Decode the shape: four questions
Two students on the same syllabus can be preparing for two different exams. The syllabus names what is covered; it says nothing about how it is asked. What decides your preparation is the shape: the form an answer must take, and the rule that turns it into marks. Four questions pull that shape out of any exam. The answers come only from the notice and one past paper, never from what other students tell you. They place the exam among six shapes; most exams are a mix of two.
Decode the shape before you start studying. What you practise carries only to work that looks like the practice: writing chosen options all evening does not prepare a hand for a scheme-marked answer. Once you know the shape, you practise in that form, at the level the marks sit. Read the notice again every year — the shape is the one thing the exam body is free to change.
Pick one exam from your list, today. Open its current notice and one past paper side by side. Answer the four questions cold. What form must you answer in, what does the marker pay for, what does a wrong or missing answer cost, and how many questions does each minute carry? Write the shape you land on in one line.
Decoding the shape is one more thing to do before I can start actual studying — I would rather just begin.
Ten minutes with a notice and one past paper is far cheaper than a term spent practising the wrong form. A student who writes long paragraphs all year for a paper that pays for named points has not saved time by skipping this step; she has spent months practising a skill the exam does not score. Decoding comes first because it decides what the rest of your studying is for.
Write to the marking scheme
A scheme-marked paper does not pay for knowing the topic; it pays for named points, the exact ones on the marker’s list. For each question the marker holds a set number of points, drawn from a term, a labelled step, a unit, a named part of the answer, and nothing beyond that list earns anything, however well written. So get the scheme itself, from a published marking scheme or a teacher who marks, and write every practice answer straight against it.
This works because the marker is not reading for understanding; in the few seconds each script gets, they are scanning for the terms and steps their own list names. A paragraph that shows real understanding but never states the labelled term earns nothing, while a few short lines carrying the right terms earn full marks. Matching the count of lines to the count of marks turns a vague answer into a checklist the marker can tick fast, which is exactly what a scheme is built to reward.
Writing to a scheme feels thin at first, almost like cheating the subject: you leave out everything you know beyond the named points, and a short, listed answer does not feel like the full page you are used to producing. That thinness is the method working, not a sign you are studying less. The scheme, not your effort, decides how many points a question carries, and a paragraph beyond its list adds nothing to the score. My own first scheme-marked answer looked bare beside the page I would have written a term earlier. It scored higher.
Weaker. Meera has a question on the water cycle worth five points: describe the stages. She writes a full paragraph, in her own order, covering evaporation, clouds, rain, rivers and the return to the sea, with two extra lines on why it matters to farmers. It reads well, and she is proud of it. The marker’s list wants five named stages, one line each, in the terms the syllabus uses. Meera’s paragraph has three of those terms buried inside long sentences, with no clean line the marker can tick against any of them. She scores two of the five, for a page she was sure covered everything.
Stronger. Same Meera, same water-cycle question, the next test. She gets the scheme first, from her teacher’s marking key: evaporation, condensation, precipitation, collection, and the return to the sea, one point each. She writes five short lines, one stage per line, naming each stage in the scheme’s own words, no extra paragraph around them. It takes half the time her first answer did. In the test the marker’s eye finds each named stage in its own line and ticks it in seconds. She writes the same five lines, in the same order, under the same clock, and this time every point lands.
- Get the scheme. Find the board’s published marking scheme for that question type, or ask a teacher who marks, for exactly what earns each point.
- List the points. Before you write, note how many points the question carries and the labelled parts or steps the scheme wants, one point per line.
- Write one point per line. Answer with the page closed, in the scheme’s own terms, matching the count of lines to the count of marks, nothing extra added.
- Mark it yourself. Check your answer against the scheme, point by point, and circle every point you left out or wrote in the wrong terms.
- Fix the gaps and repeat. Rewrite the same question from the list of what you missed, then try a new question of the same type from scratch.
Take one recent answer you wrote and lost marks on. Find or rebuild its scheme from a marking key or a few past papers of the same question type. Count how many of the scheme’s points your answer actually names, one by one. Then rewrite the answer as a list of exactly that many lines, each naming one point the scheme pays for.
A page of short, listed lines feels lazy, like I am not showing everything I know about the topic.
