[chongli]

Obviously, any such classification would be subjective, but there’s probably a “reasonable” limit to solvability. (Without guess and check)

It’s totally subjective. Since sudoku is a closed (fixed size grid, fixed number of symbols), guess-and-check is just a catch-all label for logical inference strategies which haven’t been named yet. New strategies are discovered (and named) all the time and one of the ways to do this is to guess and then, if the guess was correct, go back and try to understand why.

This bias against guess and check seems to be some deep-seated issue from our culture. I know a lot of mathematics teachers frown on guess-and-check as well and that can rub off on their students, possibly implanting the bias for life. Unfortunately, having this bias can really damage a person’s ability to learn and succeed at mathematics in university. It turns out that guess-and-check makes a triumphant return as a strategy for quickly completing proofs when one is unfamiliar with established theory. Sound familiar? That’s just like sudoku!

Working mathematicians, in contrast to poor math teachers at lower levels, have a healthy relationship to the guess-and-check strategy. And that’s good news for them, since they are often working in areas where there is insufficient established theory to make any progress.

[rhdunn]

It's all down to whether or not there is any logic to the guess and check strategy. Most of the computer algorithms work by picking a number in the possible values, following the eliminations, and repeating. If they hit an error, they backtrack to the last decision, choose a new number from the set and repeat.

If you look at some of the Cracking the Cryptic puzzles on YouTube with "computer" in the title, that covers cases where the computer resorts to a guess and check strategy but where better underlying logic is available in the solve path.

[chongli]

The finite nature of sudoku puzzles guarantees that there is an underlying logic to a given move, whether the solver is aware of it or not. Ultimately, all sudoku strategies are backed by an overriding master logic:

"Does this branch of the game tree contain the completed solution or not?"

Whether or not humans have come up with a name for a given strategy depends, almost entirely, on the depth of that branch.

[olliej]

No that’s not the case. Having a single solution does not mean that there is a logical path to it.

If I have a puzzle where I’m down to a bunch of different options and the only way to make progress is to just try fixing different numbers until the puzzle resolved that not a logical progression , it’s an exhaustive search.

[53x15]

If the solution is unique then there is a logical path to it, though maybe not one feasible to discover for an agent with bounded time and (especially for humans) bounded working memory.

Consider the search tree of a backtracking algorithm operating on a SAT-based representation of a puzzle. A leaf of the tree corresponds to some clause that becomes empty after eliminating literals assigned on the path to that node. If you flip the tree upside down and start from the clauses at the leaves, then two leaves with a common parent represent two clauses with complementary literals that can be resolved upon. So any branch of the algorithm's search tree can be viewed as a tree-structured resolution-refutation proof eliminating some value given a set of prior assignments on the path to the branch, and the tree as a whole can be decomposed into an ordered set of proofs for each elimination in the path to an overall solution.

Or, if you prefer, start again from the SAT-based representation and run a prime implicates algorithm like Tison's. Such an algorithm never advances a "guess" or generates a clause that is not entailed. It simply finds one logical consequence after another until it has found all of them, throwing away subsumed clauses along the way until all that's left is the unique solution.

Either way you will always produce a path to a solution and proof that the solution is unique using elementary rules of propositional logic that anyone can recognize. That, I think, is what your parent means.

Of course, the existence of a proof doesn't mean that a human can find that proof, and that's what we usually mean when we say there's no logical solution. That is, nothing new can be proven by scanning the puzzle looking for pattern matches against a fixed library of known proof templates taken to be the universe of logical techniques, and nothing can be proven from less structured heuristic search within the limits of human working memory and patience.

[Barrin92]

guess and check for a human works when the problem space is small enough. You can't guess and check yourself through a 300-move forced mate in chess or some bizarre sudoku before dying of old age.

Yeah in theory a human can brute force or guess like a computer, but that's obviously not the point of a meaningful game or problem to solve because it's mindless.

[chongli]

Nothing about the guess and check strategy implies that you have zero prior information and must guess blindly in an attempt to win the lottery for picking the correct move. Typically, people using guess and check have other principles, rules, or heuristics they use to narrow the search space down to a handful of options.

