TMUA Common Mistakes: Where Strong Candidates Lose Marks

The TMUA differs from school maths exams in one fundamental way: it is built to be hard. UAT-UK states openly that its tests are designed to be challenging, that they need to separate highly capable applicants from one another, and that candidates should not expect to score as highly as they do in school exams. In other words, the people writing the paper know exactly where students with strong A level results go wrong, and they set questions around those pressure points.

The evidence is written into the specification itself. Section 2 of the official content specification contains two entries devoted to mistakes. Err1 requires you to identify errors in purported proofs. Err2 goes further and names two specific invalid deductions as examples: concluding b = c from ab = ac, and concluding A = B from sin A = sin B. It is rare for an admissions test to write your most likely mistakes into its own syllabus. When one does, the message is clear: these are not random slips. They are systematic, predictable and, with the right training, entirely removable.

This guide works through the highest-frequency TMUA mistakes in three groups: invalid deductions in algebra and trigonometry, direction and quantifier errors in the logic questions, and strategic misjudgements, including the habit of setting score targets from the pre-2024 scale. For each one: why it is wrong, how it appears in the exam, and the habit that prevents it.

Cancelling a from ab = ac means dividing both sides by a, and division is only legitimate when the divisor is non-zero. If a = 0, then ab = ac holds for every possible b and c, and the conclusion b = c has no support at all. That is why the official specification lists this as its first named example of an invalid deduction.

The correct move is to rearrange and factorise: ab - ac = 0, so a(b - c) = 0, which gives a = 0 or b = c. The conclusion is a genuine "or". Both branches must be kept alive unless the question explicitly rules out a = 0.

In the exam this error almost never appears in its raw form. It appears in disguise: dividing both sides by x, by (x - 1), by the difference of two expressions that might be equal, or by a logarithmic or trigonometric expression that might vanish. In spot-the-error questions, the flawed line is very often a cancellation that looks completely innocent. Purported proofs that "establish" something absurd usually smuggle the contradiction in through exactly this step: somewhere, both sides were divided by a quantity that is secretly zero.

The preventative habit is a single question, asked every time you are tempted to cancel: could this factor be zero? Better still, retrain the reflex itself. Whenever you solve equations in practice, replace cancellation with rearrangement and factorisation. The same habit protects you twice over: it stops you discarding roots when solving, and it makes the invalid step jump out at you when reading someone else's proof.

The sine function is not one-to-one: infinitely many angles share the same sine. The full meaning of sin A = sin B is that A = B + 360k degrees or A = 180 - B + 360k degrees for some integer k. Taking A = B alone keeps one branch and silently throws the rest away, which is why the specification names this deduction as its second Err2 example.

The error belongs to a whole family, and the exam draws on all of it:

What unites the family is the attempt to invert an operation that is not invertible. In count-the-solutions questions, where you must say how many solutions an equation has in a given interval, the cost is immediate: drop one branch of the general solution and your count is simply wrong, and the wrong count will usually be sitting among the answer choices waiting for you.

Two protective habits. First, before undoing any function, ask whether it is one-to-one on the domain in play. Second, flag every step where you squared or took a root, and check candidate solutions back in the original equation at the end.

Arg1 of the specification requires you to know the converse of a statement, the contrapositive, and how their truth relates to the original. The rules are short. A statement and its contrapositive always have the same truth value. A true statement tells you nothing about its converse, which may be true or false independently.

The classic losing move happens in questions that ask which conclusion must follow from a given statement. The correct answer is very often the contrapositive; the most tempting wrong answer is almost always the converse. A homely illustration: "if a number is a multiple of 4, then it is even" is true, and so is its contrapositive, "if a number is not even, then it is not a multiple of 4". The converse, "if a number is even, then it is a multiple of 4", is plainly false.

The error is rooted in everyday language, which treats "if A then B" as a loose claim that A and B travel together, so the direction dissolves under pressure. The specification says you will not be tested on symbolic notation or formal truth tables, but nothing stops you using arrows in your own rough work. Write the statement as an arrow from A to B. The contrapositive is the arrow from not-B to not-A, the same statement in different clothes. The converse is the arrow from B to A, a different statement altogether. A two-second arrow diagram catches most of this error class before it happens.

Arg1 lists four conditional forms: if A then B, A if B, A only if B, and A if and only if B. The middle two are where marks go to die.

In "A if B", B is the condition, so the direction runs from B to A. In "A only if B" the direction is the reverse: from A to B. Most candidates read "only if" as a slightly emphatic "if", flip the direction, and lose the mark without ever feeling the mistake.

The reliable way to internalise "only if" is through necessary conditions, which Arg2 tests explicitly. "A only if B" says that without B there is no A: not-B implies not-A, and taking the contrapositive gives exactly A implies B. So "only if" introduces a necessary condition, while plain "if" introduces a sufficient one. In the same vocabulary: "A is sufficient for B" means A implies B, and "A is necessary for B" means B implies A.

