What does a top-grade percentage leave out?

Two schools, one summer

Both tables are from the summer of 2025, the A-level results of St Paul’s Girls’ School and of St Paul’s School, each in its own grades. The SPGS table carries its source’s provisional label, a word that leaves the door open, and the label stays part of the example.

A-level grades at A*: SPGS’s recorded result was 7.5 percentage points higher than St Paul’s in 2025.

A-level grades at A or above: SPGS’s recorded result was 3.8 percentage points lower than St Paul’s in 2025.

A-level grades at B or above: SPGS’s recorded result was 2.7 percentage points lower than St Paul’s in 2025.

Move the line and the direction changes, so a comparison at the top grade alone would miss a meaningful difference in the broader recorded distribution. The tables describe examination entries, though, and why the distributions differ they don’t say.

Beyond the top grade

An A* share asks one narrow question, how often the highest grade was awarded. Widening the band to A*–A and then A*–B turns it into a broader one, about how far strong grades reach across the entries, and a school can look different at each step. A*–A includes the A*s and isn’t the share awarded an A alone; getting at the As alone needs compatible grade counts, or figures for both bands precise enough to subtract, and a chart’s shape can’t supply what the source never published.

How often, and how many

Every percentage stands on a number. A high share may describe a small number of entries while a lower share comes with many more awards, and setting the count beside the rate helps tell them apart, bearing in mind that each pupil can contribute several entries. SPGS publishes its entry total; the St Paul’s figures above have no total recorded, and we leave it unknown, since a pupil count from elsewhere wouldn’t mend the gap and multiplying a rounded percentage by an entry total need not give back an exact number of grades.

What the comparison cannot see

Subject choices, earlier attainment and admissions policies all help shape a year’s results, though no grade table shows them. A large share at A*–B says nothing about the lowest grade any pupil received, and a large share at A* tells a new pupil nothing about what they are likely to achieve. One subtraction is especially tempting. GCSE and A-level results from the same summer usually describe different pupils, so the gap between the two percentages is no measure of progress; measuring that would mean following the same pupils through suitable assessments and allowing for where each of them started.

Moving the line

In the comparison below, the schools and years can stay put while the line moves from A* to A*–A or A*–B, and what matters is whether the reading changes, and whether the change survives a look at any years left out. Our example stays fixed to one dated summer, provisional label and all; a comparison of your own need not.

Where these figures come from

Each record gives the full definition, the documents it came from and what we checked.

1. St Paul’s Girls’ School, A level 2025: 66.8% of 389 entries graded A*

   The school’s A-level grade share, published as provisional. The number of A-level subject entries, rather than the number of pupils.

   Checked against the original: A*, A*–A, A*–B and entries.

2. St Paul’s School, A level 2025: 59.3% of entries graded A*

   The school’s A-level share at the same grade band.

   Checked against the original: A*. Not yet checked: A*–A and A*–B.

• This piece builds no league table and no composite school score.
• The Bayesian and regression calculations in the earlier research are not reproduced here.