A claim we can actually check
Everywhere else on this site, the difficulty label names the technique a puzzle requires. Here it names an absence. Every grid on this page has been run against all eighteen techniques we implement — singles, subsets, locked candidates, the fish, the wings, colouring, unique rectangles, almost locked sets — and none of them finishes it.
Some are more hostile than that. Sample the opening positions and you'll find grids with no naked single, no hidden single, no useful naked pair and no pointing pair anywhere on the board. There is no first move in the ordinary sense.
Every other level promises that a named technique will finish the grid; this one certifies that none of them can.
What remains true: there is exactly one solution, verified during selection. You are not being handed an ambiguous grid and told to pick.
Why patterns stop working
The techniques below this level all share a shape. Each finds a small, local contradiction — a digit with nowhere else to go, a set of squares that must consume a set of digits — and converts it into an elimination.
A well-made extreme puzzle is constructed so that no such local contradiction exists in the opening position. The information needed is spread thin across the grid, and no single neighbourhood carries enough of it. To get anywhere you have to combine implications that run through many squares.
Chains and forcing
The tool that survives is the chain. Take a square with two candidates and follow the consequences of one choice: if this is a 3, then that square isn't, which forces this other one, and so on. You're not committing — you're tracing.
Two things you might find at the end of the trace. The chain may return and contradict itself, which proves the starting assumption false and gives you a genuine elimination. Or two different starting assumptions may both force the same digit into the same square — in which case that placement is certain regardless of which was right.
That second form is forcing, and it's the most reliable way through these grids. It's also slow, and there's no honest way to make it fast: you're doing bookkeeping that a pattern would have done for you if a pattern existed.
Testing an assumption properly
At some point most solvers stop tracing and start testing — pick a square, commit to a digit, and play on until it works or breaks. There's nothing wrong with it, but there is a right way and a wrong way.
Choose a square with exactly two candidates, never three or more; you want a clean fork where disproving one proves the other. Pick one in a crowded region, so consequences arrive quickly and a contradiction surfaces sooner. And know your way back — on the board above, undo will walk you out of a failed line one step at a time.
What you should not do is guess in a sparse corner and grind forward for twenty placements. If the line is wrong you'll discover it a long way from home, and if it's right you won't be sure why.
Is that still solving?
Worth being honest about, because opinions differ and both are defensible.
One view: a proof by contradiction is a proof. Assuming a digit, deriving nonsense and concluding the digit was wrong is exactly the reasoning a chain performs — you're just doing it on the board instead of in your head.
The other: the pleasure of sudoku is that every step is known before you write it, and a test replaces knowing with trying. By that standard a puzzle needing one isn't a proper puzzle.
We take no position beyond labelling accurately. If you want grids that never require this, evil is as hard as we go while still guaranteeing a findable pattern at every step.
What the hint does here
It tells you the truth. When no technique applies, it says so rather than inventing a pattern, and offers a correct digit if you want one. Taking it is not failure — on some of these positions, no amount of looking will surface a step, because there isn't one to surface.