A student I know spent three months on past MAT papers before realising something uncomfortable: his answers were often right, but he could not explain, in writing, why they were right. On the multiple-choice section he was scoring well. On the long questions he was losing most of the marks — not for wrong answers, but for unsupported ones. That gap is the whole exam. And it is the gap most preparation ignores.
The MAT is the admissions test used for Oxford Mathematics and Computer Science, and also taken by applicants to Imperial and Warwick mathematics courses. It is 2.5 hours, no calculator, no formula sheet. It is not a test of how much maths you know. It is a test of whether you can think with the maths you know, under time pressure, and show your reasoning to a stranger reading it cold.
Why the exam is structured the way it is
Two sections, two different jobs. Question 1 is ten multiple-choice problems worth 40 marks total, compulsory, covering polynomials, algebra, calculus, geometry, sequences and trigonometry. Questions 2–7 are open-response, 15 marks each, and you choose four of six. That choice is not a courtesy — it is part of the test. In 2.5 hours, you have roughly 150 minutes for 100 marks. Question 1 alone, at 40 marks, deserves around 50–55 minutes. Four long questions at 15 marks each leave you about 22–24 minutes per question. Anyone who tries all six long questions has already lost.
The design intent is clear: the multiple-choice section rewards fluency and speed; the long section rewards depth, proof and construction. The people who fail are rarely weak at maths. They are strong at maths and weak at deciding — which questions to attempt, when to abandon a promising line, how much working to write down.
A worked problem, and what breaks when you rush it
Take a typical long-question shape: prove that for any positive integer n, a certain expression is divisible by 3, then find all n where it is divisible by 9.
A careless candidate substitutes n = 1, 2, 3, sees divisibility, and writes "it holds for all n." Zero or near-zero marks. Three examples prove nothing.
A competent candidate does this: rewrite the expression in a form where the structure is visible — say, factor out the part that is obviously a multiple of 3 — then show that the remaining factor is always an integer. That is a two-line argument, but it requires you to choose the rewrite before you start. The rewrite is the work; the algebra afterwards is mechanical.
The second part, divisibility by 9, is where the marks hide. You need the exact condition — often something like n being a multiple of 3, or n congruent to a particular residue modulo 3 — and you must state it as a set, not as "some n." Examiners award marks for the precise answer, and a vague answer loses them even when the thinking is right.
Where preparation actually breaks
Failure one: practising answers, not reasoning. Working past papers and checking against the mark scheme feels productive. It teaches you to recognise solutions, not to produce them. The mark scheme tells you what earns marks; it does not tell you how someone found the idea. Spot this early: if you can read a solution and nod, but cannot reproduce the first two lines from a blank page a day later, your practice is not working.
Failure two: no external check. This is the structural problem with self-study for this exam. You can mark multiple-choice yourself. You cannot reliably mark your own proof — you will accept your own gaps, because they are invisible to you. A tutor or a knowledgeable reader who says "this step is asserted, not shown" is worth more than ten more past papers. By the time you notice the problem, it is often late: the exam is in weeks, not months.
Failure three: ignoring the clock. Strong candidates who never timed themselves walk into 2.5 hours of no calculator and discover that arithmetic they do fluently on paper takes four times as long under pressure. Do full papers under exam conditions at least a month out, not a week out.
Failure four: attempting everything. Six long questions in 2.5 hours is a trap. Choose your four early — after a five-minute scan — and commit. Candidates who drift between questions finish none of them fully, and partial answers on five questions score worse than complete answers on four.
Learning paths, honestly compared
| Path | What it costs | What it gives | Where it fails |
|---|---|---|---|
| Self-study with past papers | Cheapest; your time | Familiarity with question style | No feedback on proofs; you mark your own gaps as correct |
| Private tutor | Expensive, hourly | Direct feedback on reasoning | Quality varies wildly; hard to find someone who knows this exam specifically |
| Structured course | Moderate; fixed time | Sequence, timed practice, feedback loop | Only as good as the feedback on open answers |
| School extension classes | Low or free | Peers and a teacher | Rarely aimed at this test's format |
The trade-off is not money versus quality. It is feedback versus volume. More papers without feedback builds confidence, not skill. Feedback without volume leaves you slow. You need both, and the scarce resource is the feedback.
If you are also weighing the physics side of Oxford admissions, the same logic applies — see Cambridge Admissions.
Who this suits, and who it does not
This suits students who are already comfortable with the core A-level (or equivalent) syllabus and want to be pushed into creative problem solving — the kind who enjoy a problem they cannot immediately solve. It suits those applying to Oxford Mathematics or Maths & CS, or to Imperial and Warwick mathematics, and who have months, not weeks, before the test.
It does not suit students still shaky on basic algebra and calculus. The MAT assumes fluency; it does not teach it. It also does not suit anyone hoping that a course replaces the need to sit timed papers — no course can do that for you.
What a well-built course does is close the feedback gap: walk through the exact reasoning behind hard questions, show where marks are won and lost on open answers, and give you a place to put your own solutions and get a response. That is the piece self-study cannot supply. The course Cambridge Admissions: MAT (Mathematics Admissions Test) — Oxford follows the MAT syllabus with personalised lessons adapted to your level, and is built around that feedback loop rather than around watching someone else solve problems.
Start with one timed paper and an honest read of your own long answers. That tells you where you actually stand — and what to work on next.