Technique Masterclasses

Pure Mathematics 3

A Star Gap Analysis

Provenance (2026-09-29). This is a synthesis/index document, not new primary-source research — no raw Pearson PDF was opened this session. Read in full this session: forge-level-threshold-guide.md (source of every boundary in Part 1 and the cross-series point in Part 1.6); forge-error-pattern-tracker.md (source of the confirmed-once-vs-confirmed-across-N-series framing used throughout, and of Part 1's evidence counts); forge-exam-technique-guide.md (§2.6, §3, §4's self-mark checklist, and §5.2's six single-series patterns, all reused in Parts 2–3 below); and forge-battle-card.md's "Top-5 danger warnings" section, used in Part 4 in place of a "danger-phrases" section — forge-exam-technique-guide.md has no section under that name (checked directly this session: neither the word "danger" nor "phrase" appears anywhere in that file outside a lesson-file's own trap-id text), and this substitution is stated here rather than made silently. signal-frequency-and-boundaries.md §2 supplies Part 0's grade-boundary figures. research/veridian/WMA13-verified-facts.md was not re-opened directly this session; every citation below is reused, with its original attribution preserved, from one of the five technique-hub documents just named. content/veridian/curriculum/wma13/ and the rest of content/veridian/technique-hub/wma13/ were checked as directory listings, both to confirm this document doesn't duplicate existing content and because a document of this exact name and shape already exists for five other papers in this course (wbs11, wbs12, wbs13, wbs14, wec14, wma12) — none of them cover WMA13. No new diagnostic category is introduced anywhere below; every category already exists, cited, in one of the five sources. Not affiliated with or endorsed by Pearson Edexcel.

WMA13 (Pure Mathematics 3) — A* Gap Analysis: Self-Diagnostic

What this document is, and isn't

This is not new research. Every document that measures why WMA13 marks are lost — the level-threshold guide, the error-pattern tracker, the exam-technique guide's own checklists and single-series findings, the battle card's danger warnings — already exists in this folder. What doesn't exist yet is one place that answers the specific question a candidate already scoring a comfortable pass actually asks: of everything already documented, which parts are about the gap between a good pass and (near-)full marks specifically, as opposed to the gap between failing and passing at all? This document answers that by re-reading the four source documents with that one filter and indexing what survives it. It invents nothing: every citation below already appears, in the same words, in one of the four sources.

Part 0 — Unlike some sibling units, WMA13's own A* is real and checkable

This matters to say first because it isn't true of every paper in this course: a sibling AS-level unit's own gap-analysis document had to open by warning that "A*" doesn't exist at that unit's level at all. WMA13 is different. signal-frequency-and-boundaries.md §2 has independently fetched and read Pearson's own published grade-boundary PDFs for 13 real series and confirms WMA13 carries its own per-unit A* row (out of 75 raw marks, on Pearson's fixed 90-UMS-equivalent scale). Across those 13 series the A* boundary has run from 58 raw marks (January 2023, the most generously graded series in the set) to 68 (June 2023, January 2024, and June 2024) — a 10-mark swing on a 75-mark paper. That source also flags, honestly, one open detail this document inherits rather than resolves: grade-boundary documents from June 2025 onward print a second, lower-boundaried "Unit 3A" row alongside the "Unit 3" row used for the figures above, and what distinguishes the two was not established. This document adds no new boundary figure of its own — the point of repeating these two facts here is only that, for this specific paper, "how far above a good pass is the top grade" is a real, source-backed question with a real, if session-to-session variable, answer — which is not something every paper in this course's hub can say.

Part 1 — The five real mark-boundary types, read as self-diagnostic questions

forge-level-threshold-guide.md organises this paper's evidence by boundary type rather than spec topic — a cut no other WMA13 document takes, and exactly the cut a gap-diagnostic needs, because a script that is "mostly right" separates from one that is "fully right" at one of these five places, regardless of which topic it happens to sit in.

