Formula and Method Bank

Provenance: written 2026-09-28, against content/veridian/technique-hub/wma12/forge-exam-technique-guide.md (read in full — the FORGE-equivalent exam-technique layer this document sits alongside, not inside), research/veridian/WMA12-verified-facts.md (read in full, §1–§9, including the per-lesson mark-scheme-bullet audit log), and the 11 real lesson files in content/veridian/curriculum/wma12/ (read for structure, lede framing and every reference-card block, specifically to identify what is already taught in full and avoid reproducing it). Beyond those, this document adds its own primary-source spot checks run directly in this session: ~/Downloads/IAL-Mathematics-Formula-Book.pdf (Issue 2, January 2021) and ~/Downloads/international-a-level-maths-spec.pdf (Issue 3, April 2019), both converted with pdftotext -layout and read directly, not taken on the facts bank's word. That check found and corrects one real, load-bearing error the facts bank itself only hedged: §8 of the facts bank flags as "not independently confirmed" the claim that "the quadratic formula and the sine/cosine rule/area-of-triangle formulae... are on the P1 booklet page." Reading the booklet directly shows this is false as stated — the P1 page of the formula booklet prints only the cosine rule and two mensuration facts (sphere surface area, cone curved-surface area); the quadratic formula, the sine rule, and the area-of-triangle formula are printed nowhere in the booklet at all. What resolves the question instead is a different, better source found in the same session: the specification's own "Notation and formulae" section for each unit (P1.2 and P2.2), which explicitly lists, unit by unit, "Formulae that students are expected to know... and will not appear in the booklet." That P1 list is transcribed in full in Part 1 below, sourced directly, not inferred.

#WMA12 — Formula & Identity Masterlist and Worked-Method Reference

#What this document is, and isn't

This is the CODEX-equivalent content layer for WMA12: a reference for what must be known cold (Part 1) and how the paper's real mark schemes expect that knowledge to be deployed, in short (Part 2). It is not a third copy of the FORGE guide's exam-craft/rubric-mechanics material (M/A/B codes, dependency chains, timing, the abbreviation glossary — all of that lives in forge-exam-technique-guide.md and is cited here, never restated), and it is not a fourth copy of the 11 curriculum lessons' full worked chains, chain-drills, marked-solutions and MCQs (those already exist, in real depth, per topic, in content/veridian/curriculum/wma12/). What is missing from both of those — a single place that lists every formula and identity WMA12 actually requires, flags which half is printed in front of you and which half has to already be in your head, and gives short, mark-scheme-accurate method skeletons for each real question shape without re-deriving a full worked example — is what this document supplies.

Every formula below is checked against one of three primary sources: the specification (international-a-level-maths-spec.pdf, Issue 3, April 2019), the formula booklet (IAL-Mathematics-Formula-Book.pdf, Issue 2, January 2021), or the verified facts bank's own primary-source extractions from real mark schemes and examiner reports. Every method skeleton in Part 2 is grounded in a real, cited mark-scheme structure — a series, a question number, and (where the facts bank's own audit log recorded the verbatim M/A/B line) the actual mark code — not a generic "how textbooks teach it" description. Genuine gaps in the source material are flagged as gaps, not smoothed over.

---

#PART 1 — Formula & Identity Masterlist

Organized by the eight spec content areas (P2.3), in spec order. For each item: the formula/identity, whether it is GIVEN in the booklet or must be MEMORIZED, and any real domain/condition restriction.

#1. Proof (spec 1.1–1.3) — no formulae; three techniques

Proof has no algebraic formula to memorize — it's a toolkit of methods, and the spec names two of them explicitly while a third is real, confirmed by mark schemes, and not separately numbered.

This third technique is worth stating precisely because a prior pass of this course's own lesson content stated, wrongly, that "a proof of that shape is not attempted on this paper at all" — corrected in the facts bank's own audit log once real anchors (June 2019 Q3(i), October 2021 Q9) were read directly. The credited alternative route for it is genuinely proof by contradiction: a real mark scheme states outright, for a "prove $(x-4)^2 \geqslant 2x-9$ for all real $x$" question, that an assumption-and-contradiction route earns full marks as an alternative to the direct algebraic one.

No booklet content applies to Proof at all — every technique is 100% memorize-and-apply.

