Deep Dive: Elementary Probability and Conditional Notation

Provenance (2026-09-29): read in full — research/veridian/WST01-verified-facts.md (§1 spec table, §2a formula-booklet contents, §2b AO weighting, §3 source archive/denominators, §5 Section 3 "Probability 3.1–3.4", §7 topic-strength summary, §9 spot-check addendum) and content/veridian/curriculum/wst01/elementary-probability-and-conditional-notation.ts in full (744 lines, including its header note and closing flag block). Grepped and spot-read — content/veridian/curriculum/wst01/conditional-probability-independence-vs-mutually-exclusive.ts (header + audited-anchor section), content/veridian/technique-hub/wst01/codex-formula-and-method-bank.md (§1.2, §2.3), content/veridian/technique-hub/wst01/forge-exam-technique-guide.md (trap-taxonomy table), content/veridian/technique-hub/wst01/signal-frequency-and-boundaries.md (full), and content/veridian/question-bank/wst01.ts (confirmed no 3.1/3.2 items exist there — its two specRefs are 2.4 and 5.3). No new PDF fetches were made; no local WST01 PDF archive exists on this machine, matching every other WST01 technique-hub document's own provenance note.

#CODEX deep-dive: elementary probability and conditional notation

#What this document is, and isn't

This is a single-topic reference for WST01 spec items 3.1 ("Elementary probability") and 3.2 ("Sample space. Exclusive and complementary events. Conditional probability"), synthesising the complete real theory, every distinct sub-case the spec actually requires, and every real exam angle this specific notation has been tested from — not a lesson, and not a restatement of one.

Two real documents already own the full worked-example treatment of this ground and are deliberately not reproduced here: elementary-probability-and-conditional-notation.ts (the primary lesson, whose own header records exactly which examiner-report citations it uses and why) and its sibling conditional-probability-independence-vs-mutually-exclusive.ts (spec 3.2/3.3, owning the tree-diagram and Venn-diagram independence-testing ground). Where this document touches a sub-case those lessons already teach in full, it gives a short, freshly-built micro-illustration only, and points to where the complete worked treatment lives, per this course's standing rule against padding a document with material that already exists elsewhere.

---

#1. Spec anchor, verbatim

Spec §S1.3, verbatim: "3.1 Elementary probability." and "3.2 Sample space. Exclusive and complementary events. Conditional probability.", with the guidance line: "understanding and use of P(A′) = 1 − P(A), P(A ∪ B) = P(A) + P(B) − P(A ∩ B), P(A ∩ B) = P(A)P(B|A)." (Pearson Edexcel IAL Mathematics spec, Issue 3, April 2019, p.53 — verified in WST01-verified-facts.md §1).

The formula booklet split matters more here than for most WST01 topics, because it draws the exact line between what a question hands you and what you must produce unprompted (verified against the real booklet, Issue 2, January 2021, pp.13–17 — WST01-verified-facts.md §2a):

The asymmetry is worth stating plainly because it is the single most exam-technique-relevant fact about this topic's notation: the general conditional-probability identity is printed for you, but the one-line complement rule — arguably the simplest fact in the whole unit — is not, and nothing on the paper will remind you of it.

---

#2. The complete theory

#2.1 Sample space and classical probability

An experiment is any process with an uncertain result; every possible result is an outcome; the full set of outcomes is the sample space $\Omega$; an event is a subset of $\Omega$. When every outcome is equally likely, probability reduces to counting: $P(A) = n(A)/n(\Omega)$. This is genuinely the whole of what spec 3.1 asks for as a standalone idea — the spec's own unit description names "the basic ideas of mathematical modelling as applied in probability and statistics" as its content, and, per WST01-verified-facts.md §5 (Section 3), there is no separate examiner-report subsection for 3.1 alone; it functions as the counting foundation everything else in Section 3 sits on top of.

Two counting mechanics do almost all the real work: independent stages multiply, not add (two rolls of a fair 4-sided die give $4\times4=16$ outcomes, not $4+4=8$), and multi-stage outcomes are ordered unless the question says otherwise, so $(2,6)$ and $(6,2)$ are two outcomes, not one.

#2.2 Events as sets: complement, exclusivity, union

$A'$ is everything in $\Omega$ not in $A$, so $A$ and $A'$ partition $\Omega$ and $P(A')=1-P(A)$. Two events are mutually exclusive when $A\cap B=\varnothing$ — a statement about whether the underlying sets overlap, fixed before any probability is attached, not something a probability calculation can create or remove. The general addition law, $P(A\cup B)=P(A)+P(B)-P(A\cap B)$, is a double-counting correction: adding $P(A)$ and $P(B)$ counts the overlap twice, and the $-P(A\cap B)$ term removes exactly one of those two copies. When $A,B$ are mutually exclusive, $P(A\cap B)=0$ and the law collapses to the simple sum $P(A\cup B)=P(A)+P(B)$ — a genuine shortcut, but only in that one case.