It is not laziness; it is aiming at what actually earns marks. The scheme pays for the named points, not for volume, and a longer answer that never states them scores no higher than a short one that does. Keep the extra understanding for revision and for questions that genuinely ask for it — on a scheme-marked question, every line beyond the scheme’s own points is effort spent on something the marker cannot score.
Building a scheme from past papers before I even start writing feels like it eats into study time I should spend learning.
It costs one evening, once, for each question type, and it saves far more than that afterwards, because you stop writing full pages that score half of what they could. A scheme built from a few years of past papers holds for the whole type, not just one question, so the same points keep paying off across every practice answer of that kind. Skipping this step just means guessing at what earns marks, test after test.
MCQ under the clock, with a penalty
A timed multiple-choice paper with a penalty scores two things: what you know, and whether you know what you do not. The penalty prices every question as a bet: with four options and a third of a mark off for a wrong one, a blind guess breaks even, and ruling out one option tips it your way. So the rule is neither attempt everything nor only what you are certain of — attempt what you can eliminate to, and leave the rest blank.
The paper is built from near-misses on purpose: an option is right only because the other three are wrong in some specific, checkable way. So the practice that builds the skill is not answering more questions; it is producing, for every wrong option you meet, the exact reason it fails, with the answer key opened only after you have tried, never before. Do this enough times and ruling out an option stops being a guess and starts being a check you can run under the clock.
Leaving a question blank on purpose feels wrong while the clock is running; every unanswered row looks like marks walking away. It is not. A blind guess under this penalty gains nothing on average and loses on a bad day, so the blank is often the winning move, not a failure of nerve. What is genuinely uncomfortable is tracking two numbers apart afterwards: how many questions you attempted, and how many of those you actually got right. Seeing that gap plainly, the first time, stings more than any blank ever does.
Weaker. Farhan sits a practice set of multiple-choice questions with a wrong-answer penalty and answers every single one, sure or not, because an attempted question at least has a chance. On the ones he is unsure of, he circles whichever option feels most familiar. When he checks the key, he has got a healthy share right, but the wrong ones, mostly blind guesses on questions he could not eliminate at all, have taken a third of a mark each. His final score comes out lower than the number of correct answers alone would suggest, and he cannot say afterwards which of his sure answers were actually sure.
Stronger. Same set, same clock, a week later. Farhan runs it in three passes: first the questions he is certain of, answered straight away; then the ones where he can rule out one option, worked through with the reasoning noted in the margin; the rest left blank, deliberately, without a second look back. He checks the key afterwards and counts two numbers: how many he attempted, and how many of those he got right. Almost every attempted one is right, because each was either a sure answer or a ruled-out bet that paid off. The blanks cost him nothing extra, and his score finally reads as good as his actual knowledge.
- Read the notice first. Check the current number of options and the wrong-answer penalty for this paper, since the break-even point moves with both.
- Pass one: the sure ones. Answer everything you know cold, straight away, and move on without lingering on any single question.
- Pass two: eliminate, then answer. For each remaining question, rule out every wrong option you can name a reason for, then answer if at least one option is gone.
- Leave the rest blank. If you cannot eliminate a single option, leave it blank on purpose. Come back only if time is left and your paper’s penalty arithmetic says a guess still pays.
- Mark and split your score. Check the key and count attempted against correct separately; the gap between the two is what to fix next time.
Take a short set of multiple-choice questions you have not seen, with the penalty on. Run three passes: sure ones first, then ones you can eliminate to, then leave the rest blank. Mark it against the key and note two numbers: how many you attempted, and how many were right. The second number is your calibration; the penalty prices it in marks.
Leaving a question blank feels like giving up marks I might have got right by chance.
A blind guess under this penalty gains nothing on average, even though any single guess might land. What feels like giving up marks is refusing a bet that does not pay. The moment you can eliminate even one option, the arithmetic flips and the guess starts paying. So the blank is for questions where nothing has been ruled out, not for every hard question you meet.
Stopping to eliminate options on every question is slower than just picking an answer and moving on.
It costs seconds per question and saves marks across the whole paper, because an eliminated-to guess pays on average while a blind one does not. The elimination itself gets faster with practice, since you are drilling the same near-miss reasoning the paper reuses question after question. Picking fast without eliminating is not actually faster once the penalty is subtracted from every wrong pick it produces.
