Technique Masterclasses

Pure Mathematics 2

A Star Gap Analysis

Provenance: written 2026-09-29. This is a synthesis/index document, not new primary-source research. Read directly this session: research/veridian/WMA12-verified-facts.md (§1, §5 — unit identity and archive scope, to ground the provenance claims below; this document does not re-mine §6–§9 independently, since the three documents below have already done that mining and this document's job is to cross-reference their output, not repeat it); content/veridian/curriculum/wma12/ read as a directory listing only (no lesson file content is reused here beyond what the technique-hub documents below already cite from it); and, in full, the following existing WMA12 technique-hub documents, which supply every specific claim below: forge-level-threshold-guide.md (the source of every threshold in Part 1), forge-error-pattern-tracker.md (the source of the confirmed-category framing in Part 1 and Part 5), forge-exam-technique-guide.md (Part 2's M/A/B system and the verbatim marking-guidance lines quoted in Part 2 below), forge-battle-card.md (its "Top-5 danger warnings," reused directly in Part 2), forge-vocabulary-precision-list.md (the source of every entry in Part 3), signal-examiner-intelligence.md (the source of every pattern in Part 4), and signal-frequency-and-boundaries.md Part 2 (the direct source of the correction in Part 0 — the only document in this hub that has actually checked what grade WMA12 carries). No independent primary-source PDF was opened this session; every mark-scheme line, examiner-report line, and grade-scale fact below is reused, with its original citation preserved, from one of these seven documents. Nothing below claims a fresh primary-source read that did not happen this session, and no new diagnostic category is introduced anywhere in this document — every category below already exists, cited, in one of the seven sources.

WMA12 — A* Gap Analysis: Self-Diagnostic

What this document is, and isn't

This is not new research. Every one of this course's WMA12 technique-hub documents that measures why marks are lost — the level-threshold guide, the error-pattern tracker, the exam-technique guide's marking-rule quotes, the vocabulary-precision list, the examiner-intelligence patterns — already exists. What doesn't exist yet is a single place that answers the specific question a candidate who is already scoring a solid pass actually asks: of everything documented across those five files, which parts are actually about the gap between a good, comfortable pass and (near-)full marks, as opposed to the gap between failing and passing at all? This document answers that by re-reading each of the five with that one filter, and indexing what survives it. It invents nothing new: every citation below already appears, in the same words, in one of the source documents named in the provenance note.

Part 0 — Read this before anything else: "A*" is not this paper's own grade

signal-frequency-and-boundaries.md Part 2 has already checked this directly, against seven official Pearson grade-boundary documents, and the finding matters enough to repeat here rather than let a candidate assume otherwise: WMA12 is graded as an AS-level unit, on a scale of A–E + U. There is no A* at unit level. A* exists only for the combined, six-unit International A-Level Mathematics qualification, calculated from UMS totalled across all six units — not from this single paper. So a document with "A*" in its own filename, applied to one 75-mark unit, is asking the wrong question if taken literally. What this document actually diagnoses is the real, unit-level equivalent: the gap between comfortably clearing grade A on WMA12 and dropping so few marks that this paper is not the unit costing you the qualification's A*. That distinction is not pedantry — treating a single-unit "A*" as a real target would mean chasing a number this qualification does not compute at this level, at the cost of the marks that actually do get totalled. This document does not repeat the raw-mark boundary table itself; it already exists, sourced, in signal-frequency-and-boundaries.md Part 2, and this document adds no new boundary claim of its own.

Part 1 — The real shape of the gap: six mark-scheme thresholds, as self-diagnostic questions

forge-level-threshold-guide.md documents six places where the real WMA12 mark scheme draws a threshold inside a mark ladder — not a right/wrong line, but a line between two things that can look identical on the page. These six are exactly where a script that is "mostly right" separates from one that is "fully right," because a candidate who has the mathematics correct can still sit on the wrong side of any one of them. Ask each question against your own last mock paper before assuming a dropped mark was a random slip.

