Mock Paper Protocol

Provenance: compiled 2026-09-29. Built from a full read of research/veridian/WMA12-verified-facts.md (§1–§9) and every existing WMA12 technique-hub document — forge-exam-technique-guide.md, forge-battle-card.md, signal-frequency-and-boundaries.md, codex-formula-and-method-bank.md, all ten codex-deep-dive-*.md files and all eleven signal-topic-brief-*.md files — plus a direct grep sweep of all twelve files in content/veridian/curriculum/wma12/ (the eleven real lessons plus paper-anatomy.ts itself) and content/veridian/question-bank/wma12.ts for every series name, done for this document to separate "a specific real question already shown to a student somewhere in this course" from "genuinely untouched." Two pieces of new checking specific to this document: (1) a direct directory listing of the three local past-paper locations named in the facts bank's own §5 — confirmed present, in this session's accessible directories, at ~/Documents/SOL/Papers/IAL_Paper2_Complete/IAL_Materials/Mathematics/ (June 2019 through June 2025, including the June 2024 "01R" and June 2025 "01a" regional variants) and at ~/Downloads/PAPERI/IAL_Unit2_Complete_2024_2025/IAL_Materials/Mathematics/ (the same series); (2) the facts bank's third archive, ~/Downloads/IAL_Maths_P2.zip (extracted path given as home/ubuntu/IAL_Maths_P2/ — a path shape that does not belong to this Mac), which the facts bank says holds October 2025 and January 2026 question papers and mark schemes, could not be located anywhere in the directories reachable from this session — flagged honestly in §2 rather than assumed available. No new primary-source PDF was opened for this document; every mark-scheme rule, quotation and per-question mark figure below traces to one of the files named above, cited at first use. The one genuinely new arithmetic in this document is §4's checkpoint table, which is the 72-second-per-mark rate mechanically applied to three per-question mark sequences that are already independently verified and cited in paper-anatomy.ts and forge-exam-technique-guide.md — not a new primary-source finding, a new use of an existing one. Not affiliated with or endorsed by Pearson Edexcel.

#WMA12 — Mock Paper Protocol

#Pearson Edexcel IAL Mathematics, Unit P2: Pure Mathematics 2

#0. What this is, and what it deliberately doesn't repeat

This is the operational layer that sits on top of five existing WMA12 documents, not a sixth re-teach of any of them. It answers three questions none of the other five fully answer: which real series is actually worth sitting cold, given what this course's own material has already shown a student; in what order to self-mark the result once the 90 minutes end, so the check runs fastest into this paper's own documented failure modes rather than a generic pass; and how to turn a mock's timing record into something more diagnostic than "finished" or "didn't finish."

What this document does not do: it does not re-explain why the M/A/B mark scheme works the way it does (forge-exam-technique-guide.md Part 2), it does not re-list the formula-booklet contents or the memorise-vs-look-up split (codex-formula-and-method-bank.md; forge-battle-card.md), it does not re-derive the paper's basic vitals — 90 minutes, 75 marks, 9–11 questions, no sections (paper-anatomy.ts) — and it does not re-catalogue the eleven lessons' own trap-taxonomy items or the topic-frequency/grade-boundary tables (forge-exam-technique-guide.md Part 4; signal-frequency-and-boundaries.md). Where any of those is the right place to look, this document names it and moves on.

---

#1. What a mock on this paper is actually rehearsing for

Three real, cross-cutting failure shapes already sit in this hub's own citations, spread across five different topics — which is precisely why a mock, not a single worked example, is the right way to catch them, since none of them announces itself as belonging to a specific spec point:

A correct method that never produces the answer, because a required later step was skipped. January 2021 Q2(b): a correctly-found second derivative earns nothing further until one of the two x-values from part (a) is actually substituted into it — "only 50% scored" the mark this gates (paper-anatomy.ts). June 2025 Q9(b): an answer carried forward "not in the correct form" broke the follow-through eligibility of the whole next step (forge-exam-technique-guide.md Part 4, item 6).

A correct final number with no visible method, which a mark scheme built entirely from M/A/B lines cannot credit. January 2024 Q5(i): the real examiner report describes candidates who used the correct sum-to-infinity formula, but a number who instead used their calculator's own summation function scored "no marks... even if [the value] was found" (paper-anatomy.ts) — the formula was never on the page for an M mark to reward. January 2025 Q5(a), independently confirmed: the real mark scheme states outright, "Note that awrt 18.8 with no working is 0 marks" (forge-exam-technique-guide.md Part 2). Two different series, two different topics, the identical mechanism.

