Training your eye for runs

After you've solved enough nonograms — somewhere in the range of forty or fifty — your eye starts doing a thing the technique articles can't quite teach. You glance at a clue list and the runs already feel placed. The leftmost run sits where it has to sit; the gap between the runs is where you expect it; the rightmost cell is forced even though you haven't done the arithmetic. The arithmetic still works, but you're no longer doing it consciously.

This is pattern recognition catching up to the formal move, and it's worth being deliberate about because it's most of what makes hard nonograms feel solvable rather than computational.

A few patterns that tend to register first.

The first is clue lists summing to near the line length. A clue of 3 1 2 on a row of length 9 has runs taking 6 cells and gaps taking 2 — total 8 — which leaves 1 cell of slack across the whole row. That single cell can sit in any of three positions, and almost every combination forces several cells. You'll learn to recognise these "tight" lines instantly because the picture they paint is mostly determined.

The second is single long runs near the middle. A clue of 8 on a row of length 12 forces 4 cells of overlap regardless of position, and those 4 cells are in the middle of the row. Same for a 9 on a row of 12, a 7 on a row of 10, etc. Your eye learns to scan the clue lists for the largest single number relative to the line length and to start there.

The third is clue lists with a single small clue at one end. A clue of 1 7 on a row of 10 forces the 7 into a tightly constrained position — its leftmost legal start is cell 3, its rightmost is cell 4 — and the middle of the run is fully determined. The 1 is more ambiguous, but you've already done most of the work.

The fourth is mirror symmetry between row and column clue lists. Every once in a while you'll glance at a puzzle and recognise that the clue list 3 1 2 appears as both a row and a column clue. This is a coincidence at small sizes, but on larger grids it sometimes signals a deliberate symmetric subject — and once you've recognised the symmetry, you can use cells you've solved on one axis to bias the other.

Training the eye is mostly volume. Solve forty puzzles in a row; you'll feel different about clue lists by the end. Solve a hundred and you'll be able to glance at a 20×20 and predict the shape of the picture from the clue numbers alone, which is a small magic that doesn't quite stop being satisfying. Until then, keep doing the arithmetic deliberately. The pattern recognition that catches up is built on the formal move, not in spite of it.