#2. Algebra and functions (spec 2.1) — factor theorem, remainder theorem

Neither theorem, nor the division identity behind them, is printed anywhere on the P2 booklet page. The restriction on divisor shape is spec-stated directly: "Only division by (ax + b) or (ax – b) will be required." Students are expected to be comfortable factorising cubics such as $x^3+3x^2-4$ and $6x^3+11x^2-x-6$ (the spec's own named examples) using these two theorems combined with division or inspection.

#3. Coordinate geometry of the circle (spec 3.1)

Nothing in this topic is printed in the booklet — the circle's own defining equation, the general-to-centre-radius conversion, and all three named circle properties are 100% memorize-and-apply. The spec adds one explicit two-way expectation worth stating precisely: students must be able to go from a circle's equation to its centre/radius and vice versa (construct an equation given the centre and radius, not only extract them from a given equation).

#4. Sequences and series (spec 4.1–4.5)

Sigma notation is a named, explicit spec requirement, not an optional convenience: the guidance column for 4.2 states outright, "Understanding of $\Sigma$ notation will [be] expected." This is a real content gap the facts bank's own audit found: at least one lesson in this course had zero coverage of translating $\sum_{r=1}^{\infty}$-style notation into an explicit first term and common ratio before that gap was closed — worth flagging here precisely because it is spec-mandated, not a minor omission.

4.3, increasing/decreasing/periodic sequences, is a genuine, checkable spec gap worth naming plainly: unlike every other row in the P2 unit-content table, the guidance column for 4.3 is empty — the spec supplies no worked example and no formal inequality-based definition for any of the three terms. The working definitions used in practice (not spec-quoted, since none exists to quote) are: a sequence is increasing if $u_{n+1}>u_n$ for all $n$ (or $\geqslant$, non-strictly increasing); decreasing if $u_{n+1}<u_n$ for all $n$; periodic with period $k$ if $u_{n+k}=u_n$ for all $n$, for some fixed positive integer $k$ (the smallest such $k$ is the period). No real, primary-source-verified WMA12 examiner-report error pattern for 4.1 or 4.3 specifically was found anywhere in this course's research to date — flagged honestly as a gap, not papered over with an invented one.

A genuine, booklet-verified trap sits inside the binomial row. The booklet's binomial-series page prints two formulae, not one: the combinatorial $(a+b)^n$ form above, restricted to $n\in\mathbb{N}$, and directly beneath it on the same page, the general real-index series $(1+x)^n=1+nx+\dfrac{n(n-1)}{1\times2}x^2+\cdots$, valid for $\lvert x\rvert<1$, $n\in\mathbb{R}$. The second form is P4 content — never required at P2, where $n$ is always a positive integer — but it sits on the same page students open in the exam, and real examiner reports confirm candidates reach for it anyway, on at least two independent series, with the examiners' own commentary naming it the harder, unnecessary route.

#5. Exponentials and logarithms (spec 5.1–5.3)

$y=a^x$'s graph shape must be known cold since nothing about it is printed: passes through $(0,1)$ for every valid $a$; strictly positive for all $x$ (never touches the $x$-axis); increasing if $a>1$, decreasing if $0<a<1$; asymptotic to $y=0$.

A precise, primary-source-confirmed distinction worth stating exactly: the specification's own "Notation and formulae" section for P2 — a section titled exactly that, stating outright that these "formulae... will not appear in the booklet" — explicitly re-lists only three of the five log laws (addition, subtraction, power). The reciprocal law and the identity law are still named as P2 content under spec item 5.2's own "to include" guidance, and are still not in the booklet, but they are not re-stated in that separate "must-know" callout — both are trivially derivable from the other three plus the definition of a logarithm ($\log_a(1/x)=\log_a(x^{-1})=-\log_a x$ by the power law; $\log_a a=1$ since $a^1=a$), so this is a minor precision, not a reason to treat either as optional.

#6. Trigonometry (spec 6.1–6.2)

Both identities fall straight out of a point $(\cos\theta,\sin\theta)$ on the unit circle. Neither is printed anywhere on the P2 booklet page (the P2 page carries only series, log, binomial, and trapezium-rule content — no trigonometry at all). Spec 6.2 names no new formula — it is a technique requirement (find one solution, then generate every other solution in the given interval), covered as a method skeleton in Part 2 below.

#7. Differentiation (spec 7.1)

No formula for differentiation technique is printed anywhere on the P2 booklet page — the power rule itself is P1-prerequisite content, examinable at P2 under the spec's own rule that P1's "associated formulae is assumed and may be tested," and it is not re-printed on the P2 page either (only on P1's, transcribed below).

