Drawing two things to scale without lying
Two silhouettes on one ground line is the oldest way of showing a size difference and still the clearest. It is also easy to get wrong in a way nobody notices, because a drawing that flatters the eye looks more convincing than one that matches the numbers.
The wrong way, which looks better
The obvious approach is to take each silhouette, fit it to a box, and set the boxes at the right relative size. It produces a balanced picture and it is wrong, because the box is not the measurement.
An outline has a width and a height in whatever units it was drawn in. Only one of those corresponds to the figure the object is judged on. Scale by the box and a long low shape and a tall narrow shape both get stretched to fill their allocation, and the ratio you were trying to show is quietly destroyed.
The right way
Pick the measured axis for each object. Compute how many screen pixels one metre is worth, once, for the whole picture. Then scale each shape so that its measured axis comes out at its real size times that number.
For a whale, judged on length, the horizontal extent of the outline becomes 27 m worth of pixels. For a tower, judged on height, the vertical extent becomes 330 m worth. Whatever the other dimension of each shape turns out to be is simply what it is, and it is not adjusted.
That is the entire method. It is three lines of arithmetic and it is the difference between a drawing that supports a claim and one that decorates it.
Every drawing on this site uses one scale for both shapes and one ground line under both. The ground line matters more than it sounds: a shared baseline turns a size comparison into a length comparison on a flat surface, which the visual system handles well even at large ratios.
Why the correct version sometimes looks wrong
Because the eye reads bulk, and bulk is not size.
A blue whale at 27 m against a T. rex at 12 m is a ratio of 2.25, and the drawing shows a vast horizontal mass beside a compact upright animal. The whale occupies far more than 2.25 times the ink. That is correct and it feels exaggerated.
The reverse case is more disconcerting. The Leaning Tower of Pisa at 56.7 m against the same whale is 2.1, and the tower, being narrow, occupies less area on the page than the whale it is more than twice the size of. A reader glancing at that drawing would guess the whale is larger. The numbers say otherwise and the numbers are what the drawing was scaled from.
This is not a flaw in the method. It is the illusion the method exists to expose. A drawing that flattered the eye by widening the tower would produce agreement between the picture and the intuition, at the cost of agreement between the picture and reality.
The baseline, and what it implies
Standing both shapes on one ground line is a strong convention and it carries a hidden assumption: that both objects are on the ground.
For a tower and an animal that is fine. For a whale it is a fiction, and for anything that flies it is close to meaningless. A drawing of Quetzalcoatlus with a 10.5 m wingspan standing on a ground line shows a span, in a pose the animal held in the air, resting on a surface it spent little time on.
The alternative, floating each object at its own natural elevation, would be more truthful about habitat and far less useful for comparing size, because the eye would then have to judge two lengths that do not share an origin. The ground line wins on balance, and it is worth knowing what it costs.
How much detail an outline needs
A silhouette has to do two jobs at once: be recognisable instantly, and be honest about proportion. Those pull in opposite directions more often than you would expect.
Recognisability pushes toward exaggeration. A stylised shark with a bigger dorsal fin reads as a shark faster. Honesty pushes toward the real outline, fin and all at its actual proportion. The catalogue goes with honesty, which is why some shapes take a moment to identify and why a few, seen alone, are genuinely ambiguous.
There is a technical version of the same trade-off. The geographic outlines carry enormous numbers of points, because a real coastline is fractal and never stops having detail. Drawing every one of them would make a page heavy for no visible gain, so the outlines are reduced to the point where the removal is smaller than a pixel at the size the drawing appears. The shape you see is the real shape at the resolution you can see.
Two colours, and why
One shape red, one blue, throughout the site and throughout the game. It is not decoration. A comparison with two shapes of the same colour forces the reader to work out which is which from context, and at ratios near one that is genuinely hard.
The colours also carry a meaning inside the game: blue is the reference held at true scale, red is the shape you resize. Keeping the same pairing on every comparison page means the drawings read the same way whether you arrived from a round or from a link.
Checking a drawing you did not make
Two quick tests will catch most bad ones.
Find the stated figures and measure the picture. If the two objects are stated at a ratio of three and the drawing shows two, the drawing has been adjusted.
Check which dimension was used. If a tall object and a long object both fill the same frame height, the picture was scaled to the box rather than to the measurement, and the ratio it appears to show is not a ratio of anything.
Neither test takes more than a few seconds, and between them they disqualify a surprising proportion of the size comparisons in circulation.