A tangram starts as one square and ends as seven pieces: two large triangles, one medium triangle, two small triangles, a square and a parallelogram. Nothing in that list sounds inevitable. Why not six, or nine? The answer is hidden in the square the pieces came from, and it is an answer about counting rather than taste.
Draw the diagonals of the original square, then keep halving. The whole square divides into sixteen identical small right isosceles triangles. Now measure the pieces against that grid: each large triangle covers four units, the medium one covers two, each small triangle covers one, the square covers two, and the parallelogram covers two. Four plus four plus two plus one plus one plus two plus two comes to sixteen. Every piece is a whole number of the same unit, and every edge lies along one of only a few directions.
Why whole units matter
That shared grid is what makes the puzzle solvable by hand. Because the edges are all multiples of one of two lengths, pieces meet flush in a surprising number of ways. A cut that produced thirds or fifths would give you more shapes in principle and far fewer satisfying joins in practice. You would spend the afternoon hunting for an alignment instead of recognising one. The grid turns a geometry problem into a small, finite set of moves.
It also explains why all seven pieces must be used. With a shared unit, the total area is fixed at sixteen, so every silhouette you make has the same area whatever it looks like. A cat, a boat and a standing person are all sixteen units rearranged. That constraint is the puzzle: you are not choosing how much material to use, only where the same material goes.
Why seven, and not more or fewer
With three or four pieces the number of recognisable silhouettes is small enough to exhaust in an afternoon, and most of them look like the pieces themselves. With a dozen or more, the combinations multiply and the rule about using every piece stops biting — there is almost always some arrangement that will fill a given outline, so the puzzle turns into patient trial rather than insight. Seven sits in the narrow band where the search is genuinely hard but the space of answers is still small enough for a person to hold in their head.
One number shows how tight that band is. If you require all seven pieces, require them to touch edge to edge, and ask for a shape with no inward corners at all, the number of possible convex silhouettes has been counted exhaustively: thirteen. Not thirteen hundred. The freedom people feel while playing is real, but it is freedom inside a very small room, and that is exactly what makes a successful solution feel earned.
The parallelogram is the odd one out and worth turning over in your hand. Unlike the triangles and the square, it is not identical to its own mirror image. Some silhouettes can only be completed if that single piece is flipped, so the puzzle quietly contains a left-handed and a right-handed version of itself. Beginners often conclude a figure is impossible when they have in fact been refusing to turn one piece over.
Like board games that travelled faster than the goods that carried them, the tangram is a portable set of rules: a flat square, a cut, and an instruction that can be explained without a shared language. The version known today took shape in China as a rearrangement puzzle, related to older traditions of fitting banquet tables together. Its material is almost nothing — a thin board, a scrap of paper — which is part of why it survived long enough to be everywhere.
There is a family resemblance to interlocking bricks, where tolerances decide what can be built, but the tangram runs on a stricter economy: no piece may be left out and no piece may be cut again. Games that change slowly tend to be games whose rules are already at a minimum. Seven pieces is that minimum with just enough room left for surprise.
So the number is not a convention. It is what you get when a square is cut so that every edge agrees, and then cut no further than a person can keep track of. Add a piece and the counting gets easier. Remove one and the counting stops being interesting.