#8. Integration (spec 8.1–8.3)

One explicit spec exclusion is worth stating precisely, since it rules out a plausible-looking question shape entirely: "$\int x\,dy$ will not be required." Every WMA12 area question integrates with respect to $x$; area calculations requiring integration with respect to $y$ are out of scope by the spec's own statement, not merely rare.

#P1-prerequisite formulae, confirmed examinable within P2

Spec P2.2 states directly that P1's specification and "associated formulae is assumed and may be tested." Reading P1's own "Notation and formulae" section directly (not inferred from what the booklet omits) gives the complete, exact list of P1-level formulae that are not printed in the booklet anywhere and must therefore already be memorized before a P2 candidate opens this paper:

One item on this list — and only one — genuinely is printed in the booklet: the cosine rule, $a^2=b^2+c^2-2bc\cos A$, on the P1 booklet page, confirmed by direct reading this session. The quadratic formula, the sine rule, the area-of-triangle formula, arc length and sector area are not in the booklet at all, at P1 or P2 — despite the quadratic formula in particular being exactly the kind of thing many students assume is "always on the formula sheet" from other exam boards' conventions. No P2 past-paper question forcing direct reliance on the sine rule, arc length, or sector-area formula was confirmed in this course's research to date (the facts bank's own §8 flags this exact check as outstanding) — the list above is sourced from the spec's own explicit statement of what P1 requires memorized, not from having found a P2 question that uses each one, so treat it as "confirmed as required background," not "confirmed as directly tested this series."

---

#PART 2 — Worked-Method Reference

Short method skeletons per real question shape, organized by the same eight spec areas. Each entry names the real mark-scheme convention it is grounded in — a series, a question number, and the mark structure where the facts bank's audit log recorded it verbatim — and points at, rather than repeats, the matching full worked-chain in the relevant curriculum lesson. None of the worked numbers below duplicate a real exam question; the method shape and mark structure are real, the numbers illustrating them (where any appear) are not.

#1. Proof

Exhaustion. (1) State the domain is finite — as scaffolding, not a scored line; real mark schemes give the first scored mark for a correct, complete list on sight, with no separate credit or requirement for narrating why the domain is bounded first. (2) List every case (this list is the B1). (3) Compute the required property (sum, product, etc.) across the cases — attempting at least two counts as the method mark (M1), even before every case is finished. (4) State the conclusion, referencing "for all the cases shown," not merely restating the claim. If the property is "a multiple of $k$" for $k>2$, the actual quotient (or an explicit equivalent, e.g. "$= 10$, a whole number") must be shown — a bare "divisible by 4" assertion with no supporting division is insufficient on a real mark scheme. Real grounding: October 2023 Q1 and October 2022 Q1, both 3-mark exhaustion questions with this B1→M1→(conclusion) structure, independently confirmed in the facts bank's proof.ts audit. Full worked exemplars: content/veridian/curriculum/wma12/proof.ts.

Counter-example. Find one failing case — more than one wastes time without extra credit. Show the calculation, then state explicitly why it fails (e.g. a factorisation showing the value is not prime). Real mark schemes bundle the example, its justification, and the conclusion into a single accuracy mark, dependent on a method mark for the attempt — a 2-mark M1(attempt)+A1(example + reason + conclusion, together) structure, confirmed on June 2022 Q3(i). Do not expect partial credit for a correct value with no stated reason.

General/direct algebraic proof (infinite domain). (1) Rearrange the inequality or identity to a "compare with zero" form. (2) Complete the square, or factorise, to express the result as a sum/product of terms whose sign is guaranteed (e.g. a squared term is always $\geqslant0$). (3) State explicitly why that guarantees the original claim — "a square number is never negative" is the kind of sentence a real mark scheme requires, not optional colour. (4) Close by referring back to the original statement. Proof by contradiction is a genuine, credited alternative: assume the negation, derive a contradiction, conclude the original claim must hold — confirmed as an accepted "Alt" route on a real mark scheme for a 3-mark inequality proof. A confirmed, common real error: treating several numerical examples of the claim holding as sufficient — it isn't, for an infinite domain, and one verified series recorded this as the single worst-performing item on its paper.

