How to get an A or A* in A-Level Maths: what actually separates the grades
An A needs roughly 65–75% overall and an A* around 75–85%, depending on the series. What separates them from lower grades is not harder maths — it is accuracy under time pressure, an applied third (statistics and mechanics) revised as seriously as pure, and a systematic error log that stops the same two or three marks leaking every paper. Students who do all three for the final term reliably move up a grade.
Ask a room of A-Level Maths students what separates an A* from a B and most will say 'harder questions'. They are wrong, and the error is expensive. The same questions are on the paper for everyone; the difference is that A* students drop fewer of the marks they already know how to earn. This is a guide to that difference — it is smaller and more fixable than people think.
What the grades actually require
Boundaries move every series, but as working targets: an A has recently needed around 65–75% of the aggregate, an A* around 75–85%. Spread over 300 marks, the difference between a mid-B and an A* is often thirty marks — that is three or four questions' worth across three papers, not a transformation in ability.
The three things that actually separate top grades
1. Accuracy under time pressure
Every examiner report says the same thing in different words: correct methods, wrong execution. Sign errors, premature rounding, a dropped 'dx'. The fix is boring and works: timed sections every week, and a check-the-last-line habit — differentiating back, substituting back, estimating whether the number is plausible.
2. The applied third, taken seriously
Statistics and mechanics are a third of the marks and a tenth of most revision timetables. This is backwards, because applied questions are more repeatable than pure — a hypothesis test and a connected-particles question follow the same skeleton every series. The A* candidates are the ones who stopped treating Paper 3 as a bonus round.
3. The error log
After every paper or section, every dropped mark gets one line: what it was and why it happened. Not 'got it wrong' — 'used P(X = k) instead of the tail' or 'forgot dy/dx upside down'. After four papers the log shows a pattern no single paper reveals, and the top grades live in fixing that pattern.
What the top-mark answers look like
A 'show that' question asks candidates to derive k = 2. A B-grade script writes two lines and asserts the answer. An A* script writes every line and notes the condition that makes it valid.
- B-grade version: 'rearranging gives k = 2' — scores the answer mark only, if that.
- A* version: each algebraic step on its own line, the division justified (the divisor is non-zero), the condition stated.
- The difference is four marks of method, and 'show that' questions appear on every paper.
- Multiply that by the three or four 'show that' and 'hence' questions per series and the grade boundary explains itself.
Top grades are won in the working, not the answer — write every line as though the examiner doubts it.
The last-term plan
- Weekly timed sections from January — 60-minute halves of real papers, marked harshly, logged.
- Every logged weakness gets fixed that week, not added to a vague list.
- Applied gets its own revision day — Paper 3 or the applied halves — every single week.
- From Easter, full papers under strict conditions, with the newest two series saved as final mocks.
- The week before each exam, the error log is the revision material — not new papers.
Mistakes that cost the grade
- Revising only what you like — the avoided topic is usually where the lost thirty marks live.
- Doing papers untimed and discovering speed was the problem in the real exam.
- Marking your own work generously — the examiner marks the page, not the intention.
- Neglecting the 'state an assumption', 'give a limitation' and 'in context' marks — free, and reliably dropped.
- Reading notes in the final week instead of working questions — recognition is not recall.
The checklist
- My last five papers were timed, no-notes and marked against the official scheme.
- My error log has a cause for every dropped mark, and I have re-tested each fix.
- Applied gets a full revision session every week — same priority as pure.
- I can write a 'show that' solution with no skipped lines.
- I know the assumption/limitation phrases by heart.
- I know roughly where the A and A* boundaries have sat, and I track my own aggregate.
The error log only works if every entry gets fixed — and 'read the topic again' is not a fix. When a step keeps not landing, that is the moment for an explanation that adapts: Lumi works the line back to the idea you actually lost, draws it, and sets a similar question to confirm the mark is banked.
What to take from this
- The A/A* gap is mostly accuracy, not knowledge — the content is already learned; the errors are not.
- Statistics and mechanics decide the top grades because most candidates under-revise them.
- An error log — every dropped mark, one line on why — is the single highest-leverage habit.
- 'Show that' and multi-step questions reward complete working; the top grades are won in the working, not the answer.
- Timing is a grade skill: 100 marks in 120 minutes means knowing when to move on.
Questions people also ask
Boundaries are set after each series is marked, but A* has recently required around 75–85% of the total across the three papers. An A has sat around 65–75%. Treat them as targets, not guarantees.
It is one of the A-Levels with the highest proportion of top grades — a large share of candidates get A or A*, partly because strong students self-select into it. The content is learnable; the grades go to students who are accurate under time pressure.
The gap is usually twenty to forty marks, and it lives in identifiable places: dropped negatives, skipped working, an under-revised applied third. An error log plus weekly timed sections closes it faster than any new content revision.
No — the A* questions are on the same specification; they are harder applications of the same methods, not different topics. The preparation is fluency on the spec, not content beyond it.