1.1 — M0→M1: is the method line written in a form the scheme can recognise, not just correctly arrived at? Oct 2020 Q1's real boundary: writing 2(2cos⁡2x−1)=7cos⁡x2(2\cos^2x-1)=7\cos x with the bracket intact earns the M1 outright; the same substitution with the bracket dropped, 2×2cos⁡2x−1=7cos⁡x2\times2\cos^2x-1=7\cos x, scores M0, because the line as written is a different, wrong equation, not because the candidate didn't know the identity.

1.2 — A0→A1: is the number attached to the one extra condition this specific mark's wording demands? Three real, distinct conditions, none reducible to "is the number right": completeness, not correctness (Jan 2022 Q2(b) — both genuine solutions found and stated, where the mark is explicitly "ONLY" the smallest one, scores less than finding and rejecting the extra); the right variable, not just the right number (Jan 2023 Q1(a) — a range stated as "x⩽9x\leqslant9" is refused outright, however correct the 9); and the right form, not a numerically equal one (Oct 2020 Q7(a) — α=\alpha= awrt 75.96°75.96° is A0 where the question fixed radians and 1.3261.326 was required).

1.3 — B0→B1: independent of method, but not independent of precision. Two shapes: a bare identity statement earns the mark until a coefficient is folded into it wrongly (Jan 2022 Q2(a) — cosec θ=1/sin⁡θ\text{cosec}\,\theta=1/\sin\theta is B1; 3 cosec θ=1/(3sin⁡θ)3\,\text{cosec}\,\theta=1/(3\sin\theta) is B0); and a single B-mark can bundle several required components with none individually sufficient (the numerical-methods sign-change mark needs both function values, a reason naming continuity, and a stated conclusion — all three, together, every time).

1.4 — The printed-answer boundary: on a "show that," the destination was never in doubt, so only a complete path is graded. Jan 2025 Q4(a) — finding AA, BB, CC correctly and simply asserting "D=0D=0" (true, and consistent with the working so far) earns nothing on that specific mark, because the scheme requires the zero to be derived, not stated. Oct 2020 Q5(a) — a triple-angle proof otherwise fully correct loses its final A1* for writing "sin" where the line requires "sin x," a cso mark tolerating no lapse anywhere in that part.

1.5 — Dependent-chain boundaries: counted per question, never assumed from topic. A step is dM1 if exactly one independent method mark precedes it in that specific question, ddM1 if exactly two do. The one documented failure mode worth self-checking against: grafting a real ddM1 quotation from one question onto a simpler worked example whose own scheme would only ever ask for a dM1 overstates what that example actually requires.

1.6 — The one comparison a topic-organised document can't show: the same slip, two different thresholds. Jan 2022 Q8(c) and Jan 2024 Q5(c) share an identical error — dropping the sign the chain rule attaches to a negative exponent, differentiating a−ta^{-t}-shaped terms. In Jan 2022 the drop costs only the final accuracy mark (a B1 M1 A0 outcome); in Jan 2024 the identical drop is folded into the method mark's own definition, and scores M0, taking every mark chained to it down too. (Jan 2025 Q5(c) supplies a third, weaker data point in the same family, explicitly flagged by the source as a single less-attested instance, not a confirmed third threshold.) The self-diagnostic this produces: knowing a trap's general shape is not the same as knowing its cost on a specific question — that has to be checked against that question's own scheme, not recalled as a fixed fact about the mistake.