#2. Algebra and functions — factor and remainder theorems

Find an unknown from a given remainder. Substitute the divisor's root into $f(x)$, set equal to the stated remainder, solve. Real convention: one method mark covers both substituting the value and forming the equation (embedding-and-equating is one mark, not two); a second, dependent method mark, tolerant of an arithmetic slip, covers solving the resulting linear equation; the final mark is the bare numeric value only, cao. Confirmed structure (January 2024 Q1, 3 marks): M1 (embed + equate) → M1 (dependent, solve) → A1 (bare value).

Given factor, fully factorise ("hence factorise"). Divide out the already-known factor (by inspection, long division, or comparing coefficients) — this whole division is normally one method mark, not two, and no separate mark exists for merely confirming the given factor is genuinely a factor before dividing. Factorise the resulting quadratic quotient — a dependent method mark, since it depends on "their" quotient however it came out. State the final answer as a product of all factors, never a list. The word "hence" is a real, mark-determining instruction: reaching the correct final factorisation by another route (e.g. solving the cubic directly on a calculator, without ever using the given factor) scores zero on a real mark scheme for that part, even though the final answer is identical to the credited route — confirmed directly on October 2023 Q4(c)(ii), where real candidates were seen doing exactly this and losing the full three marks of that part.

Method choice. For a single "find the unknown" question, examiners' own commentary confirms the remainder-theorem substitution route is measurably faster and safer than long division — candidates who chose division more often gave up partway through. Full worked exemplars for both shapes: content/veridian/curriculum/wma12/factor-and-remainder-theorems.ts.

#3. Coordinate geometry — the circle

Centre and radius from general form. Complete the square on the $x$-terms and the $y$-terms separately. On real mark schemes, the centre is an independent mark (B1) — no method required, scorable even from an otherwise-wrong line — while the radius is a dependent mark (A1 or A1ft), since it genuinely needs the completed-square method to have been carried out. Confirmed independently on two different real anchors (January 2021 Q9(a), January 2023 Q6(a)), each with a different circle presentation (one already in centre-radius form with an unresolved parameter, one in true general form).

A parametrized circle, solved later for a specific value. If a circle's equation carries an unresolved constant across several parts, and a later part solves for the specific value of that constant, that specific value must be substituted back into the earlier general/parametrized expression for the centre or radius — not re-derived from scratch, and not left in its general symbolic form. This is a genuinely non-obvious, real, heavily mark-losing step: on one verified real series, fewer than 60% of candidates who had already found the general centre correctly went on to substitute the specific solved constant into it.

Tangent at a point on the circle. Find the gradient from the centre to the point, then take the negative reciprocal — real mark schemes bundle both of these into one method mark, not two (do not split "find the radius gradient" and "take the reciprocal" into separate scored steps, and do not take the reciprocal a second time by mistake, a confirmed real error). Form the tangent line through the point using that gradient — a further, dependent method mark. State the equation — the accuracy mark. Using the radius-⊥-tangent property this way is confirmed faster and more reliable than implicit differentiation of the circle's own equation, which very few real candidates who attempted it made good progress with, even though it is creditable.

Area of a right-angled triangle formed by a tangent/normal and the axes. When the triangle is axis-aligned or otherwise right-angled by construction, base × height is both faster and equally credited compared with $\tfrac12ab\sin C$ or a shoelace-style coordinate formula — real examiner commentary confirms candidates who reach for the more general formula "not appreciating the geometry of the situation" don't lose marks for it, but do lose time. Full worked exemplars: content/veridian/curriculum/wma12/the-circle.ts.

#4. Sequences and series

Prove the arithmetic sum formula. Write $S_n$ showing at least three terms including the first and the last, with the last term written as $a+(n-1)d$ — this is a standalone, independent mark (B1). Write the same sum reversed and add the two versions term-by-term, showing at least two matching pairs to establish the pattern — the method mark (M1); reversing without then adding does not satisfy it. Show every pair sums to the same constant $2a+(n-1)d$, reach the line $2S_n=n\{2a+(n-1)d\}$, and divide by 2 — this is confirmed as one combined accuracy mark, not two; do not split "reach the $2S_n$ line" and "divide by 2" into separate marks. Confirmed exact structure, January 2020 Q8(i): B1 → M1 → A1*, 3 marks total. Do not start the proof from a form of the result itself (a confirmed, named real error — circular reasoning), and if pairing terms within a single un-reversed sum rather than reversing-and-adding, the odd-$n$ case needs explicit handling or the accuracy mark is lost.