The speed paper: pick, then answer
On a speed paper the questions are many, cheap and equal, and the marks go to whoever answers the most correctly in the time. A hundred questions in an hour is under forty seconds each, so the shape being tested is not knowledge but automaticity — a fact pulled out as one chunk, a step run without thinking, a type known on sight. So you run two passes. Pick first, answering what comes at sight and marking the rest; then a second pass, cheapest first.
Anything you have to work out here costs about three questions’ worth of time, so speed is not a bonus, it is the whole test. That is why practice always runs on a clock. Timed sets, the seconds per question actually measured, and the slow type drilled alone until it stops needing thought — deliberate practice aimed at speed, not at difficulty. With or without a penalty the pick rule holds; a penalty only moves where the second pass stops.
Timing yourself on a fresh type feels bad. You are slower, you miss questions you would have got right untimed, and the accuracy on the sheet looks worse than the day before. That drop is the honest number; untimed practice was only hiding it. The discomfort is the work — drilling a weak type again and again, on the clock, is what turns a worked-out step into one you recognise. I ran the same forty questions timed for a week; the log, not the feeling, showed which type was slow.
Weaker. Karthik has a hundred-question science mock on Saturday morning, ninety minutes on the clock. On Thursday evening he sits with the same hundred, no timer, working each one slowly and checking the textbook when unsure. He gets ninety-four right and closes the sheet, satisfied. On Saturday the ninety minutes end with forty questions still unread; the ones he did answer he answered as carefully as Thursday, at nearly three times the seconds each one gets. The score that comes back is fifty-one, not ninety-four.
Stronger. Same Karthik, same mock, the week before. He runs the hundred questions with a sixty-minute timer, marking anything that would take a second look instead of stopping to solve it. First pass: sixty-one answered at sight. Second pass, cheapest first: eighteen more from the marked set, in the twelve minutes left. He checks the answers, finds his weak type is unit-conversion questions, and drills twenty of those alone, untimed then timed, until each one takes under twenty seconds. On Saturday the real ninety minutes feel roomy, and the sheet comes back eighty-nine.
- Time every set. Give each practice set the exact minutes the real paper allows, and note the average seconds each question is actually getting.
- Pick first, don’t solve. One fast pass through the set. Answer only what comes at sight; mark anything that needs working out, and move on immediately.
- Second pass, cheapest first. Return to the marked ones in order of how fast you can finish them, not the order they came in; skip one the moment it drags.
- Isolate the slow type. Find the question type that ate the most seconds this set, and drill twenty of that type alone, untimed once, then timed.
- Log attempts and accuracy. After every set, write down how many you attempted and how many of those were right, as two separate numbers, not one score.
Pull any twenty-question practice set you already have. Set a timer for the time a real paper gives twenty questions, no more. Run the pick-first pass, mark what you skip, then the second pass, cheapest first. When the timer ends, write down two numbers — attempted, and correct — instead of one score. That gap is where next week’s drilling goes.
Racing through a set on the clock feels like guessing fast, not like actually learning anything.
Some of it is guessing, and the two numbers tell you which part. If attempts go up and accuracy holds, the extra speed comes from recognition you built drilling the slow type — that is automaticity, not luck. If accuracy falls as attempts rise, you are only guessing faster, and the fix is the same drill, done slowly first, then timed again.
Drilling twenty questions of one weak type over and over feels slower than moving on to fresh material.
It is slower today and cheaper every day after. A type you have not drilled costs three questions’ worth of time each time it appears, for as long as it stays untrained — that adds up over a whole paper of it. Twenty questions drilled once until they come at sight removes that tax for every set you run afterward, which is where the time actually gets paid back.
The long answer: a bank and a clock
A long written answer is a skill with knowledge inside it. The rater reads hundreds of scripts, so marks go to what shows in a minute’s read — a structure visible at a glance, every part of the question covered, specific content, not general sentences, and the length asked for, in the time allowed. The knowledge shape is a bank — for each theme, a few facts, an example, a number, an argument on each side, held as chunks you can pull in a minute.
