Why Are A-Level Maths Grade Boundaries Going Up?

Every August a student works out what they needed, compares it with last year, and finds the bar has moved again. In June 2019 a grade A in Edexcel A-Level maths took 165 marks out of 300. This summer it took 210.

The usual explanation is that students keep getting cleverer, so the bar keeps rising. That is not what moved it. There are two bigger reasons, and they change how you should revise.

Everything below is Edexcel A-Level maths, specification 9MA0, marked out of 300 across three papers. The AS is a separate qualification with its own cohort and its own boundaries, and it is not in these figures.

---

The paper changed, not the students

Every June is a new set of papers, and no two sets are equally hard. When the papers are kinder, everyone scores more marks, so the mark you need for a grade has to rise. Otherwise far more students would get the grade than it is meant for.

Look at the whole run first. In June 2019, 40.6% of students got a grade A or better and the boundary was 165. In June 2026, 41.5% did and the boundary was 210. Forty five more marks. Nine tenths of a percentage point more students.

Pearson Edexcel A-Level maths 9MA0

Marks needed for a grade A, out of 300

155

2019: 165 2026: 210

215

164 to 214 marks across the 6 series

Share who got a grade A or better

30%

2019 and 2026: 40.6% and 41.5%

50%

40.6% to 46.9% across the same 6 series

If the rising boundary were the year group getting stronger, the bottom rail would move with the top one. It did not.

A single year shows it even more clearly. From 2024 to 2025 a grade A cost 9 more marks, rising from 205 to 214. The share getting an A or better did not rise. It slipped, from 41.4% to 41.3%. The 2025 papers were simply easier.

Ofqual puts it in one line:

If an exam is easier than in previous years, the grade boundaries for that paper will be higher. If it is harder, the grade boundaries will be lower.

That is how a grade A this year is kept worth the same as a grade A in 2019, whatever this year’s papers turned out to be like. A higher boundary tells you about the papers. It does not tell you that you have to beat last year’s students.

We have written separately about what boundary movement means for your revision target, which is the part that actually affects your marks. GCSE maths does the same thing, and a grade 9 has ranged over 27 marks across its run: why GCSE maths grade boundaries are going up. Further Maths shows the same two reasons, and there you can watch results dip when its new course arrived and then climb back: why Further Maths grade boundaries are going up.

Students got better at the exam, not at maths

This A-Level was first sat by a full year group in June 2019. Teachers had just one earlier set of real papers to learn from, sat by a small group in 2018. A grade A took 165 marks.

The next normal summer was 2023, and a grade A took 196. The share getting an A or better had barely moved, 40.6% before and 40.9% after.

Part of that climb is what Ofqual calls the sawtooth effect. Results dip when a new exam arrives, then climb back as teachers learn how questions tend to be worded, where the method marks sit, and how an examiner wants working set out. By 2023 there were several years of past papers to practise from.

For A-Level it cannot be separated out as neatly as it can for GCSE, because the pandemic sits in the middle: no summer exams in 2020 or 2021, and a deliberately generous 2022. But the direction is the same, and so is the lesson.

That is real progress, but it is progress at the exam. Knowing how to write an answer an examiner will credit is worth learning. It is just not the same thing as understanding more maths.

Can we tell whether A-Level students are getting smarter?

No, and it is worth knowing why. Before the papers are marked, Ofqual works out roughly what share of this year group should get each grade, from their GCSE results two years earlier, and sets the boundaries to match. A number held steady like that cannot tell you whether anybody improved.

GCSE has a way round this: a separate test, the National Reference Test, that is allowed to move. Nothing like it exists above GCSE. So these figures show exactly what the boundary did, but not whether the 2026 year group was stronger at maths than the 2019 one. Anyone who tells you that, in either direction, is guessing.

