What the split test measures
A shape appears and you drag a single straight line all the way across it. The line is then used to divide the silhouette into two pieces, the area of each piece is measured, and you are scored on how close the split came to even. Five shapes make one run, and your result is the average.
The task looks trivial and is not. Human area judgement is poor in a specific, predictable way: we read a shape by its outline and its width, and we systematically under-weight regions that are short but wide while over-weighting parts that are long but thin. A shape with a fat lobe on one side and a narrow tail on the other will look balanced well before it actually is.
There is a neat piece of mathematics behind the puzzle. For any bounded region, and for any direction you choose, at least one line in that direction divides the area exactly in half. This follows from continuity: slide a line across the shape and the fraction on one side moves smoothly from zero to one, so it must pass through a half somewhere. A perfect cut always exists at whatever angle you pick. Finding it by eye is the hard part.
What counts as a good cut
| Split | Accuracy | Verdict |
|---|---|---|
| 50 / 50 | 100% | Perfect |
| 48 / 52 | 96% | Excellent |
| 45 / 55 | 90% | Good |
| 40 / 60 | 80% | Fair |
| 30 / 70 | 60% | Poor |
| 20 / 80 | 40% | Well off |
| Average over five cuts | Percentile |
|---|---|
| Below 76% | Bottom fifth |
| 76 – 82% | Below average |
| 82 – 88% | About average |
| 88 – 93% | Above average |
| 93 – 96% | Strong |
| Above 96% | Top few percent |
How to cut closer to even
- Aim for the centre of mass, not the middle of the shape. The midpoint of the bounding box is almost never the point that halves the area. Look for where the shape is heaviest and move your line toward it. This single correction is worth more than everything else combined.
- Weigh width more than length. A short, wide bulge usually carries more area than a long, thin limb, even when the limb looks bigger. Ask which side would be heavier if the shape were cut from steel plate, not which side looks longer.
- Pick your angle before you pick your position. A perfect cut exists at every angle, so the angle is free. Choose the one that gives you the simplest comparison, usually the one that cuts across the shape's bulk rather than along it.
- Distrust symmetry that is only partial. Shapes that are nearly symmetrical are the ones people miss most, because the eye settles on the axis of symmetry and stops checking. The small asymmetry is exactly what moves the true line.
- Squint. Blurring the outline removes the detail your eye is over-reading and leaves the distribution of mass, which is the thing you are actually trying to judge.
Frequently asked questions
How is the split measured?
Each silhouette is a polygon. Your line is used to clip that polygon into two pieces, and the area of each piece is computed geometrically. It is exact rather than estimated from pixels, so a 49.7 / 50.3 result really is 49.7 / 50.3.
Is a perfect 50/50 cut always possible?
Yes, and at any angle you choose. For any bounded shape and any direction, there is at least one line in that direction that halves the area. Sliding a line across the shape moves the fraction on one side continuously from none to all, so it must cross exactly half at some point.
What is a good score?
An average above 88% over five cuts is above average, and above 93% is strong. Perfect runs essentially do not happen, because five shapes in a row leave no room for a lucky guess to carry you.
Why do my cuts always land off to one side?
Almost certainly because you are aiming at the middle of the shape rather than the point that balances its area. Most people cut through the centre of the bounding box, which is biased toward whichever side has the longer thin section.
Are the shapes always the same?
No. Runs alternate between a set of drawn silhouettes and shapes generated fresh each time, so the pool cannot be memorised.
Does this measure anything real?
It measures visual area estimation, which is a genuine and well-studied perceptual skill. It is not an intelligence test and it does not predict anything beyond how well you judge area.