How many past papers should I do before an exam?

Fewer than you think, reviewed far harder than you think. Eight to ten papers fully reviewed beats twenty papers skimmed. A paper you mark honestly and diagnose takes roughly twice as long as sitting it, so budget three hours of review for every two-hour paper. If you are running out of papers before you run out of weeks, generate fresh exam-style questions on your weakest topics instead of re-doing half-memorised papers.

Should past papers be timed or untimed?

Untimed early, timed late. More than six weeks out, an untimed paper tells you what you do not know, which is the useful information at that stage. Inside six weeks, sit them under full exam conditions — timing, no notes, no phone, one sitting — because by then the failure mode is pacing and pressure, not knowledge.

How do I mark my own past paper honestly?

Mark against the mark scheme line by line, award marks only for what is written on the page, and never give yourself credit for 'I knew that really'. The three rules that fix most self-marking: mark in a different colour pen, mark the whole paper before reading any solutions rather than question by question, and write the reason for every lost mark next to it — not just the correct answer.

What should I do after marking a past paper?

Diagnose each lost mark into a cause — did not know it, knew it but misread the question, knew it but made an arithmetic slip, ran out of time — and log it by topic. The score itself is nearly worthless information. The list of causes is the whole point of doing the paper, and it is what you revise from for the next fortnight.

When should I start doing past papers?

As soon as you have covered a topic, do the questions on that topic from old papers — you do not have to wait until the course is finished to use past papers. Full papers under timed conditions make sense from about eight weeks out, once most of the syllabus is in place.

How to use past papers effectively (most students do it wrong)

The score is the least useful thing a past paper produces

You sat a paper and got 62%. What do you do with that? Nothing, honestly. It tells you roughly where you are, which you already suspected, and it tells you nothing at all about what to open tomorrow morning.

The valuable output of a past paper is not the score. It is the list of reasons you did not get the other 38% — and that list only exists if you go and build it. Sitting the paper generates the raw material. Reviewing the paper is the actual study. Almost everyone does the first part, treats the second as admin, and then wonders why paper eleven scores the same as paper four.

The rule of thumb. If you are not spending at least as long reviewing a paper as you spent sitting it, you are collecting scores, not learning.

How many past papers should you actually do?

Fewer, reviewed harder. Eight to ten papers taken apart properly will beat twenty skimmed, every time, and the students who do twenty are usually doing it because reviewing is unpleasant and starting a fresh paper feels like progress.

Budget roughly two-to-one. A two-hour paper wants two to three hours of marking and diagnosis, ideally not immediately after — sit the paper in the morning, review in the afternoon, so your marking eye is not still full of the answers you convinced yourself were right.

The other reason to slow down: past papers are a finite resource and once you have seen a question you can never un-see it. Burning through the entire archive in September leaves you with nothing but half-remembered papers for the run-in, where redoing them tests your memory of last month rather than your subject.

Timed or untimed? Depends how far out you are

Distance from exam	How to sit it	What you are testing
8+ weeks	Untimed, open syllabus, one topic at a time	Whether you know the content at all
4–8 weeks	Full paper, timed loosely, no notes	Whether you can retrieve it cold, unprompted
Under 4 weeks	Full exam conditions — timing, silence, one sitting, no phone	Pacing, stamina, decision-making under pressure

The common mistake is going to strict exam conditions too early. Eight weeks out, a timed paper mostly produces a demoralising score plus a lot of blank space you would have filled if given ten more minutes — and blank space is bad diagnostic data. You want to know whether you can do a question before you find out whether you can do it in four minutes.

The opposite mistake is more expensive: never sitting one under real conditions. Every year people who can do every question in the paper lose fifteen marks to timing, because they have practised the questions and never practised the exam.

Marking your own paper without lying to yourself

Self-marking is where the method usually dies. You read your answer, read the mark scheme, recognise the idea, and award the mark. Recognition feels identical to knowledge from the inside, which is the same illusion that makes re-reading notes feel productive. Three rules kill most of it:

Mark in a different colour. Trivial, and it works — it makes it visually obvious what you wrote in the exam versus what you added afterwards while "correcting".
Mark only what is on the page. Not what you meant, not what you would have written with more time, not "I knew that really". An examiner cannot read your intentions and neither can you, three hours later.
Mark the whole paper before you fix anything. Correcting question 3 as you go contaminates how you mark question 7. Get the honest score first, then correct.

If the mark scheme awards a mark for something you did not write — a unit, a justification line, "hence" — that is a lost mark. This category is enormous and almost entirely recoverable, which is exactly why you have to be strict about counting it.

The post-paper review that produces marks

This is the part that pays. Every lost mark gets a cause, and the cause is not "got it wrong". Four buckets cover almost everything:

Did not know it. A genuine content gap. Goes on the revision list — this is the only bucket that more revision fixes.
Knew it, misread the question. Answered "evaluate" as "describe", missed a condition, used the wrong variable. A comprehension habit, not a knowledge gap, and it does not improve by re-reading the chapter.
Knew it, slipped. Sign error, dropped unit, transcription mistake between lines. These are not random — each person's slips repeat in a small number of predictable shapes.
Ran out of time. A pacing problem. Fixed by changing how you allocate minutes and when you abandon a question, not by knowing more.

Now count the buckets. A paper where twelve of the eighteen lost marks are slips and misreads means your content revision is basically done and you are studying the wrong thing entirely — every additional hour of notes buys you nothing, while an hour of careful working buys you a grade. Most students never find this out, because they never counted.

Write each one down as a line: the topic, what you did, and what the right move was. One sentence each. That list is an error log, and it is the single artefact worth keeping from every paper you sit.

Making the paper come back

A reviewed paper filed in a folder is a reviewed paper you will never see again. The final step is scheduling the return:

Three days later, redo the specific questions you lost marks on — cold, from blank paper, not by reading your corrections.
Two weeks later, do fresh questions on the same topics. Not the same question; you have memorised it by now. Generated exam-style questions work well here precisely because the numbers and framing change while the skill does not.
Before the exam, read only your error log. Not your notes, not the papers. The log is the concentrated version of everything the papers taught you.

Why spacing matters more than volume. Getting a question right immediately after seeing the solution proves nothing. Getting it right three days later, cold, is the only evidence that the gap actually closed.

Where the admin goes to die

Everything above is correct and almost nobody sustains it, for one reason: the review is twenty minutes of typing after three hours of exam, at 9pm, and the marginal fun of that is negative. The log gets kept for two papers and then quietly stops.

That specific failure is what errorboard exists to remove. You photograph the question and your working — phone, laptop screenshot, whatever is nearest — and it reads the page, works out what actually went wrong, categorises it, and files it into a per-topic log without you typing anything. The review quiz then brings each mistake back on a spacing schedule, so the "three days later" and "two weeks later" steps happen without you having to remember to run them.

You can also hand it a whole marked paper: it will mark the working page by page, and everything it finds goes into the same log, tagged by topic.

The short version

Do fewer papers, review them twice as hard.
Untimed while you are still learning; full conditions in the last month.
Mark strictly, in another colour, whole paper before corrections.
Give every lost mark a cause, then count the causes — that count tells you what to study.
Log it, and schedule the questions to come back cold.

The paper is not the study. The paper is where you collect the evidence. What you do with the evidence in the next hour is the entire difference between eleven papers that changed nothing and four that moved a grade.