Prove the geometric sum formula. Write $S_n$, form $rS_n$ by multiplying every term through by $r$, subtract one from the other so all but the first and last terms cancel, factor out $S_n$ and solve. This is the standard proof examiners' own commentary says students should already know by name, not derive fresh under time pressure.

Sum to infinity. Valid only for $\lvert r\rvert<1$ — state this explicitly when it's the reason a sum exists at all. A confirmed real error: misidentifying which term is genuinely "$a$" once a series has been re-expressed or shifted from its original indexing. A calculator's own summation function, used with no formula shown, scores zero on a real mark scheme even when the numeric answer is correct — the formula must be quoted or visibly used.

Reading $\Sigma$ notation. Substitute $r=1,2,3$ directly to unpack the sum into explicit terms, reading off the first term $a$. One substituted term alone is not sufficient evidence of a genuine geometric sequence on a real mark scheme — check a second consecutive ratio, or write out at least three terms, before asserting the value of $r$.

Mixed arithmetic-then-geometric context. When a geometric series "picks up" from an arithmetic series' final term, do not restart the geometric series from the arithmetic series' own first term, and do not double-count the shared overlap term — a confirmed, specific, and non-obvious real trap, verified on a real series where it recurred across two consecutive parts of the same question. Full worked exemplars: content/veridian/curriculum/wma12/sequences-and-arithmetic-series.ts and geometric-series.ts.

#5. Binomial expansion

Coefficient of a specific term. Use the combinatorial form $(a+bx)^n$ — never the general $(1+x)^n$ series form, which is unnecessary and confirmed harder for P2's positive-integer $n$, despite sitting on the same booklet page. Identify the value of $r$ that matches the required power exactly — off-by-one or off-by-two mismatches between the term index and the target power are the dominant real error type, confirmed as more common than any conceptual misunderstanding of the expansion itself. Raise the entire bracketed term to the power $r$, including any constant multiplying $x$ — a confirmed, specific real slip writes $(4x)^7$ as $4x^7$ (raising only the $x$, not the coefficient), which changes the answer by a factor of $4^6$. After computing, explicitly isolate the single coefficient or term the question asked for — a correct value left buried inside a full expansion, a list of terms, or an un-annotated Pascal's-triangle row forfeits the final accuracy mark on a real mark scheme even though the right number is present on the page. Confirmed exact structure, January 2024 Q2 (3 marks): M1 (combines the correct binomial coefficient with both bracket terms, each correctly raised) → A1 (correct unsimplified coefficient) → A1 (final, isolated numeric coefficient) — a flat chain with no dependency, distinct from the dependent chains seen elsewhere in this paper. Full worked exemplars: content/veridian/curriculum/wma12/binomial-expansion.ts.

#6. Exponentials and logarithms

Solve $a^x=b$. Take logs of both sides (any consistent base — $\log$ or $\ln$ are both accepted). Use $\log_a(x^k)=k\log_a x$ to bring the exponent down. Solve the resulting linear equation for $x$ — stopping at an unrearranged expression is a confirmed, common real way to lose the final marks even after correctly reaching that expression. If the question specifies the final answer in a particular combined single-log form (e.g. "$x=\log_a k$, $k$ rational"), reaching a correct but un-combined exact expression (e.g. $\log_a p - q$) earns only an earlier accuracy mark on a real mark scheme, not the final one — the combining step (rewriting the constant via $\log_a a=1$, then applying the addition/subtraction law) is genuinely a separate, later accuracy requirement. Confirmed exact structure, January 2021 Q3(i) (3 marks): M1 (rearrange, take logs) → A1 (any correct exact expression for $x$) → A1 (the further, distinct mark for the fully-combined required form).

Combine log terms, then solve. Three specific, named, and mechanically distinct real misreadings recur across series: treating $\log x - \log y$ as $\log x/\log y$ (log subtraction is not division of the log expressions themselves); expanding $\log(x+y)$ as $\log x+\log y$ (a log is never distributed across a sum inside its own argument); and confusing $(\log_a y)^2$ with $\log_a(y^2)$ (squaring the whole log is not the same operation as the power law). Forming the log-free equation is its own standalone accuracy mark on every real anchor checked (four independent series: January 2021 Q3(ii), June 2023 Q5, June 2024 Q3(i), June 2025 Q6(i)) — distinct from both the combining step before it and the solving step after it. A representative real structure (June 2025 Q6(i), 5 marks): B1 (one correct log law applied) → M1 (correctly combines two of the original terms) → A1 (the correct log-free equation, standalone) → dM1 (solve the resulting quadratic, dependent) → A1 (final root only, with the invalid root explicitly rejected). Any root that would make an original log's argument $\leqslant0$ must be explicitly rejected in the final line, not silently dropped — every one of the four anchors above states this as part of the final mark's own criterion.

