Proportion Traps

Two ways an opening gets the size wrong

Wrong openings are built two ways, and both can leave the proportions off. One is a straight copy of a true outline, shrunk. The other is the shadow of a different object — this object with a single feature altered — which throws the proportions off wherever that feature shows. Neither is a freehand distortion: every wrong opening is the outline of something real, or a scaled copy of one.

The shrunken copy

About three items in five carry one. The generator takes a true view of the object — sometimes the very view that is the answer — and scales it by a factor between 0.68 and 0.82, uniformly on both axes at once. The copy comes out 18 to 32 percent smaller than the shape it was taken from, with every proportion inside it untouched.

Nothing about the outline's character has changed: same corners, same steps, same internal ratios. Only its size. That makes it the one trap a ratio check cannot touch, and the one trap a size check settles instantly.

Never a hairline

The floor of that range is the point of it. The gentlest shrink still takes 18 percent off both dimensions, which costs about a third of the opening's area. And no wrong opening of any kind may look nearly like the answer: each has to differ from the correct outline by at least a tenth on the generator's similarity measure — the largest of their relative differences in area, width, height and corner count — and by the same margin from every other option. Candidates inside that margin are discarded and rebuilt. The same measure caps the difference too, so a wrong opening is never a wild shape either.

Twins with a feature resized

The remaining wrong openings are near-miss twins. The generator rebuilds the object from its own recipe with one feature edited, then traces a true shadow of that variant. One of the four available edits is a resize — a feature stretched or squashed along a single axis, multiplied by roughly 1.25 to 1.65 upward or 0.45 to 0.8 downward.

A twin's proportions are wrong only where the edited feature shows, which is exactly why it survives a glance at the overall silhouette. Note which way these lean. The generator tries the shrinking edits first, because growing a feature usually grows the shadow, and an opening larger than a true outline still admits the object — so those candidates fail the passability check and are rejected before you ever see them.

Judge size against the other openings, never against the object

Here is the rule that makes size usable at all. The five openings are drawn at one shared scale — the generator measures the largest of them and fits all five to that one number — so comparing openings against each other is exact. The object's drawing is not part of that arrangement: it is fitted to its own box. An opening that looks small beside the object's drawing may be perfect; an opening that looks small beside the other four is small.

Size alone will not settle a twin, though, because only one of its features is off. For those, take one measurement across and one up within a single drawing and compare that ratio against the object's: a feature stretched along one axis changes every ratio that crosses it. A later topic turns that into a routine you can run in seconds.

On a real item

Here is an item from the generator that serves the practice page, with the answer hidden. Compare the five openings against each other for size first, then hunt for the one feature that does not match.

The takeaway

Two families put the proportions wrong. A shrunken copy of a true view keeps every internal ratio and loses 18 to 32 percent of both dimensions at once — catch it by comparing the openings against each other, never against the object's drawing, which is on an unrelated scale. A near-miss twin has one feature resized, so its proportions are wrong only locally — catch that by checking a height against a width where the feature shows.