Sudoku

[pinkagape]

Hi

I want to get better at sudoku and speed up at classics. My current plans are:

Solve more classic sudoku each day
Watch Cracking the Cryptic

Does anybody have a more nuanced training schedule or a place to pick up on developing some more advanced techniques. I don't consciously use any named techniques at the moment but I've heard of things like X wings. I read about them and occasionally half-heartedly look for them but then forget what I'm doing and go back to grinding it out.

[pinkagape]

Is there a good source of quality sudokus or will any old sudoku get the job done?

[detuned]

I’ll reply in more detail later… but I’ll add for now that I don’t think cracking the cryptic is going to be all that helpful in speeding up on classics, beyond introducing “Snyder” notation. That’s a good first start but if you want to get quicker, you also need to understand its limitations in a speed solving context. The main one being that you are going to spend a lot of time not getting far with puzzles requiring hidden pairs/triples if that’s the only technique in your arsenal… it’s no coincidence that the harder puzzles requiring what I’ll euphemistically call competition-friendly techniques often rely on hidden pairs and triples.

Practice is key, but needs to be coupled with a good understanding of why you are slow on particular puzzles and whether this indicates a particular blind spot in your scanning.

[detuned]

I could probably go on and on with regards to techniques, but with regards to improving speed I think the salient points are these:

“Easy techniques”
- Having an idea of what “naked” vs “hidden” deductions mean. Naked uses cell-focussed logic, whereas hidden uses one of row/column/box focussed logic. So if the only candidate digit in a particular cell is a 1, then that’s a naked single. If the 1 can only go in one cell within a row, then that’s a hidden single.
- this generalised to pairs and beyond. However, whilst I’d argue that naked singles are harder than hidden singles, generally it’s the other way round for pairs. Hidden pairs aren’t so bad if you’re looking for the only places 1,2 can go within a box, but a lot harder within columns. Naked pairs are a bit easier once you’ve started marking candidates in promising looking cells (this is what I’d call McNeill notation rather than Snyder notation).
- the real secret is that on one level most sudoku techniques are the same thing. Hidden pairs are the same thing as naked pairs, and pointing pairs and box reduction and x-wings. All you are doing is permuting the interactions of the four different styles of logic - ie cell focussed, and then row/column/box focussed. Where you interact cell logic with eg row, that’s giving you pairs; where you interact row with column that’s a bit harder but gives you x-wing.
- related to this is the observation that a hidden pair in a row is the same thing as a naked septuple in the other 7 cells in the row (and vice versa)
- understanding how these feel different in practice when solving vs knowing they are all the same thing I think is really helpful in speeding up. Eg if you are doing lots of cell specific candidates (ie naked) and getting nowhere, then it probably points to needing to spot something hidden. Vice versa if you are cross-scanning individual digits across the grid (ie hidden) and seeing nothing, then chances are there’s a naked pair somewhere instead. Knowing when to change up how you are looking at the grid is a big big factor in speed.

“Harder techniques”
- you probably don’t need to know the concept of alternating inference chains, but if you are familiar with the setup then you can note on one level an equivalence between the sorts of harder techniques you might come across in competition: y-wings, skyscrapers, colouring etc. Basically they are all just quick-ish ways of observing contradictions arising from making placements of certain candidates, which then logically allow you to remove those candidates.
- it’s worth being aware of these and what they look like for sure, but they don’t come up all that often, and if you are light on this then I’d say it doesn’t really matter if you are good at guessing instead.

“Guessing”
- at some point it’s just quicker to guess. Or “Bifurcate” as the old euphemism goes. The idea here is to make bifurcation sound a bit more scientific than guesswork, but there’s no distinction
- the real secret is knowing where to start your guess. “Bifurcation” at least gives you a clue here - general start with a cell with only two candidates and see where that gets you.
- I’ll concede there’s some observable art to this: some people are reliably a lot quicker and “accurate” in finding the right place to guess, beyond what might statistically be expected. I think this is mostly down to having a good feel for the structure of the puzzle. I don’t think you get this unless you are really used to guessing (ie practice) but also this is where constructing classics really helps. If you know which clues are really going to constrain a puzzle down from many solutions to just 1, this can guide your intuition in making a guess in a puzzle which seems full of possibility as well.

[pinkagape]

Thank you so much for your detailed thoughts. I will bear that in mind as I give Sudoku renewed focus

[Calump]

Good thoughts from Tom, I think the point about hidden/naked logic and changing up the way you're looking at a puzzle is important when things aren't flowing smoothly.

With regards to sourcing classics, I feel that competition puzzles feel different to the computer generated ones you'll find.
The Logic Masters India website has an Expo section where people can publish puzzles. There's plenty of classic sudokus there, and you can compare times with some of the best solvers and watch replays as well if you want to have a look how they've gone through the solve.
Other than that, I think it's probably fine to go back to old championships or GP rounds even if you took part originally. I think it's hard to remember a solve path from just a classic sudoku