Compare Two at a Time

Nobody asked you for a number

The item says: rank these four from small to large, and pick the choice with the correct ranking. The answers are orderings of the labels. Not one of them contains a degree value.

That matters more than it sounds. If you look at a panel and think "that's about seventy degrees," you have done extra work and imported extra error: you converted a shape into a number you were not asked for, using a conversion you cannot calibrate, and you will now rank the numbers instead of the angles. Every judgment in this section is a comparison between two specific angles — larger or smaller, nothing else.

Comparison is also the operation you are good at. Human estimates of absolute angle are poor and drift with orientation; human judgments of "which of these two is more open" are far sharper. Ranking only ever needs the second kind.

Slide one onto the other

Two angles drawn on different headings are hard to compare as they sit, because your eye has to undo the rotation and judge the opening at the same time.

Two angles on the headings the item happened to hand you.

So undo the rotation deliberately. Pick one ray from each angle and imagine both angles turned until those two rays point the same way — then the only thing left to read is where the other two rays fall. Whichever one swings further round belongs to the larger angle.

The same two angles turned until their upper rays share a heading; only the lower rays still disagree, and the one that swings further round is the larger angle.

Here both angles have been turned until their upper rays share a heading. The lower rays no longer agree, and the one that reaches further round is the bigger angle. That is the whole comparison, and it survived the fact that these two panels still have quite different ray lengths — you were reading heading, not reach.

A second version of the same move, for people who do not visualize rotation comfortably: picture each angle as a hinge and ask which one is opened further. A hinge has no length — you cannot ask how long a hinge is — so the question forces your attention onto exactly the property being ranked.

Coarse anchors, while the spread allows

When four angles are spread widely, you can save comparisons by first sorting them against an angle you already carry in your head. The right angle is the reliable one; a half-right angle is a usable second.

A remembered right angle, then a widely spread set: two panels fall clearly under it, one clearly over, and one sits too near the line to call.

Against the right angle, panels 1 and 2 are obviously under, panel 4 obviously over, and panel 3 is close to it. That single pass has already told you panels 1 and 2 both sit below panels 3 and 4, which is four pairwise comparisons you no longer have to make. Two comparisons remain: 1 against 2, and 3 against 4.

But notice where the anchor failed. It could not tell you much about panel 3, the one nearest the line — and there is a reason that is always true. The two angles that sit closest to your anchor are neighbors in the ranking, so they are separated by one gap: at most 16° apart in an easy item, 8° in a medium one, 3° in a hard one. Whichever of them is nearer the line is therefore within a few degrees of it. An anchor bins the angles that are far from it and is silent about the ones that are near it — and the near ones are exactly the pair whose order is in doubt.

Why anchors fade and comparison does not

An anchor pays off in proportion to how far the set spreads out, and spread is precisely what difficulty removes. Tighten the four angles and sooner or later all of them land on the same side of the right angle, at which point the anchor returns the same verdict four times over.

A tight set: every angle is obtuse, so the right-angle anchor returns the same verdict four times and orders nothing.

That is not a rare corner. Across the generated items, roughly two-thirds of easy sets already have all four angles on one side of a right angle, and by the hard level it is about nineteen in twenty. Reach for the anchor when the spread hands it to you, and expect it to be gone by the time you actually need help.

Direct comparison has no such ceiling. It never referred to an external standard, so there is nothing for tightening to take away. It gets harder in the same way everything gets harder — the differences shrink — but it keeps returning the same kind of answer at every level, which is why it, and not estimation, is the operation to build the habit around.

Try it

Rank the four angles from SMALL to LARGE.

1 - 2 - 3 - 4

1 - 3 - 2 - 4

3 - 1 - 4 - 2

1 - 3 - 4 - 2

A generated item with a wide spread — the kind where an anchor pass earns its keep.

Take the anchor pass first: panels 1 and 3 read as clearly under a right angle, panel 4 as clearly over, and panel 2 as close to one. That already separates the pair {1, 3} from the pair {2, 4}. Now two comparisons finish the item — 1 against 3, then 2 against 4 — and neither one needs a number.

1 - 2 - 3 - 4

B ✓

1 - 3 - 2 - 4

3 - 1 - 4 - 2

1 - 3 - 4 - 2

In degrees: #1 = 69°, #3 = 79°, #2 = 95°, #4 = 111°. Ray length is a distraction — compare the openings near each vertex.

The same item with the answer marked and the true degrees reported.

Build the habit on unlimited generated items in the angle ranking practice generator.

The takeaway

You are asked for an order, never a magnitude, so never produce a magnitude. Compare two angles at a time: mentally turn them until one pair of rays agrees and read which remaining ray swings further, or ask which hinge is opened wider. A remembered right angle is a cheap first cut when the four angles are spread out, but it goes quiet exactly where the item gets hard, and pairwise comparison is what remains.

Summary