How to Mark a Maths Past Paper and Build a Mistake Log

Do not use a past-paper mark only as a score. Mark each lost mark to the first point where your reasoning stopped working, classify the cause, write a repair action, and retest that exact skill later. A useful mistake log is short enough to review and specific enough to change what you do next time.

Mark evidence, not just answers

Finish under the intended conditions: record the qualification, course or tier, paper/component, session, time limit, calculator status, and date. Before opening the mark scheme, write your raw score and flag guesses. A guessed correct answer is not secure evidence.

Then mark against the official mark scheme for that exact paper. For Cambridge IGCSE Mathematics 0580, the current syllabus describes both mathematical techniques and the analysis, interpretation and communication of mathematics as assessment objectives; its command-word guidance also explains what prompts such as calculate, show, determine and sketch expect. For IB Maths AA or AI, identify the exact course, level and paper before using the corresponding official guide and markscheme; do not import a rule from a different course or session.

For every question, record marks earned, marks available, and the first line at which your work diverged from a valid route. If later lines are wrong only because they carry forward an earlier value, log the first cause and note the dependent marks. Accept equivalent methods or answers allowed by the scheme; compare with the published standard, not a model solution's wording.

Use one colour for missing or invalid work and another for corrections; never overwrite the original attempt.
Circle command words, units, rounding instructions, diagram labels and calculator requirements that you missed.
Write a one-sentence diagnosis before reading a full solution: “I used ___ when the question required ___.”

Use four error labels, not one vague ‘careless’ label

These four labels are a study tool, not official Cambridge or IB categories. They stop “careless mistake” from hiding a fixable cause. Give each error one primary label and, if useful, one secondary label.

Knowledge (K): you cannot recall or explain a definition, identity, theorem, formula, notation or prerequisite. With the paper closed, can you state what the symbols mean and produce a simple example? If not, repair with retrieval and basic questions, not another full paper.

Method (M): the ingredients are familiar, but you chose the wrong representation or strategy—for example, expanding when factorisation is needed, or applying a trigonometric rule to the wrong triangle. Name the feature that should have triggered the better method.

Execution (E): the plan was sound but a sign, arithmetic, algebra, transcription, graph or calculator-entry error broke it. Rework from the last correct line and add a check aimed at that failure. A calculator can reproduce wrong input, so “the calculator gave…” is not a diagnosis.

Exam technique (T): the mathematics may be available, but the response missed the task: wrong command word, omitted reasoning, early rounding, missing units, misread scale, unauthorised calculator mode, time loss or wrong final form. Cambridge’s syllabus includes communication, accuracy and command-word conventions; check current official material rather than treating a private checklist as a rule.

K = cannot explain or start the underlying idea.
M = knows relevant ideas but selects the wrong route.
E = route is right; an operation, transcription or calculation breaks it.
T = route may be right; the exam response misses the instruction, format or constraint.

Build a log that forces a next action

One row should describe one repairable event, not an entire bad question. Keep the question number and enough wording to find it again. Cause and prevention are the valuable columns: they turn review into a decision about your next attempt.

Use this template in a notebook, spreadsheet or database:

Add frequency only if it changes your decision. Three identical sign errors justify an intervention; a tally of every lost mark does not. Group rows by tag and skill after marking, then choose at most two high-frequency or high-cost repairs for the next study block.

Do not log a solution you copied without first writing your own diagnosis.
Separate ‘not attempted’, ‘guessed’, ‘wrong’, and ‘correct but insecure’ where that distinction affects the repair.
Keep a ‘proof of repair’ link or paper/question reference beside each row so a claim of improvement can be checked.

Run the correction loop immediately

A mark scheme is feedback, not a substitute for solving. For each meaningful error, close the scheme and run this six-step loop:

1. Reconstruct: cover the answer and write the givens, target and first step. 2. Locate: compare with the scheme and mark the first invalid line. 3. Explain: state the cause and attach K, M, E or T. 4. Repair: write the smallest rule, micro-example or checklist that prevents this error. 5. Retrieve: solve a close variant without notes, then a less familiar one if correct. 6. Update: record the result and schedule a delayed retest.

For an execution error, do not automatically reread the whole chapter. Recalculate the same expression with an estimation check, reverse operation or substitution. For a method error, compare two candidate methods and write why one fits the structure of the question. For technique, rewrite only the final response in the required form and annotate the trigger you missed.

A correction is not complete when the answer looks familiar. You must produce the reasoning from a blank page and explain why the wrong route was tempting. Retrieval practice can improve later retention, not merely measure it; Roediger and Karpicke found the advantage most clearly on delayed rather than immediate tests.

Retest on a schedule, but do not worship the dates

Use spacing as a default, not a universal prescription. A practical starting sequence is: one close variant in the same session; a fresh question after roughly 2–3 days; a mixed question after about 7–10 days; and a final check 2–4 weeks later. Move the next attempt earlier if the skill is foundational or the first retest fails. Stretch the interval when the answer is produced fluently twice and the error does not recur in mixed practice.

These intervals are an implementation heuristic, not an official Cambridge or IB timetable or a promise that every learner needs four attempts. Reviews support distributed practice, while showing that the best gap depends on the desired retention interval, material and learner. The key contrast is spaced retrieval versus rereading a correction while it remains in working memory.

At each retest, hide the old solution, change at least one surface feature, and record a binary result plus a confidence note: solved independently, solved with a prompt, or failed. A prompt is useful evidence but not a pass. If you fail, preserve the same row, add the new first-invalid line, shorten the next gap and change the intervention. If you pass one near-identical question but fail a mixed one, the repair is not yet generalised.

Pass standard: correct method, accurate execution, and the required final form without looking.
Promotion rule: remove a skill from the urgent list only after it survives at least one delayed mixed retest.
Review the log weekly by recurrence and mark loss, not by how recently a row was written.

Know when to ask a teacher or tutor

Escalation is part of the method, not an admission of failure. Ask for help when the same tagged error appears in three independent questions, when you still cannot start after a closed-book retrieval attempt, when two official materials seem to conflict, or when your score is flat after two correction–retest cycles. Also ask sooner if time pressure, anxiety, vision, motor or reading difficulties make the pattern broader than a single topic.

Bring the original attempt, the exact paper and markscheme, the relevant log rows, your corrected solution, and the retest results. Ask a narrow question: “What is the first invalid inference here?”, “What feature should trigger this method?”, or “What response does this command word require in this course?” Ask the teacher to watch one fresh question rather than simply showing you another finished solution.

Afterwards, write the agreed rule in your own words and test it on an unseen question. If uncertainty concerns a syllabus, calculator policy, command word, permitted method or assessment requirement, verify current official Cambridge or IB documentation and your school’s instructions; do not let an old forum post become your marking standard.

Frequently asked questions

Should I record every mark I lose?

Record lost marks initially, but promote repeatable causes into active rows. A recurring sign, method or command-word failure needs a full correction loop.

What if I get the final answer right for the wrong reason?

Log it as correct but insecure, write the valid reasoning, and retest with a changed question. A lucky answer needs investigation.

How many past papers should I do before reviewing my log?

Do not wait for a fixed number. Review after each paper and choose one or two recurring repairs before the next. Full papers diagnose; targeted retests repair.

Should I use the mark scheme before attempting a question?

Not for a diagnostic attempt. Use it after recording independent work and a first diagnosis. For targeted learning, study a step if needed, then close it and retrieve the method before counting the question as repaired.

Sources and verification