Page 1 of 1

Classic sudoku speed

Posted: Tue 08 Jun, 2021 4:02 am
by pinkagape
I want to echo Paul’s request for ideas on getting faster at classic sudoku. Similarly, when I have extra constraints, I get faster. I think I must be missing some key classic techniques.

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 8:06 am
by Mblount
I don't have much experience with the sudoku gp (or sudoku in general), but for what it's worth I notice "pairs" come up in the classic puzzles relatively often and have started keeping more of an eye out for them.

For me, the "variant vs classic" comes down to two things - the first is that spotting the break-in for variants (eg a 2-cell sum of 17) is likely to be easier for me than trying to scan rows/columns boxes of a classic sudoku. Gmpuzzles have some solving videos of classic sudoku, and there's sometimes some mention of how the "structure" of the grid informs where to look first. I've not tried doing this yet (and suspect the advantage would only come with experience), but I can see that developing an intuition about how the solving path is likely to go will prove faster than just scanning everything

The other aspect (which I think applies more to puzzle GP) is that my score is effectively a comparison against the test solvers' speed. I tend to be slow at all puzzles, whether classic or variant, and so if there's a variant that seems new, I'll consider attempting it in the hope that test solvers struggled with it more than usual. I guess this ties in with the strategy of going for high-pointed puzzles, but I don't trust myself to solve those inside 90 mins yet!

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 9:30 am
by david mcneill
I think one largely unspoken factor in classic Sudoku speed-solving is knowing when to start guessing and in having a good system of keeping track of tentative solving chains. Guessing isn't usually as beneficial in variants but is still an important last resort. I know on CtC there is ongoing banter about Mark's predilection for guessing, but it makes perfect sense if your primary aim is to solve puzzles more quickly.

Once anything more complicated than an X-wing comes along, I'm always going to start guessing. I never start out intending to guess, but my best intentions tend to go out the window when my head starts to hurt.

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 10:30 am
by eratosthenes
I saw a comment on a previous sudoku set along the lines of "I see that the unwritten rule of no X-wings without signposting them has been done away with"

How *would* you signpost such a thing?

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 12:05 pm
by Feadoor
By putting them in rows/columns that are largely filled, typically. I think I did an okay job of it on puzzle 2 of round 2, last year.

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 1:16 pm
by Feadoor
There's also this puzzle of mine where I think the X-Wing is quite clear, if you'll forgive the superlative introduction to the video:

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 1:36 pm
by Feadoor
And to avoid derailing the thread, here are my thoughts on speed in classic sudoku:

There are basically two ways to improve, IMO.

Solve lots (thousands) of easy puzzles, which will do two things. Firstly, it will improve your raw speed at the essential skills like being able to see which numbers are missing from a particular area, and being able to spot singles quickly. Secondly, it will develop your sixth sense for where the restricted parts of the grid are likely to be. It's also worth practising using minimal notation; after all, if you can solve without additional markings then you simply spend less time writing, and if the grid gets especially cluttered it can be a hindrance to scanning.

In parallel to this, develop a good heuristic for when you should abandon logic and simply make a guess, and a good notation for making such guesses and unwinding failed ones. Note that the point at which you should abandon logic in an environment where speed matters is well below the threshold at which you would cease to be able to solve the puzzle logically in a casual environment. Rather, if you find that easy progress has dried up and you don't have a sense of where to look next, a guess is often quicker. Not only do you immediately make progress, but if it fails you will probably have learned something about why it did so.

The top solvers are as fast as they are mostly because they're very good at logic, rather than because they're good at guessing. I don't mean that they know very advanced techniques, just that they're exceptionally good at the basic ones. Their raw speed is phenomenal and their intuition for where they should look is also very well-refined. But being able to guess efficiently is still a necessary skill for speed.

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 6:20 pm
by detuned
Some assorted thoughts from me:

Most of your placements in the grid are going to be either hidden singles (ie where there’s only one cell left for a number you are looking at) or possibly naked singles (ie where there’s only one possible number left for the cell you are looking at).

Hidden is easier/quicker than naked; but it’s good to make sure you’re in the habit of getting everything that’s available in the grid. A good way of looking at that is considering groups of 3 rows or 3 columns that overlap with the 3x3 box structure of the grid. Some people call these chutes. Anyhow where you have 2 of a particular number placed in such a group, then the 3rd will be restricted to a max of 3 cells so that’s always a handy place to look. With practice you can also get comfortable with row/column interactions too.

Naked is still quick with practice - the trick is finding weak areas of the grid to look at. This means nearly full boxes, rows or columns. Sometimes they can be sneakier, eg 3 candidates ruled out by the row, 3 by the column and 2 in the box, but again this is something that gets easier with practice.

Being quick when harder techniques comes down to two things. Firstly being confident that you haven’t missed something easy - sounds obvious in theory but in practice this separates solvers at the very top - and second understanding structures which might induce more involved techniques. Eg you might do pointing pairs or box reduction, but then you want to follow it up and see if it’s giving you a naked single. Or you might see structures that look like x-wings or swordfish are likely in play. Or similar for colouring techniques, and even things like y-wings (especially if you have a bunch of pairs notated). The key with all of that is practice and feeling comfortable with it.

One tip is if you struggled with a puzzle on a particular step - this is a good learning opportunity. Try resolving the puzzle whilst it whilst it’s still fresh, so that you start registering the places in the grid you needed to look at to get past a sticking point.

I tend to prefer minimising notation. I have known people who do tons of notation on computer and beat the pants off of me - although interestingly are slower on paper. I tend to think the lesson is experiment a bit and see what is quickest/most comfortable for you.

Re: Classic sudoku speed

Posted: Tue 08 Jun, 2021 6:23 pm
by detuned
Ah only other thing is what I call solving rhythm. Using that last number you placed to naturally move onto the next. The more you can build up that natural rhythm through the puzzle, the quicker you’ll be. For an easy level puzzle and the top solvers, that rhythm basically means you don’t stop writing from the 1st number placed to the last. They'll also tend to fill in whole clusters of cells within boxes at once, rather than just individual cells. Once you start doing that, you can get to sub-60 second solves on paper, and I think that's starting to get towards world class (although you still probably have another 10-15 seconds to go after that!)