention. However, this evening I did three in a row: an 11-by-11, a 12-by-12, and a 13-by-13. No doubt, these weren't terribly difficult puzzles, but it's unlikely that I hit three really easy ones in a row, so I must be improving. This comes as a great surprise to me; I had decided that the hard puzzles were just too hard for me. I wish I had recorded the puzzle numbers.

[denis_berthier]

Hi saul,

Nobody has posted for a while.

Nothing new on my part (I've occasionally spent a little more time on Slitherlink).
Kakuroatk didn't say anything useful about his rating principles, so there's not much to discuss about them - but the overlap between levels is obviously much larger than suggested by his graphics.

I wish I had recorded the puzzle numbers.

If you try more (and you keep the numbers), I can compute their W rating. It could be interesting to see how it correlates to your impression of difficulty in manual solving.

[saul]

If you try more (and you keep the numbers), I can compute their W rating. It could be interesting to see how it correlates to your impression of difficulty in manual solving.

Thanks, I'd like that. I just solved H33931. I'd say it was about at the upper limit of my ability, without using trial and error. I used surface sums two or three times, and a lot of ad hoc arguments to eliminate candidates.

[denis_berthier]

I just solved H33931. I'd say it was about at the upper limit of my ability, without using trial and error. I used surface sums two or three times, and a lot of ad hoc arguments to eliminate candidates.

With or without surface sums, I found the W rating is W7 - which I think is really hard for a manual solver, significantly harder than the puzzles you used to solve at the start of this thread.

For surfaces sums, I used only the following partial results obtained by solving the ne and sw corners separately and then injecting these digits into the puzzle. So, there may still be more simplifications due to surface sums.

Code: Select all

. . . . . . . . . 6 9 8 5
. . . . . . . . . 1 3 7 2
. . . . . . . . . . . 6 1
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
7 1 . . . . . . . . . . .
. . 2 . . . . . . . . . .
. . 4 . . . . . . . . . .


[saul]

With or without surface sums, I found the W rating is W7 - which I think is really hard for a manual solver, significantly harder than the puzzles you used to solve at the start of this thread.

Thanks for taking the time to analyze this, Denis. It's encouraging to know that I've progressed, even though psychologically, I don't feel much like I have. I've just solved H61471, which was quite a struggle for me.


After I found that r12 c5-c7 was 987 and r13 c5-c7 was 746, I computed that r12c2+r12c3+r13c2+r13c3=10. After some ad hoc arguments based on the candidates I had for the cells in the SW, I was able to infer their values, and this propagated to the whole W side of the board. I had been staring at the board for a long time before this surface sum occurred to me.

Unfortunately, in trying to complete the puzzle, I made an incorrect inference, and ended up with a contradiction. I repeatedly tried again (I had been saving the puzzle as I went along) but I kept making the same mistake. Finally, I started again from the very beginning, and I realized that I had been incorrectly inferring that r9c12 could not be 5. Of course, I realized that it must be 5, and I was able to confirm this by "mental trial and error" and complete the puzzle.

This seems like quite a hard puzzle to me, though not as hard as I made it, of course. I would now be able to recreate the solution expeditiously, but only because I remember where the two "sticky bits" are, and how to attack them.

[saul]

Kakuroatk didn't say anything useful about his rating principles, so there's not much to discuss about them - but the overlap between levels is obviously much larger than suggested by his graphics.

I couldn't agree more. I just did H12361 and it's clearly a medium in my view, and not a terribly difficult medium at that. Kakuroatk must be rating the puzzles programmatically, since it wouldn't be practicable to so manually, and there there are some anomalies in the system, no doubt. It would be interesting to know some details of the rating system. It appears from his comment that his rating system doesn't use surface sums. Since there are several surface sums in H12361, that accounts for some of the discrepancy between the rating on the website and my subjective rating.

[denis_berthier]

Hi Saul,

I've just solved H61471, which was quite a struggle for me.

I already had H61462 in my database (it is obtained from H61471 by vertical symmetry + sector complementation).
In the W rating, it doesn't look so hard: W4

I just did H12361 and it's clearly a medium in my view, and not a terribly difficult medium at that.
[...]
It appears from his comment that his rating system doesn't use surface sums. Since there are several surface sums in H12361, that accounts for some of the discrepancy between the rating on the website and my subjective rating.

Without using surface sums, the W rating is 6 (but with only one bivalue-chain of size 6).
But there are so many surface sums that they undoubtedly simplify this puzzle (I didn't have time to try).
Moreover, the size is much smaller than in the previous puzzle - which always makes a difference in solving time.

[saul]

Hi Henri,

I just solved H81632, but only with extensive trial and error. Even after doing this, I don't think I could go back and solve the puzzle another way. Sometimes, after doing trial and error, I think, "Oh, if only I had seen this, I could have done it my head," but not this time. If you have the time and the inclination, I'd be interested to know how this rates on your scale.

Saul

[denis_berthier]

Hi Henri,

I just solved H81632, but only with extensive trial and error. Even after doing this, I don't think I could go back and solve the puzzle another way. Sometimes, after doing trial and error, I think, "Oh, if only I had seen this, I could have done it my head," but not this time. If you have the time and the inclination, I'd be interested to know how this rates on your scale.

Hi Saul,
Who is Henri ?

I had already solved H81642, an isomorph of H81632 (via a 90° rotation). No wonder you find it hard: it is in W10 - i.e. very hard.
There are naked triplets r10{c7 c10 c11}{n1 n2 n3}, hidden pairs c8{n1 n2}{r5 r6}, naked quads r5{c6 c11 c8 c9}{n4 n3 n2 n1}, again hidden pairs c5{n1 n2}{r4 r6}, but it still requires whips[10].
(I use no surface sum, as there's no obvious one.)

[saul]

Who is Henri ?

I beg your pardon. I don't where that came from. I don't even know anyone named Henri, so far as I can recall.

[denis_berthier]

Who is Henri ?

I beg your pardon. I don't where that came from. I don't even know anyone named Henri, so far as I can recall.

There's no harm. We (in France) even had kings named Henri. (But that's no longer a good job here - too risky !)