STEP Probability and Statistics: How to Play Section C

Start with the official architecture. STEP 2 and STEP 3 each carry twelve questions: eight pure mathematics in Section A, two mechanics in Section B and two probability and statistics in Section C, every one of them marked out of 20. The scoring rule underneath this article needs stating precisely, because it is widely misremembered: you are not asked to pick six questions and commit. Everything you attempt is marked, and your grade is then built from the six questions you answered best, out of a maximum of 120, with no restriction on how many questions you attempt from any section.

That puts Section C in exactly the same position as mechanics: not a compulsory module but two seats in a candidate pool of twelve. The grade calculation is completely indifferent to provenance. Six pure answers, or five pure plus one probability, or four plus one plus one, all score identically if the marks are the same. For a sense of scale, in the 2026 series grade 1 on STEP 2 fell at 62 out of 120, a little over ten marks per question across six answers. Against that backdrop, the right question about Section C is never "do I need to learn statistics for STEP". It is "can these two seats supply me with an answer worth twelve or more marks".

The arithmetic first: Section A alone offers eight questions and only six count, so a pure-only plan is entirely legitimate, and plenty of candidates set out with exactly that intention. The hidden assumption is that at least six of those eight pure questions will actually be workable for you on the day. In practice one will land on your weakest topic, and another will simply refuse to open at the first part. If your workable pool is exactly six, a single collapsed question forces you into territory you never prepared, at full exam pressure.

Train one Section C question type to reliability, counting being the classic choice, and the effect is to widen your candidate pool from six to seven or eight, turning selection from forced acceptance into genuine choice. Hold the official caveat alongside this: you may attempt as many questions as you wish, but OCR is explicit that it is not advisable, and that candidates who attempt more than six rarely score as well as those who concentrate on six. So the point of widening the pool is not to answer more questions. It is to raise the quality of the six you finally submit. What you buy by practising Section C is not the marks of a seventh answer; it is the right to choose your best six.

Two hard exam-room rules tell you most of what Section C can and cannot be: calculators are neither permitted nor required, and no formulae booklet is issued (the booklet was discontinued from 2019). An examination that physically bans calculators cannot ask you to process a data set, compute sample statistics or read values from tables, which excludes the formats most familiar from A level statistics lessons. What remains is derivation: translate the wording into a probability model, derive distributions, compute expectations by argument, prove general results.

OCR's description of STEP's question character applies with full force here: applying mathematical knowledge in novel and unfamiliar ways, often drawing on several specification topics, demanding insight, ingenuity, persistence and substantial sequences of algebraic manipulation. In Section C that plays out as questions which begin as probability and turn into pure mathematics halfway through. A sum over a distribution becomes a series manipulation; a continuous model becomes an integral; a bound becomes an inequality argument. Most STEP "statistics" questions are pure questions wearing probability clothing: the first half examines whether you can build the model, the second half whether your pure engine can carry it home.

A level statistics chiefly trains you to use things: apply given formulae, run calculator procedures, follow standard methods on data. STEP Section C examines almost none of that. It examines construction: define your own events, set up the model yourself, derive the distribution and the expectation from first principles, step by step. The overlap between those two skill sets is smaller than most students expect, which is why a strong A level statistics performance predicts very little about STEP Section C.

The converse also holds. A candidate who "dislikes statistics" but has a solid pure base and enjoys counting arguments can find Section C surprisingly profitable, because what the section actually consumes is pure technique and modelling precision. The marking works in your favour here too: OCR explicitly rewards good progress towards a solution, so a correctly built model and cleanly derived early parts earn real credit even if the final part escapes you, and a correct answer receives full marks whatever the method used. The right preparation mindset is to learn Section C as a new question type from scratch, not to revise your A level statistics notes.

The cleanest way to see this is from the syllabus level. STEP 2 is based on A Level Mathematics and AS Level Further Mathematics; STEP 3 on A Level Mathematics and A Level Further Mathematics. At that level the probability toolbox available to a question setter is genuinely small, and within it, counting and conditional probability are the two areas with the lowest prerequisites and the greatest reasoning depth.