"Hence," or "using your answer to (a)," ignored — either by recomputing from scratch or by reaching for an unrequested, easier-looking method. October 2023 Q4(c)(ii): a candidate who calculator-solved a cubic instead of using the given factor scored zero on that part, regardless of whether the final root was correct (forge-battle-card.md, danger #4). June 2025 Q5(b): candidates who redid the trapezium-rule calculation in part (b) instead of reusing their part (a) answer, as the question required, "gained no marks" for it — and the examiner report separately notes the habit cost time as well as marks (forge-exam-technique-guide.md Part 4, item 6).

A mock's real value is catching one of these three shapes live, on a question where the trap hasn't been named for you in advance — because in each real case above, the algebra up to the point of failure was genuinely correct. §6 below builds the self-mark order specifically to surface these three shapes early, not as an afterthought once the "did I get the right number" pass is already done.

---

#2. Choosing which real series to sit

What's actually reachable, checked directly for this document. Two local archives hold real WMA12 question-paper/mark-scheme/examiner-report material and are confirmed present in this session's own accessible directories: ~/Documents/SOL/Papers/IAL_Paper2_Complete/IAL_Materials/Mathematics/ and ~/Downloads/PAPERI/IAL_Unit2_Complete_2024_2025/IAL_Materials/Mathematics/, both holding the same span — June 2019 through June 2025, including the June 2024 "01R" and June 2025 "01a" regional variants. A third archive, listed in the facts bank's own §5 as holding October 2025 and January 2026 (~/Downloads/IAL_Maths_P2.zip, extracted to a path — home/ubuntu/IAL_Maths_P2/ — that isn't a real macOS path), could not be located anywhere in this session's accessible directories. That doesn't mean it doesn't exist; it means a mock candidate should confirm it's actually reachable before building a study plan around sitting either of those two most recent series, rather than assuming the facts bank's earlier finding is still sitting on disk somewhere. As of this document, the facts bank also confirms no examiner report has yet been published for either series (§5), so even once located, neither can be self-marked against examiner commentary — only against the raw mark scheme.

Method for the tiering below. A grep sweep of all twelve WMA12 curriculum files (the eleven real lessons plus paper-anatomy.ts) plus the question bank, done for this document, counted how many separate files cite each real series by name. This is a floor on how much of a series a student may have already seen inside this course, not a measure of how exhaustively any series has been mined — a series with zero citations may still have been read cover-to-cover by an earlier research pass without a specific question making it into a lesson.

Heavily drawn on (four or more separate lesson files each citing a specific named question from this series) — sit the rest, discount the named question(s) when scoring:

Lightly touched (one or two files — close to cold outside that single topic): January 2022 (2 — geometric series, proof), June 2021 (2 — paper-anatomy timing table, trapezium rule), June 2023 (2 — logs, proof), June 2024 (2 — logs, stationary points), October 2021 (2 — proof, trig equations), October 2022 (3 — proof, stationary points, trig equations).

Nearly cold (one file, or a citation that isn't full question content): January 2025 (1 — trapezium rule only), June 2022 (1 — proof only), October 2024 (1 — paper-anatomy.ts cites its cover-page rubric wording only, not any question content — this series is close to fully cold for actual mathematics).

Already seen as a full worked question, not just cited in passing. October 2019 Q10 and October 2021 Q10(i) are each a complete RealPastQuestion entry in content/veridian/question-bank/wma12.ts, with the real question text, mark scheme and examiner-report finding shown directly — treat these two specific questions as review material, not a cold test, even though October 2019 and October 2021 are otherwise lightly touched (1 and 2 curriculum-file citations respectively).

Practical sequencing. Sit the nearly-cold series first while they're genuinely available — January 2025, June 2022, October 2024 are the closest this course currently has to an unseen full paper. Once those are exhausted, move to the lightly-touched tier, reviewing the one or two already-cited questions separately rather than letting them inflate a mock score. Keep the heavily-drawn-on series for last, specifically to practise full-paper timing and self-marking mechanics once genuine cold-content diagnosis is no longer the goal.

