It all comes up. This is where the marks have fallen, not what to leave out.
This is a record of what has appeared in real exam papers and how often. It is not a list of what to skip, and it is not a prediction: exam boards do not work to a rota.
Built from 270 real OCR, AQA and Edexcel GCSE (9 to 1) Maths papers (June and November 2017 to June 2025), every question-part tagged to a topic and cross-checked against the mark scheme. Marks on multi-topic (synoptic) questions are split across their topics. Spec strands are assigned by a topic classifier. The overall Assessment Objective mix is mandated identical for every board (Ofqual: F 50/25/25, H 40/30/30) and the estimate is calibrated to match it (verified against AQA's published grid); the per-topic split is AQA-calibrated by question type. How students score is the average share of marks scored on a topic in real papers, from AQA and Edexcel question-level analysis: a descriptive past-performance figure, not a prediction. Rising and falling is a linear fit of a topic's marks per paper year on year and is shown only with six or more sittings behind it.
The Topic sheet, Papers & search and Literacy, board comparisons, trends over time and the CSV download are free for teachers with an account.
Free, and takes a few seconds. The hand-out stays open to everyone.
Signed in as you. We will email occasional teacher updates (new analysis, free resources). Unsubscribe any time. No spam.
One row per topic: how often it appears, the marks it carries a paper, how students score on it, when it last came up. Hardest topics that come up a lot come first; the rest of the list still carries its marks.
From the boards' examiner reports: on each topic in view, the mistake that most separates stronger from weaker students, in the record's order. What to teach against, not what to drop.
The most common short question types (whole questions and sub-parts) across the tier and boards in view, with the mark-scheme note; counted over every paper type, so the paper filter does not apply here. Every paper has them; none of them is the whole paper.
Pick a topic; every panel on this sheet is about it. Boards, tier and paper follow the head sheet.
Open the teacher toolkitShare of each tier's sitting marks (ignores the tier filter).
The average share of marks students actually score on this topic in real papers (question-level analysis, AQA and Edexcel, both tiers). Lower is more demanding. A past-performance figure, not a prediction.
Routine (AO1), reasoning (AO2) and problem-solving (AO3) marks for this topic, calibrated from AQA's published per-question grid. A model, not official per-paper tags.
What examiners report students get wrong on this topic, paraphrased from OCR, AQA and Edexcel examiner reports. A starred line separates stronger from weaker students.
The task types for this topic (board, tier and paper as set above), with the mark-scheme answer or method.
Share of each board's paper marks on this topic. A share, not a count: OCR papers are out of 100, AQA and Edexcel out of 80.
Marks a paper on each type of paper. Two of every three papers are calculator, so this is per paper, not a split of the total.
Marks by position: early questions are usually easier, later ones harder.
How many marks each question on this topic is worth.
Topics this one is combined with in synoptic questions, across all papers.
Share of each sitting's marks, as filtered.
Every task type and mark-scheme note matching a word, for example Venn, iteration, show that.
Reacts to the head sheet.
Marks a paper across the papers in view. Click a bar to open that topic's sheet.
A straight-line fit of a topic's average marks per paper, year on year (years where the topic did not appear count as 0). A description of the past record, not a prediction. n is the number of sittings behind the fit; fewer than six and the topic is not shown.
Regular topics (in five or more of the sittings in view) ranked by how long since they last came up. This describes the past record. It is not a prediction that they are due; exam boards do not work to a rota.
The synoptic combinations that come up most across all papers, from the roughly 8% of question-parts tagged with two or more topics: the strongest signals, not an exhaustive list.
What students are asked to do, per hundred questions in view: solve, prove, explain. Inferred from each question's description, not the verbatim wording, so read it as indicative; phrasings that match no standard command word are left off.
Routine, reasoning and problem-solving marks across the papers in view. The overall balance is mandated identical for every board (Ofqual: F 50/25/25, H 40/30/30) and this estimate is calibrated to match it.
Pick any paper to see its topic breakdown by marks.
Most frequent hard non-technical words (a general-English barrier, separate from maths vocabulary): accurately, calculate, estimate, possible, decimal, different, minutes, significant, coordinates, figures, average, represents, statement, rectangle, following, centimetres, represent, integers, transformation, proportional, another, interval. Ranked by word count, a plain count and not a reading-age formula.
| Wordiest question | Q | Topics | Context words | Marks |
|---|---|---|---|---|
| OCR 2020 Nov P2 F | 6 | compound-measures | 654 | 78 |
| OCR 2022 Nov P6 H | 11 | expanding-factorising | 510 | 59 |
| OCR 2023 Jun P6 H | 12 | bounds | 478 | 52 |
| OCR 2023 Nov P6 H | 1 | scatter-graphs | 177 | 7 |
| OCR 2023 Nov P3 F | 19 | scatter-graphs | 177 | 7 |
| AQA 2021 Nov P3 F | 11 | probability-basics | 177 | 4 |
| AQA 2018 Nov P4 H | 8 | estimation, scatter-graphs | 161 | 6 |
| AQA 2018 Nov P1 F | 27 | estimation, scatter-graphs | 161 | 6 |
| OCR 2024 Nov P4 H | 4 | compound-measures, scatter-graphs | 159 | 8 |
| OCR 2024 Nov P1 F | 19 | compound-measures, scatter-graphs | 159 | 8 |
Question wording is never shown for embargoed series; these are topic and word-count metadata only. Reading-age data is withheld pending validation.
Source: MathedUp crawl of real OCR, AQA and Edexcel GCSE (9 to 1) Maths papers, mark-scheme cross-checked.