What these tests actually measure, and what they do not
A score is a measurement of a skill, not a verdict on a person.
It is worth being straight about what abstract reasoning tests do, because both of the usual positions are wrong. They are neither a readout of your worth nor meaningless hoops.
What they genuinely measure
Abstract reasoning tests target what psychologists call fluid reasoning — the ability to find patterns and solve unfamiliar problems without relying on stored knowledge. That is distinct from crystallised knowledge, which is what you have learned.
The design goal is a test that does not advantage the student with the larger vocabulary, the better-resourced school or the more familiar cultural background. Using abstract shapes with no names is how they try to get there.
It works partly. These tests do correlate with problem-solving performance in genuinely unfamiliar situations, which is why they are used in selection at all.
Four honest limits
They are trainable
Scores improve substantially with practice — sometimes dramatically. That is awkward for a test claiming to measure something innate, and it is why a first score and a fifth score are not comparable. It also means a score partly measures exposure, which is not culture-free at all.
They measure a narrow slice
Pattern-finding under time pressure is one component of thinking. Not measured: persistence, creativity, judgement, collaboration, writing, memory for detail, or the ability to stay with a hard problem for a week. Most real achievement leans harder on those than on anything a matrix puzzle can see.
Speed is doing a lot of the work
Most of these tests are timed, so what gets measured is reasoning-per-minute rather than reasoning. Give the same person unlimited time and scores rise considerably. A careful, slower thinker can score lower than a quick, shallower one — and outperform them at almost anything that is not a timed test.
Working memory is a confound
As covered elsewhere, a meaningful share of the difficulty is holding several things in mind at once. Working memory capacity varies between people, is largely fixed, and is not the same thing as reasoning ability — but the test cannot separate them.
Which is exactly why the techniques work. Writing rules down, taking one attribute at a time, using the page — these do not cheat the test. They remove a limitation the test was never trying to measure, so what is left is closer to your actual reasoning.
So is practising worth it
Yes, for two separate reasons.
The practical one: if a test stands between you and something you want — a school place, a job — then improving at it is straightforwardly worth doing. The test is a gate, and gates are opened rather than argued with.
The better one: the underlying skill is real. Separating variables, checking one thing at a time, describing a relationship precisely enough to apply elsewhere, noticing what stays constant, resisting the first plausible answer. Those are genuinely useful, and they transfer to debugging, to experiment design, to arguments, and to any situation where you have to work out what is going on from incomplete evidence.
The puzzles are a compact way to practise that. The score is a side effect.
If your score is lower than you expected
It means you are currently slower at a narrow, trainable, timed pattern-finding task. That is all it can mean, because that is all the test can see.
It does not measure how good you will be at mathematics, at writing code, at building something people use, or at any of the things where being right matters more than being fast. Plenty of people who are excellent at all of those score unremarkably on timed reasoning papers, and the reverse happens just as often.
Use the score as a measurement of a specific skill you can improve, and nothing else. That is both more accurate and more useful than either treating it as a verdict or dismissing it entirely.