---

#3. Setting up a real sitting

• Exactly 90 minutes, one sitting, no pausing. If a genuine interruption happens, log the real elapsed time in §8's timing audit rather than pretending the clock stopped.
• Yellow booklet only — Mathematical Formulae and Statistical Tables, Issue 2, January 2021 — and nothing else in front of you. forge-battle-card.md's formula masterlist names exactly what's on the P2 page (arithmetic/geometric series, the combinatorial binomial expansion, change-of-base logs, the trapezium rule) and what isn't (both sum-formula proofs, the log addition/subtraction/power laws, sin²θ+cos²θ≡1, tanθ≡sinθ/cosθ, and anything for proof, the factor theorem, circle geometry or differentiation/integration technique). A mock sat with a personal formula sheet on hand measures a different, easier skill than the real paper does.
• Check your calculator is Pearson-compliant before the sitting, not the night before the real exam — no symbolic algebra, no symbolic differentiation/integration, no retrievable stored formulae, confirmed verbatim on every cover page reviewed, January 2020 through January 2026 (WMA12-verified-facts.md §3).
• Use a real question paper, sourced from one of the two archives in §2 or directly from qualifications.pearson.com — never a paper reconstructed from this hub's own short, attributed trap-taxonomy quotes, which are fragments of real questions, not full ones.
• Answer every question, in printed order, by default. "Students must answer all questions" — no choice, no Section A/B/C structure, confirmed on every real paper (WMA12-verified-facts.md §3). A mock that skips a question "to save time" isn't testing this paper's actual rules.
• Write your checkpoint numbers at the top of the script before starting, using §4's method below.

---

#4. A real checkpoint-timing method, built from three fully-verified series

paper-anatomy.ts and forge-exam-technique-guide.md between them have independently extracted and verified three complete real per-question mark sequences — January 2021, January 2024 and June 2025, each directly counted off its own primary PDF and each independently summing to 75. No document in this hub has yet turned those three sequences into a running-clock checkpoint table, so this section does that: applying the paper's own real 72-second-per-mark rate (90 minutes ÷ 75 marks) to each cumulative mark total. This is arithmetic on already-verified figures, not a new source claim.

The method, not the table, is what transfers to whichever real series you actually sit: after finishing each question, add its printed mark value to your running total, multiply by 1.2, and that's the elapsed clock-time you should be at or ahead of. The three rows above prove the method is real and self-consistent (all three independently reach exactly 90.0 at Q10) — they are not a claim that every WMA12 paper has 10 questions or this exact shape. forge-exam-technique-guide.md Part 1 already confirms question count runs 9–11 and the per-question spread varies from a 2.75× range (Jan 2021, Jun 2025) to a 4.7× range (Jan 2024); write your own paper's real printed mark values into this table's structure before you start, rather than memorising one of the three rows above as if it generalises.

One fixed anchor holds regardless of question count: the halfway mark, 45 minutes, is exactly half of 90 — worth a glance at the clock on any sitting even before you've built your own full table.

---

#5. The moment the 90 minutes end, before opening anything

Before consulting the mark scheme, go back through the paper and, for every question you feel uncertain about, write a one- or two-word guess in the margin at what might be wrong — "substitution," "sign," "wrong form," whatever comes to mind first. Don't look anything up yet. This is this document's own procedural suggestion, not a Pearson-sourced convention: the reasoning is that a guess written after opening the mark scheme is retrofitted to match what the scheme says, not a genuine record of what you actually suspected on your own working — and comparing your honest pre-scheme guess against §6's actual finding is a useful check on how well you can already self-diagnose without help.

---

#6. The self-mark order — what to check first, and why

Check in this order, not question-by-question through the paper. The ordering follows §1's finding that this paper's dominant real risks are cheap to self-check mechanically before the mark scheme is even open, with the slower, literal M/A/B pass held until step 3.

Step 1 — completion. Was every question attempted? A blank answer scores zero with certainty; even a wrong attempt at a quoted formula has a real chance at a method mark. Confirm this before anything else.