#7. Trigonometric equations

Mixed sin/cos/tan equation. Identify which single identity collapses the equation to one function: $\tan\theta=\sin\theta/\cos\theta$ where a lone $\tan\theta$ term sits alongside $\sin\theta$/$\cos\theta$ terms (then multiply every term through by $\cos\theta$ as its own explicit, separately-scored step — a confirmed real slip multiplies only some of the terms, not all); $\sin^2\theta+\cos^2\theta=1$ where a squared term of one function meets a linear term of the other. Collapse to a 3-term quadratic in a single function and factorise — never divide out a common factor of $\sin\theta$, $\cos\theta$, or $\tan\theta$, which silently discards an entire solution branch (a confirmed real error, structurally identical to dividing a quadratic equation through by $x$ instead of factorising it). Confirmed real structure combining both identities on one equation (June 2025 Q8(ii), 5 marks): M1 (substitute $\tan\theta=\sin\theta/\cos\theta$) → dM1 (multiply through by $\cos\theta$, on at least two of the three terms) → A1 (reach the 3-term quadratic, using $\sin^2\theta+\cos^2\theta=1$) → ddM1 (solve and reject any invalid root — doubly dependent, since it needs both earlier method steps to have produced a genuine quadratic) → A1 (final angles).

Find all solutions in a given interval. Find the calculator's principal value, then use the quadrant-sign (CAST) pattern to generate every further solution inside the stated interval — not just a second value by rote. If the argument carries a coefficient or is a compound angle (e.g. $\tan(2x+\pi/5)$), the search must run over the scaled interval — the original interval multiplied by that coefficient — before converting back to the original variable; this genuinely produces more candidate angles than an uncoefficented equation would, and any candidate that converts to a value outside the original interval must be explicitly rejected, not silently dropped. Confirmed real structure (June 2025 Q8(i), 3 marks): dM1 (correct order of operations — isolate, invert, generate enough candidates across the scaled range, even one that "does not need to be in the given range") → A1 (enforces the original interval, rejects the out-of-range value, states the exact $k\pi$ form).

Squared compound-angle equation, e.g. $\tan^2(2x+\pi/4)=k$. Take both square roots first, then run the full compound-angle, range-scaled method on each sign separately — dropping either sign is a confirmed, real, and specifically named way of losing the last two accuracy marks on this exact question shape (real examiner commentary: candidates who correctly considered $\pm\sqrt{k}$ but then failed to find all four resulting roots still lost the final mark).

Exact vs. decimal final form. If radians are demanded "in the form $k\pi$," the final answer must be left as the exact standard-angle fraction of $\pi$ — a correctly-rounded decimal does not satisfy this requirement, a distinct rule from the general "don't round early" caution, and confirmed lost on a real series where candidates worked correctly but left a decimal. Full worked exemplars: content/veridian/curriculum/wma12/trigonometric-equations.ts.

#8. Differentiation

Find and classify stationary points. Differentiate to $dy/dx$. Set $dy/dx=0$ and solve — this is a real, independent method mark, scorable even from an incorrect derivative, provided the resulting equation is then correctly solved. Keep the resulting roots exact — a fractional or surd root, correctly found by factorisation, still loses the accuracy mark if written as a rounded decimal; the method mark survives regardless, but the accuracy mark does not. Only compute the corresponding $y$-coordinates if the question actually asks for them — doing so when only the $x$-coordinates were requested is confirmed, real, unrewarded wasted effort. Differentiate again to $d^2y/dx^2$, then substitute each found $x$-value into it — this substitution is the single most commonly missed step on a real, verified anchor question, where roughly half of candidates who correctly reached the second derivative never went on to substitute a value into it. This substitution step is a real, dependent mark, gated on the specific $x$-value found earlier surviving into it. Confirmed exact structure (January 2021 Q2, 7 marks): part (a) M1(differentiate)→A1(correct derivative)→M1(independent, set$=0$, solve)→A1(exact roots); part (b) M1(differentiate again)→dM1(dependent, substitute one found value, draw a consistent conclusion)→A1(correct values and conclusions at both points).