A counting problem depends on almost no theory. Everyone has "covered" it, yet it can demand real insight and strict case discipline: what exactly is being counted, is anything counted twice, is anything missed, with the whole structure built by you. Conditional probability is naturally layered: information arrives in stages and the model must update in stages, which makes it the sharpest test of whether you translated a verbal scenario into a precise mathematical object. Both match OCR's stated question character, novel contexts, cross-topic reach, insight and persistence, better than any other probability topic. Their recurrence in Section C is structural, not coincidence. Build your foundation on these two areas and practise them to depth, rather than spreading effort evenly across every probability topic.

The two papers differ first in their syllabus bases, and only after that in perceived difficulty:

For Section C, the reasonable expectation is that STEP 3 draws more freely on Further Maths level pure machinery, with heavier series work, integration and algebraic manipulation, and asks more of continuous models and general proofs, while STEP 2 stays closer to single-maths probability reasoning. Both papers share the same derivation-first character. Read the numbers carefully too: STEP 3's lower grade 1 boundary and higher pass proportion should not be read as "Section C is easier on STEP 3". OCR explains that boundaries are set so that the offers met closely match the places available at the official user institutions (Cambridge, Warwick and Imperial College London) and therefore move with each year's cohort; each paper is set its own boundaries, so comparing boundaries across papers is not a measure of difficulty. Boundaries move each year, so treat the 2026 figures as context, not as permanent thresholds.

Do not decide by taste. Walk through four questions:

For most candidates the conclusion is to position Section C as a second string: a dependable source of one or two candidate questions and insurance against a pure collapse, not a pillar co-equal with pure. The eight pure questions remain the first priority, always.

At the start, read all twelve questions, including both Section C questions, even if your plan is pure-first. Do not skip them out of section prejudice. The scoring rule means switching carries no sunk cost: everything you attempt is marked and only your six best answers count, so an abandoned attempt cannot drag your grade down; it only costs the time you spent on it.

The practical routine: after the read-through, choose six questions to attack and designate one substitute. Section C often makes a good substitute because the opening part of a counting question tends to have a friendly entry point, letting you convert quickly after a switch. When a main-line question jams, switch decisively rather than fighting it.

Keep the hard exam-room rules in mind as well. STEP issues no scratch paper: rough work goes in the answer booklet itself, and you simply cross out what you do not want. Crossed-out work is not marked unless you clearly indicate that it should be, and each page you write on must be labelled with the question number as the booklet instructs. Finally, resist the scatter strategy of attempting seven or eight questions. OCR is explicit that candidates who do so rarely score as well; the right use of three hours is to finish six questions properly.

The official requirement is blunt: solutions must be clear, logical and legible, with working fully set out, and marks may be lost if examiners cannot follow your working, even when the final answer is correct. Probability is where this rule bites hardest, because probabilistic language is naturally slippery: undefined events, an unstated meaning of "at random", a case analysis with a missing branch. Habits that survive A level marking will sever the credit chain entirely in STEP.

The discipline for writing Section C comes down to three rules. First, define your events and random variables in notation before you argue, and then argue in that notation throughout. Second, state the assumptions you are using; independence and equal likelihood are premises to be written down, not defaults to be assumed silently. Third, make counting arguments auditable: say what is being counted and how you know nothing is double-counted or missed. Two rules work in your favour meanwhile: progress towards a full solution earns marks, so set the model up in full and secure the early parts; and any correct method earns full marks, so an elegant counting argument and a laborious complete enumeration are worth the same, and you should take whichever route you can write most clearly. In STEP, clarity is itself worth marks.

Build your training on the official free resources first. The STEP Support Programme, developed by the Cambridge Faculty of Mathematics, is free and open to everyone, with online modules, hints and full solutions. Cambridge has also written fully worked solutions for STEP 2 and STEP 3 papers from 2019 onwards, and recommends using them to assess progress on timed papers once the STEP 3 modules are done.

For Section C specifically, work in themed passes rather than by year: extract the probability and statistics questions from recent past papers and work them in blocks by type, counting, conditional probability, discrete distributions, continuous distributions. Blocked practice builds pattern recognition faster than chronological grinding, and it exposes quickly where your genuine advantage lies, which is exactly the data the decision framework above asks for. One caution when digging through archives: STEP 1 has not been offered since the June 2021 sitting, so when you run into old STEP 1 papers, do not treat them as samples of the current examination.

Finish with full timed papers, and treat the decision process itself, the read-through, the selection, the substitution, as part of what you are training. Force yourself to execute the full twelve-question scan and name a substitute in every mock. In STEP, choosing questions well is worth as much as solving them well, and that skill only develops under full-paper timed conditions.