Step 2 — a fast, mechanical pass, done before opening the mark scheme, targeting the specific patterns §1 and forge-exam-technique-guide.md already confirm are real and repeated:

• Every question carrying "show all stages of your working" or a calculator-technology warning: is a formula or method line visible before any number is plugged in? A correct final value with nothing written above it is a confirmed real zero, twice, on two different series (Jan 2024 Q5(i); Jan 2025 Q5(a), §1). This is also the geometric-series lesson's own named trap, calculator-summation-shown-without-formula.
• Every exact-form or surd answer: still exact, or did it become a rounded decimal partway through? Real, checked cost: Jan 2021 Q2's own A1 line reads "do not accept 2.67 but 2.6[recurring] can be accepted" (forge-exam-technique-guide.md Part 2); the same rule cost marks on a real circle question the same way — January 2023, "a common error was giving the answer r = √20" as a decimal instead of the exact surd (forge-exam-technique-guide.md Part 2, reproduced in forge-battle-card.md danger #2).
• Every "using your answer to (a)" or "hence" part: did you actually reuse the earlier result, or recompute it, or reach for an easier unrequested method? Both real, quantified costs from §1 apply here — one scored zero on the part regardless of the final root (Oct 2023 Q4(c)(ii); matches factor-and-remainder-theorems.ts's own trap hence-factorise-reached-without-using-the-given-factor), the other cost marks and time in the same move (Jun 2025 Q5(b)).
• Every trig equation on a stated interval: correct angle mode used throughout, and the final form matching what was actually asked (degrees vs. exact kπ radians)? Confirmed independently on two series, Oct 2022 Q5 and Jun 2025 Q8(i) (forge-exam-technique-guide.md Part 4, item 3).
• Every final answer buried inside a working line or a list: is the actual required value isolated, boxed or underlined on its own? Real, checked cost: Jan 2024 Q2's final mark was commonly lost by candidates who "failed to isolate the required term from a list" — matches binomial-expansion.ts's own trap, required-term-not-isolated-from-a-list.
• Every optimisation or geometry answer: physically possible? A negative length, radius, or a root outside the question's own stated domain is a real, confirmed cost when not checked — June 2025 Q9's examiner report records candidates reaching a negative value and "failed to realise" it was impossible. Flagged honestly: this specific pattern has no matching trap-taxonomy entry yet in any lesson file (forge-exam-technique-guide.md Part 4, item 5, names this directly as unbuilt) — if it costs you a mark, log it as a new, single-instance observation in §7 below, not a forced match to an existing name.

Step 3 — the literal mark-scheme pass, applying the real M/A/B conventions to your own script as strictly as an examiner would, not as leniently as you're tempted to:

• Every A mark is dependent on its own M — if you can't point to the specific line where the method was genuinely earned, don't award the accuracy mark that depends on it, whatever your final number says.
• ft (follow-through) only ever tags A and B marks, never M — a carried-forward wrong value can still earn a correct-ft A mark, but the M mark it would depend on stands or falls purely on the method shown.
• cso ("correct solution only") is stricter than cao — the mark scheme's own wording requires "no errors" anywhere in that part of the question, not just a correct final line — don't award yourself a cso-flagged mark for a derivation that had a slip you happened to correct later.
• The real misreading rule caps the penalty rather than zeroing the part outright: a misread that doesn't materially simplify the question costs a flat two marks off the A/B total for the affected part, per the mark scheme's own general guidance to "deduct two" in that case — apply exactly that deduction on a mock, not a blanket zero and not full credit.
• isw (ignore subsequent working) protects work written after a correct answer already appears on the page — it does not restore a correct answer you later crossed out or changed. A real, checked instance: June 2025 Q3's mark scheme accepts an equivalent, un-simplified form of a correct circle equation and explicitly disregards what's written after it, but that protection does not run backward to a crossed-out correct line — mark the version you actually submitted, not the earlier correct one you "really meant."
• The real multiple-attempts rule is more generous than most candidates assume, and worth applying as written rather than harshly: a clean, un-crossed-out attempt next to a crossed-out wrong start is marked normally, but if two conflicting attempts are both left uncrossed (or both crossed out), the rule is to mark every attempt and award the highest single score — not the lowest, and not the first one written. Apply this rule on a mock exactly as written, rather than assuming an uncrossed false start silently costs you the mark.

---

#7. Logging every dropped mark

For each mark lost in steps 2–3, record it against the matching lesson's own trap-taxonomy item, using the locator table already built in forge-exam-technique-guide.md Part 4 (organised by lesson file, e.g. the-circle.ts's eight named traps, trigonometric-equations.ts's ten). Where a loss genuinely matches a named item, log it there. Where it doesn't — as is currently true for the impossible-root pattern named in step 2 above, or for anything else that surfaces on your own mock and doesn't fit an existing name — write it up plainly as a single observation from your own sitting, not a confirmed pattern; a pattern that recurs across two or more of your own mocks is worth raising for this hub's own next pass to consider folding in, not something this document (or one mock) can establish on its own.