"Prove no turning points" via the discriminant. Set $dy/dx=0$, rearrange to a 3-term quadratic in $x$, and show $b^2-4ac<0$ (or complete the square to the same effect) — then state explicitly that this means no real solutions exist, hence no stationary points. A real, confirmed dependency-scoping quirk worth knowing: on a verified anchor where the curve's constant term ($b$, in $y=ax^3-3x^2+3x+b$) is found from a separate earlier condition, the discriminant mark can survive an error in that constant (it's eliminated by differentiation before this step is even reached) but not an error in the coefficient that's still present after differentiation — the two errors are not scored the same way, because they don't have the same downstream effect on the method.

Increasing/decreasing intervals. Solve $f'(x)>0$ (increasing) or $f'(x)<0$ (decreasing) from the factorised derivative and its critical values. State the interval with a consistent inequality direction throughout — reporting the wrong side of a critical-value inequality (the "outside" range instead of the "inside," or vice versa) is a confirmed, real, and common error. Where a factor is required in integer-coefficient form (e.g. $(2x-7)$, not $(x-\tfrac72)$), the non-integer-coefficient version — though mathematically identical — is marked wrong if the question specifies integer coefficients; a confirmed, genuinely non-obvious "equivalent but marked wrong" trap.

Real-world optimisation from a stated constraint (e.g. a fixed volume constraining a surface-area formula). Use the given constraint to eliminate one variable from the quantity being optimised, differentiate the resulting single-variable expression, set the derivative to zero, and solve — then justify the nature of the resulting stationary value with the second derivative. A genuinely recurring, real error family on this exact question type, confirmed on two independent 2025 series: misreading the physical shape of the diagram (e.g. treating an explicitly open prism as closed, adding an area that shouldn't be there); a factor-of-2 slip in a cross-sectional area calculation; and, when the critical-value equation reduces to $x^3=k$ rather than a quadratic, taking a square root instead of a cube root — confirmed independently on two different series as a named, specific error, distinct from any quadratic-based stationary-point trap. A resulting negative or otherwise physically-impossible root must be rejected explicitly, not carried forward. Full worked exemplars: content/veridian/curriculum/wma12/stationary-points-and-curve-sketching.ts.

#9. Integration (spec item 8, numbered to match Part 1 above)

Definite integral by hand. Integrate each term (raise the power by one, divide by the new power), substitute the upper limit, then the lower, and subtract — the full method must be visible on the page. A calculator's own numerical-integration function returns a decimal with no method attached, and scores zero on a real mark scheme even when the value is correct, especially where an exact answer is demanded (permitted calculators are numerical-only and cannot return an exact symbolic result at all).

Area between a curve and a line, or two curves. Establish which graph is on top across the whole interval — sketch if unsure, and check specifically for a sign change partway through, which forces the integral to be split rather than evaluated in one piece. Form (upper − lower) as a single expression and integrate once, or integrate each function separately and subtract — real mark schemes credit both as valid, but combining first is confirmed, by examiners' own relative-efficiency commentary, as the faster route: subtracting after two full separate integrations needs more terms carried through the working. If a computed difference comes out negative, that's a signal to check the subtraction order before reporting a result, not simply to flip the sign.

Area against an axis boundary (not a solved intersection). When one boundary of a region is an axis (e.g. the $y$-axis, $x=0$) rather than a curve/line intersection, that limit is stated directly, with no solving required and no method mark attached to it — real mark schemes award this as its own standalone accuracy mark, separate from the other limit, which usually does need solving. Do not confuse a function's value at that axis (a $y$-value) with the $x$-limit itself — a confirmed, real, and specifically named error on a verified anchor question. Confirmed structure (January 2020 Q4, part of a 6-mark question): one standalone B1 for the solved intersection limit, a separate standalone B1 for the axis-given limit requiring no solving at all.

Trapezium rule. Identify $n$ from the number of strips (gaps between consecutive ordinates), not the number of ordinates themselves — a confirmed, recurring real error across at least four series miscounts this by one, taking $h$ from the point count instead of the interval count. Compute $h=(b-a)/n$. Sum as $\tfrac{h}{2}\times[(\text{first}+\text{last})+2\times(\text{sum of every interior ordinate})]$ — close the bracket only once every term is inside it; a confirmed real slip closes the bracket after the first addition and adds the remaining terms outside it, forfeiting the accuracy mark even though the arithmetic that follows is otherwise correct. When ordinates must be read from a labelled graph rather than a table or formula, watch specifically for zero-valued endpoints — a real, confirmed source of difficulty distinct from, and harder than, reading the same information from a table.