#2.3 Conditional probability, and a general proof worth stating explicitly

$P(A\mid B)$ restricts the sample space to $B$: once $B$ is known to have happened, only outcomes inside $B$ remain relevant, so $P(A\mid B) = n(A\cap B)/n(B) = P(A\cap B)/P(B)$ for $P(B)>0$. This is the spec's own definition, printed in ratio form on the formula sheet.

One property of this definition is worth deriving rather than just asserting, because it is the exact mechanism behind this topic's single most-repeated real exam error (§4 below). Since $A\cap B \subseteq A$, it is always true that $P(A\cap B) \le P(A)$, and therefore:

$$P(A\mid B) = \frac{P(A\cap B)}{P(B)} \;\le\; \frac{P(A)}{P(B)}$$

with equality iff $A\cap B = A$, i.e. iff $A$ is entirely contained in $B$. In every other case, the raw ratio $P(A)/P(B)$ strictly overshoots the true conditional probability — it can never fall short of it. This is a clean, general statement of what several real examiner-report citations describe as an isolated error (§4): treating $P(A\mid B)$ as $P(A)/P(B)$ is not a "sometimes wrong" shortcut, it is a value that is provably always too large unless $A\subseteq B$, and this is a fact about the definitions themselves, not a fact that needed a past paper to establish — the past papers below simply confirm how often, and in what disguises, it actually gets used.

---

#3. Every distinct sub-case the spec requires, each with a short micro-example

3.1 — Classical counting with independent, ordered stages. Two fair coins: outcomes $HH, HT, TH, TT$ ($2\times2=4$, ordered). $P(\text{exactly one head}) = 2/4 = 1/2$.

3.2 — Complement of a compound (multi-region) event. A sample space partitions into four mutually exclusive regions with $P(\text{region 1})=0.1$, $P(B)=0.3$, $P(\text{region 3})=0.2$, $P(\text{region 4})=0.4$ (summing to 1). $P(B') = 0.1+0.2+0.4 = 0.7$ — the complement of a compound event is the sum of every other region in the partition, not a single subtraction from a single number. This is exactly the mechanism behind the real anchor below (Jan 2021 Q1(a)), illustrated here with fresh numbers rather than reproduced.

3.3 — General addition law vs. the mutually-exclusive special case. Overlapping: $P(A)=0.5$, $P(B)=0.3$, $P(A\cap B)=0.1 \Rightarrow P(A\cup B)=0.7$. Mutually exclusive: $P(C)=0.2$, $P(D)=0.15$, $C\cap D=\varnothing \Rightarrow P(C\cup D)=0.35$ — the same law, with its correction term already zero. This second form is the real mechanism behind Jan 2021 Q1(b)'s mark scheme, which explicitly lists "$0.35+0.35$" as an accepted route precisely because the two named events there don't overlap on that question's own diagram (WST01-verified-facts.md §5, per the mark scheme bullet for part (b)) — deliberately not reproduced in the primary lesson (its own header records this as an intentional omission, to avoid duplicating its prequestion epn-pq-2), and not reproduced with the real numbers here either, for the same reason.

3.4 — Conditional probability, simple case. 12 balls: 5 red (2 of them shiny). $P(\text{shiny}\mid \text{red}) = 2/5$.

3.5 — Conditional probability where the conditioned event is a compound complement: $P(A\cap B')$, $P(A\mid B')$. Partition: $P(A\cap B')=0.2$ (A, but not B), $P(A\cap B)=0.1$, $P(B\text{-only})=0.3$, $P(\text{neither})=0.4$. Then $P(B')=0.2+0.4=0.6$ and $P(A\mid B') = 0.2/0.6 = 1/3$. This is the compound form the real Jan 2021 Q1(a)–(c) mark scheme's own M1 mark specifically requires — $P(A\cap B')/P(B')$ with a genuine attempt at the numerator — and which, per this course's own coverage audit of the primary lesson (2026-09-12), had been quoted four separate times for its raw-ratio error before that gap was closed. The micro-example above is a fresh, independently-checked illustration of the same mechanism, not the real question's own numbers.

3.6 — Conditional probability whose numerator must be built from more than one route. Abstracted micro-illustration of a real, two-part error family: suppose the numerator event $M$ can occur via two mutually exclusive routes inside the conditioning event $N$, with $P(\text{route 1}\cap N)=0.05$, $P(\text{route 2}\cap N)=0.03$, and $P(N)=0.4$. The correct value sums both routes: $P(M\mid N) = (0.05+0.03)/0.4 = 0.2$. Two distinct, separately-verified real error mechanisms live here: (i) using only one route — $0.05/0.4=0.125$, an undercount, matching the error Pearson's own examiner report records on a three-branch tree-diagram question as using a product of two probabilities where three were needed (Oct 2021, Q4(c)); and (ii) folding a spurious extra copy of the conditioning probability itself into the numerator — $(0.05+0.03+0.4)/0.4 = 1.2$, which (like the real case) overshoots 1 and self-diagnoses. This second mechanism is a direct, deliberately-simplified abstraction of a real, independently re-verified error: on a real tree-diagram question, both of the examiner report's own quoted wrong answers used the correct denominator but wrongly re-added or re-multiplied that same denominator value back into the numerator (Jan 2023, Q2(d) — re-verified against the actual mark-scheme PDF rather than trusted from an earlier, since-corrected summary; see §4 below for the correction's own provenance).