A topic you understand but cannot put on the page in ten minutes scores as if you did not know it. So bank and skill train apart. Build the bank from what past papers keep asking, and drill it page-closed, like any by-heart material. The skill grows only by writing — one timed answer a day, to the word limit, marked against a model or a reader, then one thing fixed before the next. Read the verb, since discuss and critically examine pay for different structures, and write the structure as headings before a line.
Writing one timed answer a day is slower and harder than reading the theme again, and the first ones come back badly cut — a paragraph you were sure was strong loses marks for missing one part of the question, or for general words where a model answer has specific ones. That cut is the point — it finds the gap before the real exam does. I revised one theme cold for a fortnight; my first timed answer on it came back with three of five parts missing, because ten minutes had never made me choose which three to write first.
Weaker. Zara has a long-answer paper in a fortnight, six themes she needs cold. Every evening she reads one theme from her notes, underlining the arguments on each side, and by the end of the first week she can talk about any of the six for ten minutes without stopping. She feels ready; the notes are full, and nothing in them surprises her when she re-reads. She has not written a single full answer to a real question in the whole fortnight, timed or otherwise.
Stronger. Same Zara, same six themes, the fortnight before. She spends twenty minutes an evening turning each theme into a bank — three facts, one example, a number, an argument on each side, written as short chunks on one card. Then she picks one past question a day, reads the verb, writes the structure as headings, and writes the full answer against a ten-minute clock. A teacher marks the first three; each comes back missing one part of the question, which she then fixes before writing the next. By day ten she is covering all four parts inside the ten minutes, from the bank alone.
- Build the bank. For each theme, write a few facts, one example, a number, and an argument on each side, as short chunks you could recall in a minute.
- Read the verb first. Before writing, check whether the question says discuss, explain, or critically examine — each one pays for a different structure.
- Structure it, then write. Put the answer’s shape down first, as headings on the page, so the rater sees the structure before reading a line.
- Write to the clock. One full answer, timed to the exact minutes and length the exam allows, with no stopping to check the notes.
- Mark it, fix one thing. Compare against a model answer or a reader who marks, note the single biggest gap, and fix only that before the next one.
Pick one theme you feel confident about. Give yourself five minutes to write it as a bank — three facts, one example, a number — on a card, nothing else. Set a ten-minute timer, pick a past question on that theme, write the structure as headings, then write the full answer from the card alone. Mark yourself against a model answer.
My first timed answers come back with points missing everywhere, which feels like proof I don’t actually know the theme.
It is proof of the opposite gap. The theme is in your head. What is missing is the separate skill of choosing which parts to put down and in what order, under a clock that will not wait. That skill only shows when you write against time, which is why the first attempts are the worst. Every later one is built on a gap untimed reading never showed you.
One full timed answer a day eats an hour I could spend covering two more themes from the syllabus.
Covering a theme and being able to write it in ten minutes are different tasks, and only the second one is marked. An hour spent reading two more themes leaves two themes you can talk about and zero you can put on a page to time. The hour spent writing one answer, marked and fixed, is the only hour that moves the actual score, because the rater reads pages, never revision.
Aptitude and language: train, don’t cover
An aptitude or language test has no syllabus to cover, only a skill to raise, and a skill rises only at its edge, over months. Reading it the night before moves nothing. What moves a reading-speed score or a speaking band is practice at the type you get wrong, in the exact form the test uses, with feedback on every attempt, because speaking carries to speaking, and choosing carries to choosing, but neither carries far to the other.
That happens because the test has no topics inside it, only question types, and each type is its own small skill. Decode the types from past papers, then record your time and your error rate on each one; the frequent, weak types are where the hours go, and that is the campaign’s triage grid again, with types standing in for topics. A language test also holds by-heart material, frequency-list words, fixed phrases, spellings, and that gets the ordinary by-heart tools, not this one.
The work itself is plain and a little dull: read against the clock and check the answer, write to the word limit against the band descriptors, speak into a recorder and listen back to your own voice. Nothing in a single session feels like progress, because a skill grows slowly and steadily, not in jumps. Measure it monthly, not weekly, since a flat week is not a flat month, and the score only shows movement once enough sessions have piled behind it.