These phrasings appear in the paper wearing every possible costume, wrapped around statements about numbers, functions or geometry. The table above is worth memorising to the point of reflex, because under time pressure you will not have the spare capacity to re-derive it.

Arg3 defines the quantifiers, and it glosses for some as "for at least one". Two high-frequency errors live here.

First, reading "for some" as "for exactly one" or "for only some, not all". Under the official definition, a property that holds for every x certainly holds for some x; the two statements are compatible, not rivals. In true-or-false questions, deciding that a "for some" statement must be false because the property actually holds universally is a classic dropped mark.

Second, negating a quantified statement without changing the quantifier. The correct rules flip the quantifier and negate the inside together:

The signature error is negating "all swans are white" as "all swans are not white": the predicate got negated, the quantifier did not. The third row deserves its own attention: the negation of "if A then B" is "A and not B", not "if A then not B". A conditional is refuted by a single case where the hypothesis holds and the conclusion fails, which is also exactly why counterexamples work the way they do.

Arg4 makes negating all of these forms an explicit syllabus requirement in its own right. This is mechanical skill, and mechanical skill is trainable: practise negating statements of every shape until the transformations run without conscious thought.

When a Paper 2 question asks whether a statement is true or false, the standard move for "false" is to produce a counterexample. Missed counterexamples cluster in a small number of value types, and it pays to check them in order:

Before hunting, fix the domain. Does the statement range over all reals, over positive numbers, over integers? A statement can be true on one domain and false on another, and the quantifier in the question tells you exactly which objects are in play.

One related clarification, because it changes how people budget their preparation time: Prf1 of the official specification lists exactly four proof types (direct deduction, proof by cases, proof by contradiction, and disproof by counterexample). Mathematical induction is not on that list, and it does not appear anywhere in the specification or in the official Notes on Logic and Proof. Time spent drilling induction for the TMUA is better spent practising counterexample construction.

Err1 questions present a purported proof and ask you to identify what is wrong with it. Under time pressure, the commonest failure is a mode error: candidates read the proof the way they would check their own work, nodding along line by line, and every step looks roughly fine. Then they either find nothing or pick a line on gut feeling.

The productive stance is the opposite one. The question type guarantees an error exists, so read as a prosecutor, not as a student. Interrogate each line against a checklist, and notice that the checklist is essentially the table of contents of this guide:

Prf4, which asks you to rearrange a shuffled sequence of statements into a correct proof, tests the same skill from the constructive side: only a candidate who sees what each line depends on can produce the one legal ordering.

On pacing: 20 questions in 75 minutes works out at about 3 minutes 45 seconds per question, an arithmetic guide rather than an official rule. Error-spotting questions tend to run slower than computational ones, so do not let them all pile up in the final ten minutes.

The TMUA reports a single overall score from 1.0 to 9.0, to one decimal place. Many candidates inherit their target from forum posts and older students, and here a fatal time lag creeps in: the scale changed in 2024. Durham's admissions pages state that a 6.5 from 2023 equates to a 5.0 from 2024 onwards, and that an old 4.5 corresponds to roughly 3.2 to 3.5 on the current scale. Carry a pre-2024 "6.5" into the current scale as your target and you have silently raised the bar on yourself by a wide margin. Mock scores then look like failure, panic follows, and technique degrades in the real sitting. It is a purely self-inflicted wound.

On the current scale, the official anchors come from the Explanation of Results that UAT-UK publishes for each sitting: typical candidates score around 4.5, and roughly 10 per cent score above 7.0. UAT-UK publishes no grade boundaries and no pass mark. The university-published data points that do exist are three: Durham (a TMUA score of 5.0 or above brings automatic eligibility to be considered for a reduced offer), Warwick Maths (the majority of last cycle's offers went to applicants on 5.0 and above), and Imperial, which published department-level figures: among 2025-entry offer holders in Mathematics the mean TMUA was 7.6, and 8.4 for overseas applicants. Imperial itself stresses that this is historic data and not indicative of future cycles. None of these numbers contradict each other; they describe institutions with very different levels of competition. Whenever you meet a target score, ask three questions: which year, which scale, and who published it.

One final strategic error belongs here: treating October and January as two attempts. UAT-UK's rule is that you may sit only once within an admissions cycle, and neither sitting carries any advantage. There is no "try in October, improve in January".

The official preparation advice is three items long: read the specification, read the Notes on Logic and Proof, and practise papers under timed conditions. To turn this guide's error list into a routine, four habits are enough.

FrontierVue's practice and mock modules provide a timed, on-screen environment with review by error category, which pairs directly with the ledger above. For the wider picture of the test, see our TMUA overview page and the rest of this guide series.