1.1 — Independence: did that step need a method, or was it a "read it off the page" mark? Two real, cross-series anchors (circle centres read from an equation already in centre-radius or general form, January 2021 Q9 and January 2023 Q6; exhaustion-proof case lists, October 2023 Q1 and October 2022 Q1) confirm the same rule: a step earns B — no method owed — when it can be produced by inspection, and M only where an actual operation has to happen. A script that writes a correct value but pads it with an unnecessary derivation is not being penalised for that padding, but a script that treats a genuine B-step as needing justification is spending time it doesn't have to.

1.2 — Correct-but-zero: does your final answer route through what the question told you to use? Two real mechanisms take a fully correct final answer to zero. The "Hence" rule (October 2023 Q4(c)(ii)): reaching the right factorisation by solving the cubic directly instead of dividing out the given factor scores the method mark and both accuracy marks as zero, regardless of correctness. The no-working rule (January 2025 Q5(a)), read by forge-level-threshold-guide.md as three tiers: a method line clears it; an omitted detail "implied by further work" still clears it; no working at all scores zero, stated twice in the same mark scheme for emphasis. This is precisely the trap a strong candidate is more exposed to than a weak one — see Part 2.

1.3 — Sufficiency floors: do you know the mark scheme's own stated minimum, not just whether you finished? Three real, cited floors, each a specific number rather than a vague "sufficient working" instruction: at least 2 of 3 terms multiplied through correctly (June 2025 Q8(ii)); at least two valid combinations worked in an exhaustion list (October 2023 Q1 / October 2022 Q1); at least two pairs of matching terms shown in the arithmetic sum-formula proof (January 2020 Q8(i)). A candidate who doesn't know these floors exist either over-works past them (costing time) or stops one short (costing the mark) — and one short of a stated floor is not partial credit, it's the mark gone.

1.4 — Exact form: does this specific mark's own line carry an exactness instruction, or does it carry awrt? January 2021 Q2's A1 requires the exact fraction 8/3 or a recurring-decimal notation (2.66‾2.6\overline{6}) and explicitly rejects the truncated 2.672.67 — the boundary is whether the notation itself certifies exactness, not whether the numeral is arithmetically close. The same shape recurs on circle radii (r=20r=\sqrt{20} confirmed as a real error when given as a decimal). Contrast: January 2025 Q3's A1 accepts awrt 3.76 outright, no exact form offered at all. A top-band script tracks which rule applies mark by mark; a good-pass script applies one rule paper-wide by habit.

1.5 — Dependency location: which one specific earlier mark does this step actually depend on? January 2023 Q6(b)'s dM1 depends, by the real scheme's own wording, on the earlier M1 (the attempt) having been made — not on that attempt's accuracy mark too. A candidate who eliminates incorrectly but then correctly solves whatever wrong quadratic results still earns the dM1. A lesson-audit pass in this course's own material once got this backwards, over-specifying the dependency — worth naming because "which mark am I actually depending on" is not intuition a strong candidate can skip checking. Contrast the doubly-dependent ddM1 chain (January 2021 Q3(ii): combine logs → remove logs → solve, independently confirmed across four series as three separately-scored stages), where a slip at the very first stage removes access to everything after it — a structurally different risk from the singular-dependency case.

1.6 — Allowance: is this deviation a bounded special case, or the flat misreading deduction? January 2021 Q9's O-as-origin/O-as-centre confusion was permitted as a named special case, but the allowance reached every method mark on that part except the final numeric answer — bounded, not blanket. The general misreading rule, by contrast, is a flat two-mark deduction from A/B marks in the affected part, with no discretion, and it stops applying at all if the misreading "materially simplifies" the question. Confusing which of the two mechanisms is in play changes how many marks a single error actually costs.

