Less accurate than most schools assume. Research on teacher forecasts for GCSEs and A Levels in England found the exact grade was right only about 43–55% of the time. Between 87% and 93% were within one grade, and most errors were optimistic. No equivalent study exists for Cambridge IGCSE, so each school should measure its own accuracy every year.
How accurate are teacher predictions?
Mostly within a grade, often not exact, and usually too high. Cambridge Assessment analysed forecast grades that schools sent to the exam board OCR in England:
| Study | Qualification | Exactly right | Too high (optimistic) | Too low | Within one grade |
|---|---|---|---|---|---|
| Gill & Rushton (2011) | OCR A Levels, June 2009 | 54.7% | 33.1% | 12.2% | 93.0% |
| Gill & Benton (2015) | OCR A Levels, June 2014 | 43.1% | 43.2% | 13.7% | 88.1% |
| Gill & Benton (2015) | OCR GCSEs, June 2014 | 43.6% | 42.4% | 14.0% | 87.3% |
Sources: Gill & Rushton, 2011; Gill & Benton, A Levels 2014; Gill & Benton, GCSEs 2014.
Accuracy also varied by grade. In the 2014 A Level study, forecasts were right 60% of the time for students who achieved A*, but 17% for those who achieved E. The authors concluded that “it was easier for teachers to correctly forecast higher grades than lower grades”.
UCAS predictions, made earlier in the year, look similar. A UCAS analysis of 2010 A Level predictions found that 42% were accurate and 48% were over-predicted by at least one grade (BIS/UCAS, 2013). UCAS says the gap between predicted and achieved A Level grades has widened over the past decade.
All of these studies use data from England. We found no published study of forecast accuracy for Cambridge IGCSE or Cambridge International A Level.
Are mocks better predictors than teacher judgement?
Not necessarily, and a mock’s accuracy depends on how it was built. A mock gives an accurate signal only if it uses a complete past series, is marked to the mark scheme and is converted with that series’ thresholds. Cambridge advises choosing components “with the same final digit” so published thresholds apply. Mock grades then need judgement about how much each student may change before the exam, which is where a school’s own accuracy data helps. Our guide to setting grade boundaries for mock exams covers the build.
How should a school measure its own prediction accuracy?
Every August, compare each prediction with the result, using three measures that Cambridge Assessment’s studies use:
- Exact accuracy: the percentage of predictions that matched the result.
- Within one grade: the percentage that were exact or one grade out.
- Direction: the share optimistic versus pessimistic, and the average difference.
Measure the average difference two ways. The “actual” difference keeps signs, so it shows whether the school over-predicts overall. The “absolute” difference ignores signs, and the researchers note it is “more useful when offers specify grades in particular subjects”. That matters for UCAS. A department whose predictions are one grade high for half its students and one grade low for the other half has an actual difference of zero, but it is still wrong for every student.
| Measure | What it tells you |
|---|---|
| Exact accuracy | How often predictions are right |
| Within one grade | How often predictions are close enough to trust |
| Actual (signed) difference | Whether the school leans optimistic or pessimistic |
| Absolute difference | How far off predictions are for individual students |
Break every measure down by subject, teacher and grade band. Low grades are where predictions are weakest.
What does UCAS offer schools?
A predicted-grades report, for some students. UCAS piloted reports with more than 1,000 schools in 2025 and “will now publish the reports annually directly to schools and colleges” to help them “benchmark their UCAS predicted grades”. Eligibility covers applicants domiciled in England, Wales and Northern Ireland. So most students at international schools are unlikely to be included, and those schools will need their own analysis.
How should the results change practice?
Use them to recalibrate, not to blame. If a department is consistently optimistic, its predictions should rest on more evidence, such as two full mocks rather than one, and on moderation by the head of department. UCAS asks that predictions be “based on demonstrable evidence of performance”. A school that knows its own accuracy can justify its predictions to universities and to the exam board. See the predicted grades policy framework.
How schools do this with AI Buddy
AI Buddy keeps every mock and practice result by student and topic, and its predicted-grade reports show the trend behind each prediction. After results, departments can compare predictions and mock grades with final grades to calculate their own accuracy. They can then see whether particular topics or grade bands were misjudged. The prediction itself remains the teacher’s professional judgement.
Frequently asked questions
How accurate are mock exams at predicting IGCSE results?
There is no published IGCSE study. Research on GCSE and A Level forecasts in England found about 43–55% exactly right and 87–93% within one grade, with most errors optimistic.
How accurate are predicted grades?
Studies of A Level predictions in England have found around 42–55% exactly right, with over-prediction more common than under-prediction.
How should a school measure prediction accuracy?
Compare predictions with results each year, calculating exact accuracy, within-one-grade accuracy, and the actual and absolute average differences, by subject and grade band.
Why are predictions usually too high?
The studies show optimism is the most common error, and accuracy is lowest for lower grades. The research does not establish the cause; schools should check their own pattern.
Discover how AI Buddy helps schools strengthen teaching, learning and evidence-informed school improvement. Or start a short consultation with our schools team using the form below — we will get back to you directly.
Sources
- Gill, T. & Rushton, N. (2011), The accuracy of forecast grades for OCR A levels, Cambridge Assessment Statistics Report 26
- Gill, T. & Benton, T. (2015), The accuracy of forecast grades for OCR A levels in June 2014 and for OCR GCSEs in June 2014
- BIS / UCAS (2013), Investigating the accuracy of predicted A level grades
- UCAS, Predicted grades: context and Predicted grade good practice