Add a student's prior attainment to model their likely A-level outcomes. Nothing you type leaves your device — want to explore first? Load a random student in one click.
Tap +/− to set a grade. Leave subjects you didn't take blank.
Select the subjects the student is taking.
Contextual factors nudge the model, exactly as published.
{{ errorMsg }}
Based on a mean GCSE of {{ meanLabel }}. Bands show the range within which each grade is most likely to land — a guide for target-setting, not a guarantee.
Upload a spreadsheet of students and get every forecast in one table. Download the sample, replace the rows with your own, and drop it back in.
Columns: Name · GCSE Grades · Subject 1–4 · Disadvantaged · EAL · Birth Month · Attendance
{{ batchError }}
No black box, no proprietary score. Every prediction is read straight from real national results — here's the whole pipeline, start to finish.
That score drops the student into a national band — from <1 up to 9+ — the rows of the matrix below.
For each subject, read the real grade distribution for students who started in that band — pooled over four years, cohort-weighted.
The distribution's centre is the predicted grade; its spread sets the tolerance band and confidence. Context factors nudge the result.
Every forecast is read from 3,113,847 real A-level results in the DfE's official 16–18 transition matrices — the grades actually achieved across England over four years, grouped by the level students started from. The grid shows one real slice: A-level Biology. Read across a row for the likely spread of grades in that band; the bright diagonal is prior attainment carrying through to results.
The core forecast is built on prior attainment alone. On top, four optional context factors refine it — and unlike the black-box products, every one is anchored to a published national statistic, no invented weightings. Here is each factor, with its source:
Every coefficient, derivation and source is served openly at /model — audit it, question it, improve it. And this is just the foundation: our roadmap takes prediction further with pupil-level research data. See where we're headed →
Grade prediction shapes real futures. Teachers deserve to see — and question — the maths behind every target they set.
Every coefficient, benchmark and formula is published. No proprietary scoring, no hidden weightings.
Students, tutors, teachers and whole schools — no licence fees, no per-seat pricing, no demo calls.
Fork it, audit it, recalibrate it on your own results. Contributions make the model better for all.
Today's model reads the best free national data. The next leap comes from pupil-level records — and we have a clear, legitimate path to get there.
3.1M real A-level results from the DfE transition matrices, plus context factors anchored to published research. Free for every school, forever.
A public-benefit project with a leading education faculty, analysing the National Pupil Database inside the ONS Secure Research Service — the only legitimate route to pupil-level data.
Predictions conditioned on subject-specific GCSEs, school type and deprivation decile — disclosure-checked, published as open aggregates, and free for anyone to build on.
We aim to work with the University of Cambridge's education researchers on the question that matters most — what predicts A-level outcomes beyond mean GCSE? The National Pupil Database can answer it: GCSE-to-A-level linkage alongside free-school-meals, ethnicity, SEN, deprivation and school type. It can't be used commercially, but a university public-benefit project can study it and publish the results openly — so richer predictions reach every school, college and local authority for free. Slower, but legitimate, and built to last.