Every mark of the 28-question paper was coded to a topic and compared with the forecast published on the site, then the untested syllabus areas were used to sharpen the Paper 2 outlook.
Algebra, numbers, equations, statistics and mensuration carried 46 of the 90 marks. Number work (11) was the single biggest topic.
Grey band = usual range (middle 80% of 2016–25) · forecast · 2026 actual (green in range, orange above, red below)
Checked question by question against the 4052 content list. Anything here and common in earlier Paper 2s is a candidate for Paper 2.
In 2016–25, matrices and sets never appeared in both papers of the same year: each year it was exactly one or the other. P1 2026 had sets, none on matrices, so matrices are very likely to appear in P2. The one-paper-only split held 10 years out of 10 (about a 1-in-1,000 chance by luck), which is why this page rates matrices “very likely” and not just a priority.
Honest result: a good planning guide, not a precise one. It was as accurate as simply averaging 2016–25 (78%) and a little better than last three papers alone (74%). Both specimen papers landed a similar share (72% Claude, 73% ChatGPT). These are shares of marks by topic, not a measure of whether questions were predicted: ChatGPT’s review, which compares question by question, found 7 specimen questions (25 of 90 marks) used methods absent from the real paper. Its coding of the real paper matches mine for 88 of 90 marks (only Q3 is split differently).
| Prediction | 2026 result | |
|---|---|---|
| Algebra is the biggest block (12 marks, range 9–15) | 10 marks, 2nd biggest topic | ✔ |
| Statistics around 8 marks | 8 marks (box plots, grouped mean, SD concept) | ✔ |
| Equations around 7 marks | 9 marks (range topped out at 7.8) | ≈ |
| Trig around 7 with bearings and a triangle | 6 marks, bearings yes, sine/cosine rule no | ≈ |
| Numbers 7 (range 6–10) | 11 marks, above range | ✘ |
| Mensuration 5 (range 3–6) | 8 marks, above range | ✘ |
| Coordinate geometry “now rare” (0–1 marks) | 5 marks, my biggest miss | ✘ |
| Percentages 3 (range 2–4) | 0 coded marks, below range (the method was used inside Q17 and Q27) | ≈ |
| Matrices and sets “coin-flips” | Sets 3 marks, matrices 0 | ✔ |
| Sequences and construction “coin-flips” | Both 0 marks | ✔ |
| “A proof” and “explain” questions | 6 reasoning questions, 13 marks | ✔ |
| No long multi-part question; stays short | Longest were 6 marks | ✔ |
| Speed/travel not above 4 marks | 4 marks (average speed) | ✔ |
16 of 21 topics inside their usual range. Missed above range: Equations & inequalities, Numbers & indices, Mensuration, Coordinate geometry. Below: Percentages (no mark coded to it, although Q17 and Q27 used percentage methods).
The forecast’s top five (algebra, statistics, equations, trig, numbers) caught four of the paper’s top five. Mensuration (8) tied statistics for fourth and replaced trig.
Lesson: single-year jumps of 3–5 marks in a topic are normal. Number work, mensuration and coordinate geometry can each rise suddenly, so treat a forecast range as the middle 80% of years, not a ceiling. This is the same “no topic is due” point made on the Analysis tab: 21 topics, 5 outside range is about what chance predicts.
Built from the 2016–25 Paper 2 history (recency-weighted) and then adjusted for what Paper 1 left untested. The “P1” column is the 2026 Paper 1 mark. Verdicts mix how often a topic appears in Paper 2 with what Paper 1 left untested, so topics in nearly every Paper 2 (statistics, mensuration, probability) rate high even though P1 covered them. Verdicts are planning guides, not probabilities: they rest on ten comparable years.
Revision priority if time is short: (1) graph plotting and tangents, (2) quadratic modelling with the formula, (3) matrices, (4) sine/cosine rule and 3-D, (5) cumulative frequency and standard deviation with probability, (6) vector or coordinate geometry, (7) composite-solid mensuration. The topics behind these seven (graphs, equations, matrices, trig, statistics, probability, vectors, coordinates, mensuration) carry about 55 of the 80 non-real-world marks in a typical Paper 2.