The New Sudoku Players' Forum

[JeJ]

I know there is not a "right" way to solve a Sudoku, some people would look for clues here and others there and all options are valid. I am not a very advanced player but I am very good at logic and mathematics stuff, so I've progressed quickly since I started. I've solved some Hodoku "hard", Simple Sudoku "Extreme" and enjoysudoku.com "Devious" puzzles but sometimes I just can't go past the eliminations and placements of the early stages of the game. I am going to present my approach to solve puzzles in hopes that people can give me advice on other tactics, I don't mean solving techniques, I've studied most of them, but when it comes to apply them, the amount of work seems overwhelming and I get stuck.

So here's my approach:

I use Hodoku (even if the puzzle is from other sources) to filter the puzzles from 1 to 9 and in the process I make the obvious eliminations and placements, basically looking for single candidates, single digits and the obvious locked candidates.

Since the elimination or placement of higher numbers can change things for lower numbers, once I reach 9, I repeat the process again from 1 to 9.

Then I use the filters again from 1 to 9 to look for pairs and try to find naked ones.

Then I use the filters again from 1 to 9 and look for units in which a specific number is only twice, since they lead to pairs (naked and hidden but mostly naked), 2-string kites, skyscrapers and x-wings and when I find them I do the corresponding eliminations.

Then I use the filters YET again from 1 to 9 and try to find again pairs and look for XY-wings and W-wings throwing in a naked triple here or there.

After this I am almost finished with my bag of tricks, then I look for Unique rectangles and that's about it.

So could people please criticize my method or recommend some adjustments to it?

Thanks.

[JeJ]

I thought this thread was going to generate some interesting discussions, but apparently not.

[David P Bird]

JeJ, perhaps I've missed them because of your terminology, but I see no reference to two very common solving techniques: Simple Colouring and Box-Line Reductions.

[JeJ]

JeJ, perhaps I've missed them because of your terminology, but I see no reference to two very common solving techniques: Simple Colouring and Box-Line Reductions.

I thought I was using the standard terminology, but it's ok.

Box Line reductions I use it when I said "looking for single candidates, single digits and the obvious locked candidates" well maybe my terminology was not that standard.

I was using colors before, but I stopped doing it when I switched from Simple Sudoku to Hodoku, but I think I'll start using it again.

Thanks for your advice.

[kifotv]

Sorry if this reply is a little late, but I just saw this thread (searching 'hodoku', as it were), and thought this sounded kinda like what I'd been thinking about asking here for a while, I think that, kinda like OP said, techniques are easy-ish to learn, and to understand, and possibly even to spot, but it can be hard to learn when to look for what, where to look, and especially learning to feel when easier techniques probably won't advance a puzzle, and that harder techniques should be used. I'm not really an expert, not even close, but I've thought about this a lot so might as well chime in (I'd be interested in hearing other peoples thoughts, too!) I use HoDoKu too, so I guess I'll start there. I might type too much, but I think about what I would say all the time so I'll get it out

I'm wondering why you'd have to make the candidate lists yourself, since it does that for you. I guess sometimes it saves you from having to spot subsets later on, but the extra work and time seems like a waste since they're easy enough to spot later. Also, if time and patience (and effort) are a breaking point, I'd suggest using Set All Singles (F11) before even starting to solve it. Using auto-candidate lists and auto-set-singles can save you from having to work through the more trivial things (especially the tedious singles-cascade that most puzzles boil down to).

Another great thing to try is using hints, especially Vague (alt+f12) and Concrete (ctrl+f12) hints. I think finishing every puzzle is a good idea, even if you need to use hints, or 'show next step'. At the very least, it will give you the answer to "What step should I use next?".

There are a couple more things which are HoDoKu specific, but might take some tweaking and time. The first is to re-order and re-score your solver steps (Edit>Preferences>Solver). HoDoKu tries techniques in that order, so try to arrange them in an order that makes sense to you. For example, I think that naked subsets are ALWAYS easier to find than hidden subsets, so I set the solver to look for Naked Quads BEFORE Hidden Pairs. You probably shouldn't put any techniques you don't know above those which you do, obviously, and you may even consider disabling techniques which you don't know, assuring that any puzzle you generate can be solved by techniques which you can actually use.

The second would be tweaking the scores in the solver. I personally think XYZ-wings are easier to find than XY-wings, but by default it's the other way around, so I tweak the scores to reflect my own tastes. If you edit it from time to time, and you order the list according to your own tastes, the total scores of puzzles (shown in the bottom of the main window) will more accurately reflect how hard the puzzle is for YOU. If you get the scores and order set how you want, try changing the level ranges (Edit>Preferences>General) to match your 'moods', to make sure you don't get a puzzle that requires 30 minutes of intense focus when you just want a casual 5-minute solve.

Oh, and use the learning and practicing modes, and not just the 'playing' mode! (check the manual)

As far as general non-hodoku-specific things, try to learn some more techniques, and read up on the ones you already know. Seems like most techniques have a more abstract base concept, outside of just the 'pattern', and understanding the logic usually helps me spot it easier than just knowing and scanning for the 'pattern'.

At some point, you may want to try searching for eliminations, rather than places where techniques can produce them. Learning chains and loops can help a lot with this, but isn't really required. If you see a 2 candidate cell, or some strongly-linked cells, or any potentially effective elimination, try looking around for a technique to make it happen. You might also be spotting an area which almost fits a technique, if it weren't for an extra candidate or two, and searching to produce those eliminations. Finding a goal, and then the path to reach it, might prove a little easier than searching for paths which produce results, but I guess it's a taste thing, too...

OP: What sort of techniques are you using regularly? Do you have a favorite technique that you look for first?