Sort each entry into one of three causes, since the fix differs for each:

• Method not shown, but you knew it — the cheapest to fix; step 2 above should catch most of these before you even reach the mark scheme.
• Genuine content gap — you didn't know the method. Route this to the matching lesson's own worked-chain and trap-taxonomy block via the Part 4 locator, not just the one question you got wrong.
• Never reached — time ran out before the question or part was attempted. Keep this separate; §8 below is specifically about it.

---

#8. The timing audit

Compare your checkpoint numbers from §4 against what actually happened. This is this document's own reasoned categorisation, not a Pearson-stated convention, flagged as such — three different readings of "over budget," rather than one undifferentiated signal:

• Over budget, but the answer was right. A fluency gap, not a knowledge gap — you have the method, it simply took longer than 72 seconds/mark allows. The fix is speed drilling that spec point's own chain-drill, not re-learning the technique.
• Under budget, but the answer was wrong. A carelessness flag, not a timing one — you finished with time to spare and still lost the mark, so the loss belongs in §7's content ledger, not a "give myself more time" plan.
• Never reached in the time available. The one genuinely timing-caused loss. If the same position in the paper, not the same topic, keeps being where time runs out across two or more mocks, the likelier cause is cumulative overspend earlier in the paper, not that specific question.

---

#9. Calibrating the result

signal-frequency-and-boundaries.md Part 2 carries seven real, Pearson-published WMA12 unit grade boundaries, October 2023 through January 2026 (e.g. A = 55/75 raw in Oct 2023, 53/75 in Jan 2024, 48/75 in Jun 2024, 51/75 in Jan 2026) — real published figures, not an estimate. Use the boundary from the series you actually sat, not a generic "A is usually about 70%" assumption: across the seven checked, the A boundary itself has moved by seven raw marks. Never treat a single boundary as this sitting's actual number — that document's own §"how to use this document" note says exactly that, and this document repeats it rather than smoothing it into a single average.

---

#10. Feeding the result back into revision

Route each §7 entry to the resource that actually addresses its category:

• Method-not-shown / presentation entries → forge-exam-technique-guide.md Part 2 (the formula-first rule, the M/A/B mechanics, the abbreviation glossary with real instances).
• Genuine content-gap entries → the specific curriculum lesson file named in the Part 4 locator for that trap — rework its own worked-chain and trap-taxonomy block, not just the single mock question.
• Formula-recall failures → codex-formula-and-method-bank.md, or the condensed version in forge-battle-card.md's formula masterlist.
• Fluency (over-budget-but-correct) entries from §8 → that spec point's own chain-drill, repeated for speed on content you already have correct.
• Never-reached entries from §8 → re-run §4's checkpoint method on your next mock specifically to confirm whether the same position in the paper recurs as the overspend point.

---

#11. Honesty notes

• No grade-boundary movement is presented as a threshold to aim for, and no likelihood or percentage appears anywhere in this document — every count above is a real, named instance or a real published figure, cited to its source.
• §2's tiering measures this course's own citation footprint as of 2026-09-29 on the files actually swept — it is not a claim about what any individual student has personally read, and a heavily-cited series is "partly seen by this course's own material," not "harder" or "easier."
• The missing third archive (§2) is disclosed exactly as found — not located in this session's accessible directories — rather than either asserted as available or silently dropped from the document.
• Every step in §6–§8 that is this document's own synthesis rather than a directly sourced Pearson rule is flagged as such in its own section (§5's margin-guess technique; §8's three-way timing categorisation) — nothing is presented as an official marking or Pearson procedural rule unless it is quoted verbatim from one.
• This document's own genuinely new contribution: the archive-access check and citation-count tiering in §2; the checkpoint-table arithmetic in §4 (built on three already-verified per-question sequences, not a new primary-source find); the self-mark ordering in §6, which resequences already-sourced material specifically around §1's three named failure shapes; and the loss-ledger/timing-audit framing in §7–§8. Every fact those sections rest on — every quoted mark-scheme rule, named series, and trap-taxonomy item — is inherited from WMA12-verified-facts.md, paper-anatomy.ts, or one of the four other technique-hub documents named in the provenance note above, not re-derived from a primary PDF independently for this pass.