Every June series of this specification

A series with a cohort under a quarter of the largest here is greyed and left off the figure above, which names it: a few thousand candidates sitting a specification early is a fact about who sat it rather than about the standard. Each share is the students who got that grade itself, so the rows do not add up to the figure Pearson publishes, which counts everyone at that grade or above. An asterisk marks a year whose shares are Pearson's first published figures, which it revises later; the boundaries on that row are final. No summer exams were sat in 2020 or 2021, so there are no boundaries for those two years. The grades came from teacher assessment instead, which is why they are the two highest lines in the table. Boundaries are from Pearson's notional component documents and shares are from Pearson's grade statistics. Both are linked on the GCSE maths past papers page, series by series.

Read each pair of columns together. At grade A, the marks needed have ranged over 50 across these sittings and the share getting that grade has ranged over seven tenths of a point, from 24.0% to 24.7%. Each share in the table is the students who got that grade itself, so it reads 24.4% for a grade A this summer. The 41.5% quoted earlier counts everyone at an A or above, which is the figure Pearson publishes and the one the rails above plot.

The years that really did move

Look at 2020 and 2021. They are the two highest lines in the table, at 23.5% and 28.5% for an A*, against 16.3% in 2023. They are also the two years with no summer exams. Grades came from teachers instead, and the autumn sittings that replaced the exams were tiny: 2,395 students in October 2020 against 79,068 this summer.

2022 is next, at 22.4%, with the lowest boundaries in the table: a grade A for 164 marks, a grade E for 28, which is 9% of the paper. That was the staging post on the way back to pre-pandemic grading.

Then 2023 proves which way round it works. The grade A boundary rose 32 marks, from 164 to 196, and the share getting an A or better fell six points, from 46.9% to 40.9%. Boundaries up, outcomes down, in the same year. That was Ofqual pulling results back to pre-pandemic levels, not a weaker year group meeting a harder paper.

So the biggest moves in the share of top grades came from how grades were awarded that year. None of them came from the papers getting harder.

The three papers do not move together

One more thing the per paper figures show, which the totals hide. Across these June sittings the A* boundary on a single paper has ranged from 70 to 89 out of 100.

June 2024 wanted 89 on Paper 3, the Statistics and Mechanics paper, and 81 on Paper 1. Whichever paper you have decided is the soft one, some series disagrees with you.

Your grade comes from the total across all three, so a weak paper can be rescued by a strong one. But it also means a paper you had written off as easy can quietly cost you a grade.

What to do with this

Stop using last year’s boundary as this year’s target. It describes last year’s papers. Yours have not been written into a boundary yet.

Aim above the top of the recent range, not at the middle of it. Across the six normal sittings since 2019 the grade A boundary has run from 164 to 214 out of 300, and this summer it was 210. Train for 214 and a hard paper cannot take the grade off you.

Do not read a high boundary as bad news. A high boundary usually means an accessible paper, which is the paper you want. It also means marks are cheaper to lose, so the careless ones cost more.

Every series, paper by paper, with its boundaries and its outcomes printed beside them for both the A-Level and the AS, is on our A-Level maths past papers page. To turn your own mock marks into a realistic grade range, use the grade predictor.

The short version

A higher boundary does not mean students got smarter. It happens for two bigger reasons.

The papers change every year. When they are easier, everyone scores more, so the boundary rises to keep the grade worth the same. From 2024 to 2025 a grade A cost nine more marks, and the share getting one slipped rather than rose.

Students learn the exam. In the first years of a new A-Level, teachers and students learn what the examiners want, and marks rise without anyone necessarily understanding more maths. Ofqual calls it the sawtooth effect.

Put together, forty five more marks since 2019 bought nine tenths of a percentage point. So the number to beat is not last year’s boundary. It is the best mark you are capable of putting on the paper.

---

Figures in this article come from Pearson’s notional component grade boundaries and Pearson’s grade statistics for 9MA0, both published after each series. The quotation is from Ofqual and the Department for Education’s guidance, GCSE and A level grading: what you need to know, and the 2022 and 2023 grading policy is set out in Ofqual’s approach to grading in summer 2022. The sawtooth effect is explained in GCSEs 2021: what is the sawtooth effect? and studied in An investigation into the Sawtooth Effect in secondary school assessments in England. The National Reference Test is described in Ofqual’s National Reference Test annual statement 2026.