Free adaptive drill — difficulty follows your answers

An adaptive paper, with the difficulty drawn as you walk it

The general-ability battery this page prepares you for is adaptive: it decides what to ask next from how you just answered, which is why it feels relentless, why two people who get the same number right can finish in different places, and why you cannot go back. This drill is free, runs eighteen generated items across the deductive, inductive and numerical families, and draws the difficulty as a live trace above the question so the mechanism is visible instead of merely described. It is preparation, not the instrument, and no published item is reproduced here.

100% free
No signup
18 adaptive items
3 item families
Live difficulty trace

Eighteen items across the three families the publisher’s general-ability test names, with one difference from an ordinary drill: the difficulty of the next item is decided by how you answered the last one, and the chart above the question draws that decision as it happens. Get one right and the next is harder. Get one wrong and it eases off.

Deductive: Rules given, conclusion required. The difficulty is the length of the chain and whether you have to run it backwards.
Inductive: Rule not given. A sequence of drawn cells changes by one, two or three properties at once, and you supply the one that comes next.
Numerical: A small table and a question that needs one, two or three steps on it. Difficulty is steps, not arithmetic.

What adaptive testing does to your preparation

Three things follow from it, and none of them are obvious from inside a sitting. The paper will feel hard the whole way through, because an adaptive test is trying to give you items you have roughly an even chance on — feeling stretched is the design working, not a sign of doing badly. Two candidates who answer the same number correctly can finish in different places, because a right answer on a hard item and a right answer on an easy one are not worth the same. And no two candidates see the same paper, which is why memorising answers is useless and why the publisher can afford to let you sit it unsupervised at home.

Practice items run flat at level 3 and never move the difficulty; they show the answer afterwards and reach no figure. Keys 1 to 4 answer, and every option is also a button.

How to read an adaptive run

Three practice items, eighteen scored, and a chart that moves under you.

Take the three practice items at a fixed level
They run flat at level 3 of 7, show the answer afterwards, and deliberately do not touch the difficulty — an untimed warm-up that moved the ladder would hand the first scored item a level you had not earned. Skip them if you have met all three families before. Nothing in them reaches any figure in the result.
Answer eighteen and watch the line above the question
Keys 1 to 4 answer and every option is a button. Above each item is the trace: a point per scored item, plotted at the level it ran at, filled where you were right and hollow where you were not. Right answers push the line up, wrong ones push it down, and the shape you should expect is a climb followed by a flat zigzag. The zigzag is the run finding your level.
Read the turnarounds, not the total
The panel names the level the run settled at and averages the last turnarounds into a figure with a spread beside it. That spread is the precision eighteen items actually bought — a wide one means the run was still climbing when it stopped. A one-up one-down ladder settles where you are right about half the time, so accuracy near 50% is the ladder working rather than a poor result.

Technical specifications

Items	18 scored, generated at run time from the difficulty in force, plus 3 optional practice items held flat at level 3 and excluded from every figure
Families	Deductive, inductive and numerical, rotated evenly so each contributes six of the eighteen — the three the publisher's own general-ability test names
Difficulty scale	7 levels, this page's own construction. Deductive difficulty is the chain length and whether the inference runs backwards; inductive is how many properties change at once, from one to three; numerical is how many steps the table question needs, from one to three
Adaptive rule	One step harder after a right answer, one step easier after a wrong one, from the shared trial module's staircase. That rule converges on the level where a candidate is right about 50% of the time, and the results panel states that target rather than leaving it implied
How it differs from the real thing	A published adaptive test selects from an item bank calibrated on thousands of people and re-estimates ability after every answer. This walks a fixed seven-rung ladder. The visible behavior is the same; the scale underneath it is not, and no level here converts to a percentile
Reported figures	Items right out of scoreable, the level the run finished at, the mean of the usable turnarounds with its standard deviation, and the hardest level reached in each family
What is excluded from the ladder	An item interrupted by this tab going to the background never reaches the staircase at all, so the difficulty is never moved by an item the visitor did not see. It is requeued and asked again at the end
What leaves the page	Nothing. Items are generated in the tab from a seed, no answer is transmitted, and the copy button writes the trace summary to your own clipboard

Frequently asked questions

Why did it feel hard from beginning to end?

Because that is what an adaptive test is engineered to feel like. A fixed paper spends its first third on items nearly everyone gets right, which is comfortable and tells the employer almost nothing; an adaptive one skips that by moving quickly to items you have roughly an even chance on, and then stays there. The consequence is that the subjective experience of an adaptive test carries no information about how it went. Candidates routinely come out convinced they did badly because they were unsure of half the questions, and being unsure of half the questions is the target state.

Why could I not go back and change an answer?

Because the question you were looking at was chosen on the basis of the answer you would be going back to change. Back-navigation is not a feature that was left out to annoy you: it is arithmetically incoherent on an adaptive test, since revising an earlier answer would invalidate the entire sequence of choices made after it. The practical consequence for a candidate is that the usual timed-paper strategy of skimming, skipping the hard ones and returning at the end does not exist here, so the decision on every item is final and has to be made at the time.

Why was I asked to sit a second, supervised test on site?

Because the design of this publisher's range pairs an unsupervised test taken at home with a shorter supervised one taken later, and the second exists to confirm the first was yours. Sitting an unsupervised reasoning test with a friend, a calculator you were not supposed to have, or somebody else at the keyboard produces a result that the supervised retest does not reproduce, and the discrepancy is the thing being looked for. This is worth knowing before the first test rather than after it: the honest strategy is to sit the unsupervised one exactly as you would sit the supervised one, because a first score you cannot repeat is worse than a lower one you can.

