Two-String Kite
The Two-String Kite ties together a row and a column. Find a row where a digit has exactly two candidate cells, and a column where the same digit also has exactly two. If one cell of the row pair and one cell of the column pair sit in the same box — the knot where the two strings meet — the remaining two cells are the kite's loose ends, and any cell seeing both loose ends loses the digit.
Work it through with 6. Row 1 holds 6 only at r1c2 and r1c8; column 3 holds 6 only at r3c3 and r8c3. The cells r1c2 and r3c3 share the top-left box, so the strings are knotted there. The loose ends are r1c8 and r8c3. Now suppose r1c8 is not 6: then r1c2 must be, which forbids 6 anywhere else in its box, so r3c3 is not 6 — forcing r8c3 to be 6. At least one loose end always carries the digit, so r8c8, which sees r1c8 by column and r8c3 by row, can never be 6.
Two details keep the pattern honest. The meeting cells must be different cells of the shared box — if the row and column pairs share an actual cell, you have a different (simpler) situation. And both strings must be genuine conjugate pairs: exactly two candidates for the digit in that row and that column, no more.
Scan for kites alongside skyscrapers — both live on conjugate pairs and both are three-link chains. Where the skyscraper joins two parallel lines, the kite joins perpendicular ones through a box; between them they cover most short single-digit eliminations.
- Track conjugate pairs per digit: one row pair plus one column pair meeting inside a box is the whole pattern.
- The meeting cells must be distinct cells of the same box, not one shared cell.
- The elimination cell is usually the intersection of the loose ends' column and row — check it first.