Log in

No account? Create an account
20 February 2010 @ 03:10 pm
What If I Wrote the Puzzles? - Hungarian Edition  
This weekend was the Hungarian National Sudoku Championship (one of the many national championships I contributed to this year). My contributions to this championship were probably the most balanced between classic puzzles and variations.

I contributed three classics (presented in increasing difficulty):




I also contributed the 6x6 Warmup Relay I originally made for Sudoku Masterpieces (but which was cut due to space). Its got a pretty clear theme. After finishing a puzzle, copy the identically located digits into the shaded cells of the next grid (so the lower-left square of puzzle 1, for example, will become the lower-left given in the puzzle with a big 2 shaded in pink).

Finally I contributed two more puzzles which are variations I've focused on recently with my construction:

Arrow Sudoku (the numbers in the circles must equal the sum of the digits along the arrows; digits can repeat within an arrow but cannot disobey the standard sudoku rules in doing so)

Thermo-Sudoku (the numbers in the thermometers must strictly increase from the bulb to the end)

Please enjoy these puzzles, and congratulations to Zoltan Gyimesi for winning the Hungarian championship.
XSG - 1 G 2 Manyxsg on February 22nd, 2010 03:04 am (UTC)
What I love most about your puzzles happens to be the thing that I think you take the most pride in (at least, I seem to recall that you've discussed it in the past): you take great pains to make them aesthetic. I think that the above are a great example of taking something that has transcended an intellectual hobby and become an art. I imagine that some day variations of your puzzles will be presented in a museum...
motrismotris on February 22nd, 2010 03:29 am (UTC)
Indeed. For so many reasons - not just to communicate to the solver there is a human behind the grid - I think it is worth pursuing the art of a puzzle.

I presented some slides in my recent sudoku talk at Silicon Valley Puzzle Day that show the predominant form of the puzzles at the time the transitions US -> Japan -> Britain -> US occurred and referred to the steps as the "Evolution of Sudoku?". The hand-crafted artisan Japanese puzzle stands out, but even the rotationally symmetric forms of Wayne Gould's/Pappocom's software is a step better than the predominant US/Shortz form which even lacks this basic symmetry. The main "elimination of givens" to improve Number Place is still true, but a lot of the quality in Japanese nanpure did not become Su Doku or then Sudoku (aka wordless crosswords). I'll maybe adapt that material up here sometime, but I will not accept lazy automated design when so much more is possible.

Edited at 2010-02-22 03:33 am (UTC)
(Anonymous) on February 23rd, 2010 10:51 pm (UTC)
Quick question: as I understand classic sudoku today, there is no real evolution from the pappocom rotational symmetry form, and so I don't understand the transition in the evolution of sudoku from UK to US - especially as you've said that

"...the predominant US/Shortz form ... lacks this basic symmetry."

If anything then, the US is a regression rather than progression from pappocom?

I'd also argue that the quality of the solve is always more important than the initial aesthetics of a puzzle. Whilst it is true that a lot of "nice" looking puzzles also solve very nicely, it is not a general rule. Conversely, some of the best solves I can recall (red magician puzzles on nikoli.com) have no aesthetic properties beyond rotational symmetry.

So whilst I abhor lazy computer generated puzzles as much as any puzzle fan - for example it's really not much to ask that a sudoku has rotational symmetry - I think it ought to be made clear that the priority is always the solving experience. Especially pleasing aesthetics are merely a bonus!

motrismotris on February 23rd, 2010 10:57 pm (UTC)
Its not strictly aesthetics, but if I'm showing a point in a talk where people aren't solving the puzzles then its certainly what I use to point it out. I've already described for you what I think the role for human construction (and at least human editing) should be and you can see that at play in my recent run of classics posted here as I aspire to demonstrate my own goals in sudoku construction. I will say as a constructor who wants to be able to sell his classic sudoku in this country for (some) money, that the only market I can imagine is puzzles that look nice (and will also solve nice) but cannot have just the solve nice property, at least until the traits of proper sudoku are recognized in the marketplace here.

The ? after the Evolution of Sudoku in the title is there specifically because I view the pappocom software as at best a slight regression of the form and the shortz as a major regression of the form, to the point St. Martin's is functionally selling "Number Place" puzzles under a new name with better marketing.

Edited at 2010-02-23 11:02 pm (UTC)
(Anonymous) on February 24th, 2010 02:03 pm (UTC)
Ah OK, I see. I viewed US=birth of numberplace -> Japan nikoli refinement -> UK pappocom and worldwide boom as progressively important stages in the sudoku phenomenon. I don't think it's fair to label the US as entirely representing the regression into lazy computer generated puzzles - that's just market forces. A lot of people are happy with any old puzzle and so that's exactly what Will Shortz gives them at the lowest possible price.

I guess my point is that if the hypothetical sudoku museum existed, it'd be a travesty if conditions of entry were both look nice and solve nice as opposed to just solve nice. Admittedly, most would be look nice and solve nice, but I don't want to rule out the exceptions here!