Justifying an over- or under-estimate. State the direction (over/under) and the mechanism (how the curve's own curvature relates to the straight chords used to approximate it) in the same sentence — real mark schemes confirm this is one combined mark, not two separately-awardable ones, on every verified instance checked. Merely restating the conclusion — "it's an underestimate because it's smaller than the true area" — does not earn the mark on a real scheme, per the FORGE guide's own paraphrase of this exact rule; the credited reasoning form instead ties the direction explicitly to the shape of the chords relative to the curve (e.g. the chords lie below a curve that is concave from above across that interval, so each trapezium's area is smaller than the true area under it). Confirmed exact single-mark structure: October 2023 Q6(c), a single B1 requiring both the label and the reason together. Full worked exemplars: content/veridian/curriculum/wma12/definite-integrals-and-areas.ts and the-trapezium-rule.ts.

---

#Flagged gaps — what this document does not claim

• 4.1 (recurrence-relation sequences) and 4.3 (increasing/decreasing/periodic sequences) have no confirmed, primary-source-verified real examiner-report error pattern anywhere in this course's research to date. The generic method (substitute successive $n$; test the sign of $u_{n+1}-u_n$ to classify monotonicity; check $u_{n+k}=u_n$ for periodicity) is standard mathematics, not something invented for this document, but it is not grounded in a real WMA12 mark scheme the way every other Part 2 entry above is. This is a real content gap in the underlying research, not something this document can close from primary sources that don't yet exist in the archive read for it.
• 5.1 ($y=a^x$ and its graph) has no confirmed real WMA12 question in the source material reviewed that directly tests sketching or interpreting this graph — the facts bank's own §6.6 material is entirely about solving $a^x=b$ equations, never about the graph itself. Flagged as a gap, not asserted as untested.
• The circle's non-degeneracy condition ($g^2+f^2-c>0$) is a genuine mathematical fact, not a confirmed exam-tested one. No real WMA12 question in the material reviewed for this document or the facts bank was confirmed to test the boundary/degenerate case directly.
• The P1-prerequisite list (sine rule, arc length, sector area, area of a triangle, quadratic formula) is sourced from the specification's own explicit statement of P1's required-but-unprinted formulae, not from a confirmed real P2 exam question using each one. The facts bank itself named this exact gap in its own §8; this document resolves what the list is, sourced directly and independently this session, but does not claim to have found a live P2 question exercising the sine rule, arc length, or sector-area formula specifically.
• Part 2's method skeletons are grounded in the specific series and questions the facts bank's own audit log already extracted in depth (the "roughly two-thirds" of examiner reports it names as closely read). Series the facts bank itself flags as only keyword-grepped, not read in full depth (October 2019, October 2021, October 2024, June 2023, June 2024, January 2025 in parts), may hold further real method-skeleton material not reflected here — this document did not re-run a fresh primary-source sweep of those series beyond what the facts bank and the FORGE guide had already independently verified.

---

#Sources used

• content/veridian/technique-hub/wma12/forge-exam-technique-guide.md — read in full; cited, not duplicated, for M/A/B mechanics and rubric conventions.
• research/veridian/WMA12-verified-facts.md — read in full (§1–§9), the primary source for spec content, booklet contents, and the per-lesson mark-scheme-bullet audit log that grounds most of Part 2's cited mark structures.
• content/veridian/curriculum/wma12/*.ts — all 11 lesson files, read for structure, lede framing, and every reference-card block, to identify existing coverage and avoid duplicating full worked chains.
• ~/Downloads/IAL-Mathematics-Formula-Book.pdf (Issue 2, January 2021) — read directly this session (pdftotext -layout), independently confirming the P2 booklet page's contents and, critically, what is and isn't on the P1 page (only the cosine rule and two mensuration facts — not the quadratic formula, sine rule, or area-of-triangle formula).
• ~/Downloads/international-a-level-maths-spec.pdf (Issue 3, April 2019) — read directly this session (pdftotext -layout), independently confirming the P2.3 unit-content table verbatim and locating the P1.2/P2.2 "Notation and formulae" sections, which give the spec's own explicit, authoritative list of formulae students must know that are not printed in the booklet — the primary source this document's P1-prerequisite table and several Part 1 "MEMORIZE" flags are built on directly, not inferred from the booklet's absences.