Weaker. Karthik has a reasoning test in three weeks, twenty minutes, forty questions, no syllabus to open. On Tuesday evening he sits with a passage booklet and does twelve passages back to back, untimed, checking nothing until the end. By nine he has done twelve more than Monday. The pile of finished booklets is thick, and it feels like real work. He totals his score once, at the end of the week, gets thirty of forty, and moves to the next booklet without looking at which four kinds of question he keeps missing. Three weeks pass this way. In the test his score sits at exactly thirty.
Stronger. Same Karthik, the same three weeks, the same reasoning test ahead. Tuesday evening he does six passages, ten minutes each, clock running, checks every one against the key straight after and writes the type beside each wrong answer: syllogism, wrong three times; data arrangement, wrong once. By Thursday the log has four rows, and syllogism is the tallest. He gives Friday and Saturday to syllogism alone, ten timed passages of that one type, nothing else, checking each. The log for syllogism drops from three wrong in six to one wrong in ten. He does the same for the next-tallest row the following week, then the next. In the test his score sits at thirty-seven, most of it from the types he drilled to near zero.
- Decode the types. Go through three past papers and list every question type it uses, not the topics: reasoning types, comprehension types, the fixed phrases a language paper asks for.
- Log time and errors. For each type, time yourself on ten questions and mark every wrong one; keep one row per type, updated after every practice session, never guessed at.
- Find the weak type. Sort the rows by how often the type appears and how many you get wrong; the frequent, weak row at the top is where the next week’s hours go.
- Drill that type alone. Ten timed questions of that one type only, checked straight after each set: reading against the clock, writing to the word limit, or speaking into a recorder and listening back.
- Measure again next month. Run the same log a month later. A flat week means nothing; if the whole month is flat, check the drill is still at the edge, with feedback on each attempt.
Pick a skill test you actually face — a spoken test, a reasoning paper, anything timed. Do ten questions of one type, clock on, and check each against the answer key straight after. Count how many you got right and how long each took. That count, not the hour spent, tells you whether this type needs more of your week.
It feels like I am not covering enough. A hundred passages should count for something.
It should, and it does not, because a skill test gives no credit for coverage. A hundred passages read the same way, unchecked, drills the same type again without telling you which is weak. What counts is not how many you did but how many you got right, sorted by type. Log the type beside every wrong answer for a week, and the pile stops looking like effort and starts looking like a map.
Logging every type and timing every set takes longer than just doing more passages.
It takes two extra minutes a set, and that is what decides whether the hour moved your score. Ten passages done blind teach you nothing about which type is weak; ten logged by type tell you exactly where next week’s hours should go. The log is what turns volume into practice; without it you are reading, not training. After a month the logged pile is smaller, and the weak row has shrunk.
The interview: rehearse the conversation
An interview or viva marks your thinking as it happens, and knowing the subject is only the ticket in. What the panel marks is what it can see live: whether you understood the question, whether your reasoning holds, and whether you stay yourself under a hard follow-up. The knowledge shape is your own record: every line on your form, thesis or CV is a question, and preparation is the why behind it, held as a chunk you can open in a sentence.
That works because the room’s skill is being watched, and no amount of reading rehearses being watched. A speech prepared in advance runs past the question the panel actually asked, and the panel marks that as not listening, since it wants a reply, not a recital. So the drill has to match the arousal of the room: a mock interview with a stranger who does not know you, recorded and played back, one thing fixed each round. Knowledge gets you into the room; only a rehearsed conversation, not a rehearsed script, gets you through it.
Watching the recording back is the uncomfortable part: your own voice running past the question, filling every silence with more prepared speech, while a stranger waits for you to stop. The room does the same thing once, with marks attached. The fix feels wrong at first: answer in two or three sentences and stop; let the panel pull the rest. Stopping early feels like giving something away. It is the only way the panel gets to see you think. When you do not know, say so in the first sentence, then show how you would find out; a guess dressed as an answer is the one thing every panel is trained to catch.
Weaker. Meera has an internship interview in a week, and she has written out an answer to every question she can think of: her strengths, her final-year project, why this company. She reads each answer aloud to herself in her room, twice a day, until every line comes smoothly. On the day, the panel asks about her project. She begins her prepared answer, three sentences in, on schedule. The interviewer asks, mid-sentence, what she would do differently now. Her voice catches; the next line of the speech does not fit the question, and she keeps talking anyway, filling the silence with more of what she had written, further from the actual question with every sentence.