3.7 — Conditional probability read off a Venn diagram where independence is wrongly assumed instead of tested. Stated at the level of mechanism only, since the full worked treatment already lives in the sibling lesson: a diagram gives every region's probability directly, but a common real error substitutes the independence shortcut $P(A)P(B)$ for the diagram's own $P(A\cap B)$ region without checking whether the two actually agree (Jun 2022, Q4(b); corroborated by Pearson's own paper-wide diagnosis that series that candidates "often assume independence when an appropriate conditional probability should be used instead"). Full worked treatment: conditional-probability-independence-vs-mutually-exclusive.ts.

3.8 — The 3.2/3.3 boundary itself. The general form $P(A\cap B)=P(A)P(B\mid A)$ is on the formula sheet; the independence-specific simplifications $P(B\mid A)=P(B)$, $P(A\mid B)=P(A)$, $P(A\cap B)=P(A)P(B)$ are spec item 3.3, explicitly listed among the formulae students must produce from memory (WST01-verified-facts.md §2a). Confusing the two directions — assuming the special case applies without testing it — is the single mechanism behind sub-case 3.7 above and the whole of the sibling lesson; this deep-dive's job is only to mark where that boundary sits, not to re-teach across it.

---

#4. Synthesis: every real exam angle this topic has actually been tested from

Denominator, stated honestly. Per-topic visibility exists only for the 6 series whose examiner reports were located and fully read (Jun 2019, Jan 2021, Oct 2021, Jun 2022, Jan 2023, Jun 2024 — WST01-verified-facts.md §3). The other 8 series' mark schemes were checked for notation/marking conventions only, not searched question-by-question for this topic. Every frequency claim below uses 6 as the denominator and names exactly which series confirm it — never a percentage, never a grade boundary (none exist anywhere in the source material for this paper — confirmed by a direct string search of the facts bank, per signal-frequency-and-boundaries.md §2).

On the Jan 2023 Q2(d) correction. WST01-verified-facts.md itself records that an earlier pass mischaracterised this question's error as "missing the correct conditioning denominator." A 2026-08-28 re-verification against the actual mark-scheme PDF (rendered as an image, since pdftotext garbles this question's fraction layout) found both quoted wrong answers use the correct denominator throughout — the real fault is a spurious extra copy of that same value added or multiplied into the numerator. This deep-dive uses only the corrected version (sub-case 3.6 above); it is recorded here so the correction's own provenance stays traceable rather than silently inherited.

Frequency, stated as a lower bound, not a percentage. A specific question or general-comment citation for this topic's own conditional-probability-notation error exists in 4 of the 6 fully-read examiner-report series: Jan 2021, Oct 2021, Jun 2022, and Jan 2023. It is not confirmed — meaning no citation was found, not that the content was absent — for Jun 2019 or Jun 2024. No standalone spec-3.1-only question (elementary probability with no conditional element at all) was found anywhere across the 14 series' mark schemes reviewed; 3.1 appears as the counting foundation embedded inside 3.2-level questions, never as its own separately-marked item — consistent with the wider Section 3 pattern (signal-frequency-and-boundaries.md §1 records the whole 3.1–3.4 block, not this narrower slice, as confirmed in 4 of 6 series with one further mark-scheme-only citation in Oct 2022).

A real, exam-technique-relevant nuance on answer form. Jan 2021 Q1(c)'s own mark scheme explicitly refuses the accuracy mark for $0.15/0.4$ left unsimplified — it must be reduced to $3/8$ or $0.375$. Elsewhere on this paper, the general marking abbreviation oe ("or equivalent") is used freely — a real Oct 2022 mark-scheme line accepts a fraction, and separately notes it will also accept "awrt 0.556 or awrt 55.6%" for the same value. The two coexist without contradiction: oe covers equivalent forms of an already-fully-simplified value; it does not license leaving a fraction unreduced when a mark scheme's own notes specifically call for the reduced form. A student relying on oe as a general licence to skip simplification on this topic would be relying on a real marking convention in the wrong place.

---

#5. What this document does not claim

No grade-boundary data exists anywhere in the verified source material for WST01 (confirmed by direct string search, signal-frequency-and-boundaries.md §2) — none is invented here. No likelihood or percentage figure is given anywhere above; every frequency claim names the specific series it is based on and the denominator it is drawn from. The 8 series without a located examiner report (Jun 2021, Jan 2022, Oct 2022, Jun 2023, Oct 2023, Jan 2024, Oct 2024, Jan 2025) were not searched question-by-question for this topic, beyond the two mark-scheme-only citations already used above (Oct 2022's oe note, Jan 2025 Q1(b)); this document's exam-angle synthesis in §4 should be read as a lower bound on what a full read of those 8 series would surface, not as an exhaustive record — the same caveat the underlying facts bank states about itself.