# Do you ever have to guess in a nonogram? Almost never > A well-made nonogram never requires guessing: one solution, reachable by deduction. What the stuck feeling really means, and when the puzzle is at fault. Canonical: https://getyournonogram.com/blog/do-you-ever-have-to-guess-in-nonograms Author: Mei Tanaka Published: 2026-07-30 --- On a well-made nonogram, you never have to guess — and every word of the qualifier matters. One solution, reachable by deduction alone, is a promise the maker builds into the grid, not a law the rules enforce. That splits the honest answer in two: on a good puzzle the next certain cell always exists, and the feeling that you must guess is nearly always wrong about which of you has failed. Nearly every tutorial gives the same advice — do not guess — and almost none says why you can afford to obey it. The reason is construction. [The rules themselves never promise one solution](/blog/nonogram-rules); a maker adds that property deliberately, and careful collections check for it. So when the grid in front of you seems to demand a coin flip, the demand is nearly always an illusion with one of three causes — and only the rarest of the three is the puzzle's fault. ## Do you have to guess to solve a nonogram? No — not on a properly made puzzle. A good nonogram has exactly one solution and an unbroken chain of certain deductions leading to it; every cell you mark can be proven before you mark it. Guessing enters only when a puzzle was generated carelessly and nobody checked that the chain exists. Honest is not the same as easy, though. [Wikipedia's solving overview](https://en.wikipedia.org/wiki/Nonogram) notes that many puzzles yield to reasoning about a single row or column at a time, while harder ones may demand "what if?" reasoning that spans more than one line at once. Deeper is not uncertain. The chain of proof still runs from the first mark to the last; it just passes through more cells before it pays. ## A stall that looks like a guess Here is the moment most coin flips happen, on real numbers. 1. **The stall.** A row ten cells wide carries the clue 3. Earlier work has proven cells 1, 2, and 10 empty, and cell 6 filled. The clue is a complete inventory with a single run, so that run must cover cell 6 — which leaves it exactly three legal homes: cells 4–6, 5–7, or 6–8. 2. **The dead end.** Compare the three homes and the only cell they share is 6 — the one you already have. The row alone has nothing more to say. This is what a forced guess feels like from the inside: three placements, one row, no way to choose. 3. **The turn.** Read the crossing lines instead. Column 5's clue is a single 2, and both of its cells were finished higher up the grid. A finished clue empties the rest of its line, so cell 5 of our row is provably blank. Mark the X. 4. **The collapse.** The homes 4–6 and 5–7 both needed cell 5; both are dead. Only 6–8 survives — fill cells 7 and 8, then close the row with X's at 3, 4, and 9. The row was never undecidable. It was waiting for one fact from the other direction, and the guess it seemed to demand would have been a coin flip over a question that already had an answer. ## What is the stuck feeling actually telling you? One of three things: a deduction is waiting in the direction you are not reading; an earlier mark is wrong; or — rarest by far — the puzzle itself is under-made. The first two account for nearly every stall between them; work the list in that order and the coin never gets flipped. The first cause is the stall above, generalized. Every mark you set in a row lands in some column, so certainty ripples across the grid in both directions, and [the fastest solving alternates between them](/blog/how-to-solve-nonograms). A solver who reads only rows is stuck twice as often, for half the reason. The second wears the first's costume. If every line you visit refuses to move, stop reading forward and start auditing backward: re-count your marked lines against their clues, X's included. [The line-by-line discipline](/blog/the-line-by-line-method-slowly-explained) treats that re-reading as half the method, because half of all stalls are a mis-read clue sitting a few moves in the past, quietly poisoning everything downstream. The third cause is real, but it has to be earned — it gets its own section below. ## Is testing a cell just guessing with extra steps? No — the difference is bookkeeping. A guess commits ink to a cell and plays on as if it were true. A test assumes a value deliberately, follows the consequences, and stands ready to unwind every one of them; when the assumption collapses, its opposite is proven. The first is chance. The second is logic's last tool. The reference pages record the same move. [Wikipedia's "contradictions" method](https://en.wikipedia.org/wiki/Nonogram) is exactly this — try a box in an empty cell, solve onward, and an error proves the cell is a space — down to its advice to work in pencil. The pencil is the whole philosophy in one object: what makes a test legitimate is the readiness to erase. On a well-made puzzle you will rarely need the move at all, and reaching for it early usually means the first two diagnoses were skipped. ## When it really is the puzzle Badly made puzzles exist, and the data on them is blunt. Jan Wolter, who ran the webpbn puzzle archive, compared puzzle sets for [his survey of computer solvers](https://www.webpbn.com/survey/index.html): among the archive's human-designed black-and-white puzzles, 81.8 percent could be solved a line at a time to a unique solution, and only 2.6 percent allowed multiple solutions. Among puzzles generated at random — every cell of a 30×30 grid painted black or white on a coin flip — 0.4 percent were line solvable, and 97.5 percent had more than one answer. The no-guessing property is manufactured, and mass generation without a checking pass manufactures its absence just as reliably. That asymmetry is your protection and your test. Before blaming the maker, earn the accusation: audit every line against its clue, in both directions, X's included. If the audit is clean and two different endings both satisfy every clue, the grid was under-clued — the failure belongs to its constructor, and finishing either way says nothing about your logic. Change puzzles, not standards. The next certain cell exists. On every honest grid that sentence stays true from the first mark to the last, and the whole craft of solving is trusting it one minute longer than frustration wants you to. Guessing is not a moral failing; it is answering a question the grid never asked.