Two of us answered the same number correctly and got different feedback. How?

Because a right answer on a hard item and a right answer on an easy one are not worth the same thing, and an adaptive test scores on the difficulty of what you got right rather than on the count. Somebody who climbed steadily and then made two mistakes at a high level can finish above somebody who was correct on more items but never left the middle of the range. This is the single most counterintuitive property of the format and the reason a raw count is not reported back to candidates: the count is not the quantity the test is estimating.

Is there any point in memorizing questions from a practice site?

None, and the format is specifically built to make it pointless. Adaptive delivery means no two candidates see the same set of items, item exposure is controlled so that popular items are not shown endlessly, and the supervised retest exists to catch a first result that did not come from the person sitting it. What is worth practising is the thing that transfers: recognising a conditional chain that has to be run backwards, spotting that two properties are changing at once in a sequence, and knowing that a share of a total can fall while the underlying figure grows. Those are techniques, and techniques survive a new item.

What does the G in a general-ability test mean?

It refers to the single broad factor that scores on unlike reasoning tests tend to share: people who do well on verbal items tend, on average, to do well on numerical and abstract ones too, and a general test is built to measure that overlap rather than any one of them. In practice it means a battery that combines several families into one result instead of reporting them separately. The practical difference for a candidate is that you cannot rely on a strong family to be reported on its own — a weak inductive section is not quarantined from a strong numerical one, it is averaged into the same figure.

Does running out of time hurt more on an adaptive test?

It hurts differently, and the difference argues against rushing. On a fixed paper unanswered items are simply zeros, so speed converts directly into marks. On an adaptive one an unanswered item is a missing piece of information about where your level is, and answers given too fast to be considered push the estimate around in ways that take several later items to undo. The behavior the format rewards is a steady pace with committed answers, which is close to the opposite of the advice that suits a fifty-item speeded paper.

What adaptive delivery changes, and what the supervised retest is for

An adaptive test is a different instrument from a fixed paper wearing the same clothes. Behind it is a bank of items whose difficulty has been estimated on large samples in advance, and a running estimate of the candidate’s ability that is updated after every answer. The next item shown is the one that would sharpen that estimate the most — which, in practice, is one the candidate has close to an even chance on. The result is a test that reaches a given precision in far fewer items than a fixed paper would need, which is the commercial argument for it, and a test that cannot be worked in the usual way, which is the candidate’s problem. There is no skimming, no coming back, and no comfortable opening stretch. The drill above reproduces the behavior with a seven-rung ladder rather than a calibrated bank, and says so on the results panel, because the point of the demo is the shape of the walk rather than a number at the end of it.

The second structural fact is the one most preparation ignores, and it changes the strategy rather than the technique. The range is designed around an unsupervised test taken wherever the candidate happens to be, followed by a shorter supervised test taken under observation, and the second is there to establish that the first was the candidate’s own work. What that means in practice is that any advantage obtained on the first sitting has to be reproducible on the second, and help that is not available in a supervised room is not an advantage at all — it is a discrepancy waiting to be found. A candidate who sits the home test honestly and slightly below their best is in a far better position than one who produces a result they cannot repeat. Nothing about this makes practice pointless; it makes memorization pointless, which is a different claim.

The three families in the drill are the ones the general-ability test names, and each has a technique that transfers. Deductive items reward noticing whether the inference runs forwards from a fact or backwards from a denial, because denying the consequent is valid and denying the antecedent is the fallacy every distractor is built from — the same trap dressed differently on a raw cognitive count is a matter of arithmetic rather than logic. Inductive items reward counting how many properties are changing before you look at the options, since above the easiest levels checking a single property is never enough; the drawn version at length is the non-verbal reasoning page. Numerical items reward reading what the question is asking for before touching the table, because a share of a total and the underlying figure can move in opposite directions. If the process you are in is a flat speeded paper rather than an adaptive one, fifty items in twelve minutes is the other tempo to have felt, and the scenario format covers the part of a graduate sift that has no right answers at all.

A reaction time here is the interval between the frame that painted the stimulus and the timestamp the browser attached to your key, both read from the same monotonic clock. What neither can see is the display pipeline behind it, so on a 60 Hz screen roughly 16 ms of every figure below is the machine rather than you. That is the timing floor: two numbers closer together than that are the same number, and this page reports no precision it cannot support.

This is a measurement exercise, not a clinical assessment. It reports what you did on this page against a stated reference and nothing more — it cannot establish an attention difficulty, dyscalculia, or a reason you would qualify for extra time on the real assessment. Only a qualified professional, working with more than a browser, can make that judgment.

Not affiliated with, endorsed by, or connected to the owner of SHL. SHL is a trademark of its owner and is used here only to name the assessment this page prepares for. The questions on this page are our own: no part of the published test is reproduced, and a score here is not comparable to a score from the real instrument. Verify and Verify G+ are marks of the same owner and are named here on the same terms.

Where eighteen adaptive items are chosen

Every number on this page is worked out by JavaScript running in the tab you are reading it in. Your answers, your reaction times and your score are never uploaded, logged or kept — which is also why the test carries on working after you disconnect from the network, and why nothing here can be held back behind an email address.

The adaptation happens entirely in this tab: the level of your next question is computed from your last answer by JavaScript running here, so no server ever learns which items you were shown, let alone how you did on them. That is also why closing the tab loses the trace, and why the copy button exists.