Anyhow, it's fair enough that you want to sell books and I think that your most likely path to success is via "puzzle editing". I can only say that at least here in the UK, the nikoli publications are the ones featuring by far the prettiest puzzles. The problem they have is that there are relatively few such publications, and they tend to be more expensive.

motrismotris on February 24th, 2010 04:36 pm (UTC)
I should probably just post my analysis at some point but another thing I tried to characterize was the range of givens I've seen from the different sources. I found a Shortz sudoku with 42 givens! Now there is likely an audience for very easy puzzles, but this is certainly more than the standardization of the puzzle style should allow. The US crossword has specific requirements on rotational grid symmetry, # of black squares, no cheater cells, etc. Shortz would never accept a grid that grossly deviated from these established standards but then I guess they were set a few decades before his time editing the NYT puzzle. I think in these early years of the sudoku boom there is a similar lack of standardized expectations for what a valid sudoku should be but I'd argue its more than just the requirement it have a single solution. I expect at least one element of symmetry and no more than 33 givens. I'm curious at what the observed range on Pappocom sudoku givens is (I only used one book to estimate it for the purposes of my talk but mean to research it more).

Edited at 2010-02-24 04:42 pm (UTC)
(Anonymous) on February 24th, 2010 05:25 pm (UTC)
I have the pappocom program, and generated a couple of puzzles in the test.

As far as I could see, the "easy" puzzles didn't exceed 33 givens - indeed most were below 32. Although the special "very easy" option did. Actually, I'd not had a play with the program for a while - I was impressed with the aesthetics of some of the easy puzzles it turned out. I'd consider using the patterns of about 1 in 4 or so as ideas for my own puzzles. This ratio dwindled quite a bit for the "medium" and "hard" puzzles

Of course that's all anecdotal, but according to this source


Wayne Gould says "Su doku puzzles usually have between (say) 26 and 32 clues." Presumably he adhered to this guideline whilst writing his program...

In fairness to this supposed lack of standardized expectations, most (not all) published sudoku here in the UK adheres to <33 givens + symmetry. I guess this is good as a bare minimum, but doesn't exactly rule out bland puzzles.

motrismotris on February 24th, 2010 05:29 pm (UTC)
It certainly doesn't rule out bland puzzles, but its one of the basic design rules that changed between Number Place where 36 givens was common (the first 2 puzzles were at this number, for example) to later sudoku where say 18-24 are more typical with hard puzzles, and 24-32 for the rest. Having more than 36 givens seems a step back in the sense of puzzle design as you can certainly make very easy puzzles with just 32 if you just throw out "mistakes" from your generator and/or have your generator correctly value hidden versus naked singles and solving path breadth. I've always thought the Pappocom engine had a little more behind it but its easy in a small sample size to be tricked by your observation. Still I think there is more than just a market failure with the best selling US sudoku source and that their generator has never been changed suggests they don't feel anyone breathing down their neck with a better product.

Edited at 2010-02-24 05:30 pm (UTC)
Hooligan: beginner's luckjedusor on February 23rd, 2010 06:39 am (UTC)
projectyl has pointed me to your logic puzzles enough times that I figured I ought to add you as a friend. Your picture looks familiar--it's entirely possible that we've met at some puzzle event or another.
Craig K.canadianpuzzler on February 27th, 2010 04:44 am (UTC)
He was walking by our solving room on Sunday of the 2009 MIT Mystery Hunt and I shanghaied him into it briefly. So yes, you've met, if briefly.
(Anonymous) on February 23rd, 2010 08:12 pm (UTC)
Re: the warmup relay, count me among those who noticed what you were doing with the 5th, but assumed there must be some extra trick to it. I expected that doing the obvious thing there would result in a contradiction in the 6th, and so I'd have to work backwards from *those* shaded squares....

...but, no. I ended up just wasting time. Ah well.

motrismotris on February 23rd, 2010 09:26 pm (UTC)
On that one, there was simply no way to draw a 5 of any shape in the grid and also add a 5 clue to give a different solution so the unusual property you discovered could be exploited (but overall none of these minis were hard puzzles to be honest).
(Anonymous) on February 23rd, 2010 10:12 pm (UTC)
Are 6x6s ever hard, though? I can't recall ever seeing any.

motrismotris on February 23rd, 2010 10:20 pm (UTC)
A 6x6 puzzle should allow a large fraction if not all of the "difficult" techniques known to solve puzzles from X/Y-wings on up. Its just rare for someone to construct (and publish) a small puzzle with hard steps. However, I think they may be the right "training" space for some of the more advanced techniques since the affected cells will not be diluted by as many irrelevant cells to a particular kind of deduction.

This very old sudoku.com post has such an example and maybe links to others.
(Anonymous) on February 24th, 2010 09:54 am (UTC)
Very nice set, it's a very pleasant feeling when you're solving nicely designed puzzles :) Also, somehow I knew you'd put a 2=1+1 in the arrow puzzle in one of those two arrows