My original point still stands. A cultural bias against guess-and-check is just another way of saying that strategies which determine an exact sequence of moves using only strict deduction from known theory are the only “acceptable” way to play sudoku. That, to me, is highly limiting and absurd. It’s also inefficient because often you’re faced with a choice between two mutually exclusive options and it’s just plain faster and less error-prone to guess and check than it is to carry out all of the deduction in your head before making the correct move.

[Barrin92]

it's not an absurd limitation. You're not playing sudoku to beat an opponent (idk maybe there's competitive sudoku where people guess if it's most efficient), you're playing to get better at reasoning. Deduction literally is the game, not filling squares with numbers.

If you're doing programming exercises you can brute force or guess a solution but there's a cultural bias against doing that, because what you're trying to learn is how to be smart about the problem, otherwise the exercise is pointless.

Guessing never adds anything to the experience in a game that's deliberately set up to be a test of skill. It's like doing an actual puzzle. Yes you can trial and error the edges of the pieces randomly, maybe that's faster, but it defeats the purpose of puzzling. A puzzle that heavily benefits from guesswork is just a bad puzzle.

[chongli]

If you're doing programming exercises you can brute force or guess a solution but there's a cultural bias against doing that, because what you're trying to learn is how to be smart about the problem, otherwise the exercise is pointless.

What exactly counts as “being smart about the problem” is culturally determined. In my view, taking ten times as long to solve the problem using pure deduction (because you’re struggling to hold all of the possibilities in your head simultaneously) is being less smart about the problem than making an informed guess and then quickly and mechanically checking whether you were correct.

A puzzle that heavily benefits from guesswork is just a bad puzzle.

That puts an upper limit on the difficulty rating of puzzles which is in some sense limited by the imagination of solvers.

[olliej]

Unless the whole point of the puzzle is to determine the logic.

Bruteforcing (aside from taking longer is just not interesting. When people are setting puzzles a lot of time and effort is put into how the logic is expected to work on the path to solving. When people post puzzles for testing if people find a spot where bifurcation is necessary they’ll generally point it out and the setter will check to see if they made an error or whether the person testing has missed some of the logic.

[chongli]

Guess-and-check is not brute forcing. I addressed this in another part of this thread.

[lupire]

IOW: intelligence is data compression.

[shagie]

Intelligence is the ability to craft a more optimal solution than a brute force approach would take.

In the domain of data compression, Kolmogorov complexity https://en.wikipedia.org/wiki/Kolmogorov_complexity is one of the approaches to try to measure that amount.

[chongli]

Guess-and-check is not the same as brute force (nor exhaustive search). No one is sitting there with an empty sudoku grid, filling the whole thing in with guessed numbers, checking to see if the solution is valid, then erasing everything and starting over. Literally no one does that, not even software sudoku solvers (it takes way too long).

Guess-and-check is literally just another name for backtracking algorithms. These do not work (for humans) without an adequate suite of rules that can be use to prune the search space.

[kazinator]

> New strategies are discovered

I highly doubt it. In Sudoku, it is either logically deducible that a particular value is to be placed into a given square. Or else it isn't; one can deduce that one of several values can be placed there, after all other exploration elsewhere has been exhausted.

Deducing what can be placed into what square is a simple elimination. If the square is in a row that already has 7, 7 cannot be placed there. If it's in a 3x3 square that has 5, 5 cannot be placed there. If it's in a column that has 4, 4 cannot be placed there. For every square of the Sudoku board, we can do this evaluation. We hit upon a square where all values are eliminated but one, and so that one is confirmed. Based on that confirmation, we can update the other eliminations and keep going. When this process cannot continue, there is no choice but to guess. OK, here we can have a 7 or a 4, and so it goes.

I don't see how you can pretend to have some strategies here, let alone ones with pet names and all.

People have different systems for tracking the information, to be sure.

[lgas]

Sure, you can always use backtracking to solve any sudoku, but people come up with new, more effective strategies all the time, eg. https://www.youtube.com/watch?v=ZLcey7qiXv8

[kazinator]

What is "more effective" than "solve any"?