Stronger. Same Meera, the same interview a week out. Instead of writing full answers, she lists, for every line on her CV, the one-sentence why behind it: why this project, why this company, what she would change. Three evenings that week a senior she has never met runs a mock interview over video, records it, and asks real follow-ups. The first playback is hard to watch; she talks past two questions before catching herself. By the third mock she answers in two sentences and stops. On the day, the panel asks what she would do differently on her project. She answers in three sentences, then says she is not sure about one part and explains how she would check it. The panel moves on, satisfied.
- List your own questions. Go down every line on your form, thesis or CV and write the one question it invites, then the one-sentence why behind it, ready to open and unfold on request.
- Find a stranger. Ask someone who does not know your work to sit across from you, video on, and put ten minutes on the clock before you begin.
- Run the mock, recorded. Let them ask real follow-ups, not the questions you rehearsed; answer each in two or three sentences and stop, and record the whole conversation.
- Watch the playback. Note the one place you talked past the question or filled a silence with more speech, and pick that as the single thing to fix next round.
- Run it again. Two more rounds, one fix each, with a different stranger if you can find one, until you stop, say you do not know when it is true, and let a follow-up land.
Pick one line from your own record, a project, a subject, a job you have held, and say out loud, in one sentence, why it is there and what you would do differently now. Time yourself: if it runs past three sentences, you are reciting, not answering. Record it on your phone and listen back once.
Recording myself and watching it back is humiliating. I would rather just wing it on the day.
It is humiliating, and that is the feeling the room will produce anyway, live, with marks attached — better on a screen you control than in front of a panel. Watching yourself talk past a question shows you where the prepared speech breaks, something re-reading notes never would. Wing it on the day and the panel finds that break for you. Rehearse on a Tuesday evening; by the third recording the flinch is gone.
Three recorded mock interviews take hours I could spend reading more about the subject instead.
Reading adds to what you know; it does not touch how you sound when a follow-up lands, and the room marks the second thing. A mock interview trains the room’s own skill: being watched, answering live, staying yourself when a question does not go where you planned. Three rounds of twenty minutes cost an evening; freezing at the first real follow-up costs the place.
Knowing it is not passing it
I know the subject, so I will pass it in whatever form they ask — that is the trap. What you practised carries only to work that looks like the practice, and the form is most of what an exam looks like. The same topic is a different task: a choice under a clock, points a marker ticks against a scheme, a timed argument, a reply across a table. Each pulls the knowledge out along its own path, and only the path you have worn stays open when the exam asks for it.
That is why a student who knows the chapter cold can fail a paper that asks for it in an unfamiliar shape. A topper in one exam can start again as a beginner in the next — not because the knowledge left her, but because that path was never worn. So after “do I know it”, ask “have I done it in this form, at this speed”. Answer with a count of timed, marked attempts, never a feeling. Zero attempts means the knowledge is there and the path is not. Decode the shape, then wear the path.
Pick a topic you are sure you know cold. Set a timer for the length one question of your next exam gets, and produce the answer in that exact form, with nothing open in front of you. Count what came out against what you expected. A recent count of zero attempts in that form is your real answer.
Training in the exam’s own form feels like exam coaching, not real learning — like I am trading the subject for marks.
Both are real, and the paper pays for only one of them. Keeping the learning and adding the form are not opposites — you need the knowledge before there is anything to put into that form. Keep the learning; add the form. Twenty timed attempts in the exam’s own shape is the difference between knowing a topic and scoring it.
Six shapes, one method
Every exam you will ever sit is one of six shapes, or a mix of two — the six on this page, each with its own one-line rule. Underneath all six sits one method: retrieve what you know, space the sessions, and practise in the exam’s own form. Check your answer against something outside your head, and rehearse the room when there is a room to rehearse. The exam chapters ahead say which shape, or mix, each exam is.
The facts that change year to year — the pattern, the marks, the penalty — live in the exam body’s current notice, never in this book. A marking scheme moves; a shape does not. Decode the shape first, from the notice and one past paper — the method to run on it is already yours. Decode the shape, then run the method.