HL Only · Paper 3 Deep Dive

When does Paper 3 push you beyond your comfort zone?

Paper 3 doesn't test extra content. It tests whether you can pick up a brand-new mathematical object mid-exam, read its definition properly, and reach for the tools you already own. Here's how to train that skill — and how to teach it.

For students → For teachers → Case study: Information theory → Get practice questions →

Part 1 · For students

What Paper 3 actually rewards

Paper 3 is Higher Level only. It's two long, open-ended problems (55 marks in an hour, 30% of your grade for the paper — but only about 20% of your final HL result). The paper introduces mathematical objects you have never seen before: an unfamiliar function, a novel probability setup, a strange geometric arrangement. Your job is not to remember something extra. Your job is to re-derive what you need from definitions, using the tools already sitting in the syllabus.

The one sentence to remember

Unfamiliar theory ≠ unavailable mathematics. If Paper 3 introduces a concept, everything you need to solve it is already in the syllabus. You just haven't seen it packaged this way.

The six moves that unlock any Paper 3 question

When a new object appears, work through it in this order — every time.

Identify the object

Is it a function? A random variable? A set? A geometric shape? Underline the definition in the exam paper.

State the domain

What values are allowed? Where does the object make sense — and where does it break?

Translate to old tools

Is it a log? A polynomial? A ratio? What's the closest thing from the syllabus you already know how to differentiate, integrate, or graph?

Interpret

Sketch it. Where is it positive, negative, zero, undefined? What does it mean in the problem's context?

Generalise

The exam usually gives you two or three concrete cases before asking a general result. Prove or extend using the pattern.

Optimise / conclude

Find the maximum, minimum, root, or key inequality using the derivatives and limits you already know.

Common mistakes that lose marks

Freezing on the definition. Read it twice, slowly. Substitute a small value ($p=1/2$, $x=2$, etc.) to check you've understood it.
Trying to remember "the theorem". There isn't one. Paper 3 questions are constructed so no memorised theorem cracks them.
Skipping the sketch. A graph of the new object almost always exposes the key feature (a peak, an asymptote, a zero).
Not linking parts. Every sub-part gives you an intermediate result the next sub-part uses. Track what you've proved.

How to actually revise for it

Do it slowly. Spend 45 minutes reading a single unfamiliar-context question and going through the six moves above. Don't rush.
Redo it fast. A week later, redo the same question under timed conditions. See how much of the framework you internalised.
Do the past papers, then rewrite the setup. Change one number, one function, one condition. Solve your own variant. This is what makes the six-move method automatic.
Focus on connections, not chapters. Calculus meets probability. Vectors meet trigonometry. Log/exp meets sequences. Paper 3 lives at the joins.

Worked example · Information theory · HL AA and HL AI

A concrete Paper 3 walkthrough

Suppose the paper opens by defining, for a probability $p \in (0,1]$, the information in an event as

$$I(p) = -\ln p.$$

You've never seen this. Now use the six moves:

Moves 1–2: Identify + domain

$I$ is a function of one variable, $p$. Domain: $0 < p \le 1$. Range: $[0, \infty)$ because $-\ln p \ge 0$ on that domain.

Move 3: Translate

You already know the log function. So $I$ is a familiar log flipped in sign. Everything you know about $\ln$ transfers.

Move 4: Interpret + sketch

$I(1) = 0$: certain events carry zero information. As $p \to 0^{+}$, $I(p) \to \infty$: rare events carry huge information. That's Shannon's entire idea — reconstructed from scratch in two lines.

Move 5: Generalise

For a random variable with outcomes $x_1, \dots, x_n$ and probabilities $p_1, \dots, p_n$, expected information (entropy) is

$$H = \sum_{i=1}^{n} p_i \cdot I(p_i) = -\sum_{i=1}^{n} p_i \ln p_i.$$

You built this from the sigma notation and expectation you already know.

Move 6: Optimise

For a biased coin with $P(\text{heads}) = p$, maximise $H(p) = -p\ln p - (1-p)\ln(1-p)$. Differentiate and set to zero: $\ln\frac{1-p}{p} = 0 \Rightarrow p = 1/2$. The fair coin carries the most information — a clean HL calculus result, but derived from a definition the exam gave you.

Everything used: log, chain rule, sigma notation, first derivative test. All syllabus. Nothing memorised beyond it.

Part 2 · For teachers

Teaching mathematical mobility

The natural teaching instinct with Paper 3 is either to (a) teach random advanced theories as "extra chapters" (differential geometry, coding theory, chaos…) or (b) grind through more past papers. Both miss the point. The IB is not testing knowledge of Shannon or Ramsey. It's testing transfer: can a student read a new mathematical object and mobilise old tools onto it?

The two errors to avoid

Error 1 — Chapter creep: Adding "information theory" or "graph colourings" to your syllabus as new content. Students then walk into Paper 3 expecting that specific topic and freeze when it's something else.
Error 2 — Skill drift: Teaching each syllabus topic in isolation and never returning to it in a mixed context. Students can differentiate a log but can't spot when a Paper 3 setup wants them to.

The pedagogical move that works

Treat the syllabus as a graph, not a list. Nodes are topics. Edges are the moves between them:

"When you see a rate, you're one step from calculus."
"When you see a sum indexed by outcomes, you're one step from expectation."
"When you see a definition involving $\ln$, sketch it and ask about monotonicity."

Every lesson should end with one edge — a link between what you just taught and something they saw two chapters ago.

A two-lesson training plan

Lesson 1 — Rebuild an unknown object (60 min)

Give students the definition of an unfamiliar function (e.g. $I(p) = -\ln p$, or $L(x) = \int_1^x \frac{dt}{t}$ if they've never seen $\ln$ defined this way). Walk them through the six moves. No formulas provided. They finish with a full analysis: domain, range, monotonicity, extremes, interpretation.

Lesson 2 — Chain to old tools (60 min)

Give the same class a Paper 3 past question. Ask each student to annotate the paper with one letter per line: D (domain), T (translate), S (sketch), G (generalise), O (optimise). Only then attempt the algebra. This forces the framework to become automatic.

Assessment shift

Replace some of your standard topic tests with context tests: 20 minutes on a novel definition, with the six-move framework as the marking rubric. Reward students who correctly stage the problem, even if their final algebra is imperfect. Paper 3 marks are heavily weighted on method.

References & further reading

Çeltik, T. (2025). When Does Paper 3 Push Us Beyond Our Comfort Zone? — the source article this page is built on.
International Baccalaureate Organization. Mathematics: Analysis and Approaches and Mathematics: Applications and Interpretation subject briefs (current edition).
Shannon, C. E. (1948). A Mathematical Theory of Communication. Bell System Technical Journal, 27(3–4). — origin of the information function used in the case study.

The framework, case study and pedagogical structure summarised here are drawn from Çeltik (2025). Any implementation errors or classroom simplifications are our own.

Practise Paper 3 style with our engines

Every HL AA and HL AI unit engine on this site tags Paper 3-style questions. Filter by "Paper 3" difficulty tier when you generate a worksheet.

HL AA course → HL AI course → Custom Paper 3 set — €5 →