Part 2 — The invisible-method trap: this paper's own "correct but scores almost nothing"

The single mechanism in this hub most likely to catch a candidate who is already strong, because it specifically punishes the candidates confident enough to skip writing down a step they can already see the answer to. forge-exam-technique-guide.md §3 gives the exact cost, not a vague one: on Oct 2024 Q8(c), a candidate who reaches the correct T=T= awrt 1.1581.158 with no algebra shown scores 1,1,0,0,0 — the differentiation itself (visible, checkable) earns its two marks, but every mark attached to solving the resulting equation is zero, regardless of how correct 1.158 is. That is 2 of 5 marks for a numerically perfect final answer. forge-error-pattern-tracker.md's T-GEN-2 confirms this is not a one-off: the identical mechanism is independently documented across six separately audited lessons spanning differentiation, exponential modelling, log-graphs, modulus/transformations, numerical methods, and algebraic division. The reasoning worth internalising: a weaker script produces the written method by necessity, because it needs the steps to reach the answer at all. A script strong enough to see the destination immediately is exactly the one that can skip writing it down — which is precisely what this rule catches. The gap-closing behaviour here is not "get better at the maths"; the maths was already there. It is the discipline of writing the line down specifically when it feels redundant.

Part 3 — Six sharp, single-series patterns: real, but not yet a confirmed recurring trap

forge-exam-technique-guide.md §5.2 names six patterns, each independently sourced but each confirmed in only one series so far — reject-the-extra-solution when a question names a specific one (Jan 2022 Q2(b)); forgetting to halve back to θ\theta from a solved 2θ2\theta (same question); a range stated in xx-notation instead of yy/f(x)f(x)-notation (Jan 2023 Q1(a)); on dy/dx=1/(dx/dy)dy/dx=1/(dx/dy), setting the reciprocal equal to zero instead of its denominator (Oct 2024, scored 0 outright despite landing on the right number for that curve); the power rule misapplied to a variable exponent instead of ax→axln⁡aa^x\to a^x\ln a (Oct 2024 Q4(c)); and terminating an algebraic division one step early, leaving a linear remainder against a quadratic divisor (Oct 2024 Q9(c)). forge-error-pattern-tracker.md's own Part 3 guidance is explicit about how to weigh single-series findings like these: real and verbatim-sourced, but "worth knowing exists," not something to drill as hard as a trap confirmed across four or five series. That is exactly the profile of a top-band-specific gap: these six are too rare individually for generic revision to surface them, but each is a documented, real zero on a script that is otherwise flawless — which is a different kind of risk from a trap common enough that most preparation already covers it.

Part 4 — What the battle card's danger warnings say about the pass/top-grade split

forge-battle-card.md's "Top-5 danger warnings" is itself already a condensed cross-reference of the level-threshold guide, the tracker, and the exam-technique guide — using it here adds no new source of claims, only a filter applied to what it already ranks. Read against forge-error-pattern-tracker.md's own confirmed-series counts, the five warnings split unevenly: #1 (continuity not named, T-NUM-1-equivalent), #2 (domain of f−1f^{-1} omitted, T-ALG-2), #3 (units/context word dropped, T-EXP-1), and #5 (angle-unit or missing/extra-solution errors, T-TRIG-7) are each confirmed across four or more series — common enough that they cost marks on the way to a pass at all, not narrowly a top-grade problem. #4 — calculator-only working on a flagged question — is the one warning that maps directly onto Part 2's mechanism above: the reason it specifically separates a strong script from a top one is that it disproportionately catches the candidate confident enough not to need the written method, which is a different profile of risk from the other four, which catch weak and strong candidates roughly alike.

Part 5 — Using this as a diagnostic, not just a reading list

This document adds no new tracking mechanism of its own — forge-error-pattern-tracker.md already has the fillable log (its Part 2 table) and the tallying method (its Part 3). What this document adds is a filter to apply before writing "silly mistake" in that log: for a mark lost on a question you were otherwise getting right, check it against Part 1's five boundary types and Part 3's six single-series patterns first. A mark lost to the printed-answer boundary (Part 1.4) or the invisible-method trap (Part 2) is not the same category of problem as a mark lost to not knowing a method at all, and the fix is different — a habit (derive the printed target explicitly; write the step down even when it feels redundant), not more practice of mathematics that, by definition, was already correct.

Flagged gaps — what this document does not claim

Sources used