Part 2 — The invisible-answer trap: a top-band-looking script can score like a blank one

This is the single most consequential mechanism for exactly the candidate this document is written for, because it disproportionately catches confident, capable scripts. Three real, independent enforcements of the same underlying rule: January 2024 Q5(i) — a calculator's summation button reaching the correct sum-to-infinity value scored nothing; January 2025 Q5(a) — the mark scheme states outright, twice, that a correct value with no working scores zero; October 2019 Q10 — a true "no solutions" conclusion, unscored, because the discriminant working was never shown, which the report itself attributes to over-reliance on calculators. forge-exam-technique-guide.md traces this to the general marking guidance's own verbatim rule: a method mark is earned from correct working with values even without the formula quoted first, but may be lost if there is any mistake in the working — and there is nothing on the M line to read at all if no working is visible. forge-battle-card.md ranks this its #1 danger warning for the whole paper for exactly this reason.

The reason this belongs in an A*-gap document specifically: a weaker candidate who needs the written method to reach the answer at all produces the working by necessity. A candidate strong enough to see the answer immediately is the one who can skip writing it down — and is therefore the one this rule actually catches. The gap-closing behaviour is not "get better at the maths," it's the discipline of writing the line down even when it feels redundant.

Part 3 — Vocabulary precision as a last-mile differentiator

forge-vocabulary-precision-list.md documents word-level distinctions a mark scheme is precise about, and a "good pass" script routinely blurs without losing the mathematical plot — which is exactly why blurring them costs marks that error in method wouldn't. Three of its sharpest entries, reused here rather than re-derived: "justify... by further calculus" vs. "deduce" — the former names the method as part of the mark (a correct conclusion reasoned from a cubic's shape alone "scores no marks," January 2021 Q2(b)) while the latter accepts a correctly-drawn conclusion from an already-established result (June 2025 Q8(a)); "factorise" vs. "fully factorise" — a cubic left as a linear factor times a still-factorisable quadratic is not fully factorised even though every shown step is correct (October 2023 Q4); and the general marking guidance's own rule that "exact", once triggered, rejects a correctly-rounded decimal outright, confirmed on the specific, verified trap of a radian answer required "in the form kπ" (June 2025 Q8(i)). A script that is mathematically sound throughout but treats these as interchangeable synonyms is losing marks a purely computational review would never catch.

Part 4 — What the examiner reports themselves say about where the gap actually sits

signal-examiner-intelligence.md finds that WMA12 reports have a consistent, quotable habit of dropping their usual hedge-word register for hard numbers on exactly one question per series — the one the report itself treats as that paper's real discriminator (January 2021 alone drops from "nearly 90% to just over 70%" between two marks on Q3(i); October 2021's Q9 states "47% scored 0/4", its worst-performing question). The practical implication: the good-pass/top-grade gap on any given series is not spread evenly across the paper — it concentrates on one identifiable question, and the report's own switch into hard numbers is the signal that it has.

The same document also finds that where a mark scheme credits two genuinely valid methods equally, the reports repeatedly volunteer, unprompted, which one real candidates handle worse — implicit differentiation on the circle "often made little progress" despite earning method marks (January 2021 Q9); long division chosen over substitution on a factor-theorem opener cost time and risk (forge-error-pattern-tracker.md 6.2, long-division-chosen-when-substitution-was-asked-for, confirmed January 2024). A script that is technically correct but consistently reaches for the harder of two equally-credited routes is not making an error the mark scheme penalises directly — it is spending marks-worth of time and risk that a script choosing the easier credited route doesn't.

Part 5 — Using this as an actual 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 tables (its Part 3 tally sheet, Part 4 detailed log). What this document adds is a filter to apply before logging a lost mark there: for any mark lost on a question you were otherwise getting right, check it against Part 1's six thresholds first, rather than defaulting to "silly mistake." A mark lost to the Hence rule or the no-working rule (Part 1.2, 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 — the fix is a habit (write the method line; check what "hence" points at), not more practice of the underlying mathematics, which by definition was already there. Use Part 3 and Part 4 the same way: before writing "careless" in the tracker's own log column, check whether the actual cause was one specific word read too loosely, or a route chosen that was valid but empirically slower — both are namable, fixable causes, not a single undifferentiated "carelessness" bucket.

Flagged gaps — what this document does not claim

Sources used