The dark art of predicting GCSE grades


It’s that time of year when I check the weather forecast before I leave the house with my dog because I don’t want to get soaked.  Generally the forecasts are accurate but if the forecast says it will rain at 11, sometimes it rains sooner and sometimes it rains later.  That’s the nature of the game – and whilst we aren’t happy about it, we accept that the Met Office have done the best that they can.  That’s the nature of a prediction.

I went in onto school last Wednesday to check my results (perks of my role).  This is only my second year of teaching GCSE science following a break of several years teaching BTEC (we only ever have one year 11 group).  You could argue that this year I had more data on which to base my predictions as I had the cohort last year as well.  My predictions were based on:

  • Mock exam result
  • Exam questions completed in class from ExamPro (so with A/C/G demand gradings)
  • ISA coursework
  • Classwork
  • Comparison to similar students last year
  • Gut feeling/aptitude

Prediction accuracy last year (numbers are percentages)


Prediction accuracy this year (numbers are percentages)



Unfortunately my predictions this year were on the high side so that my accuracy rate was lower.  I accept that I was a little optimistic for a couple of students but I had a much larger cohort this year due to Y10 and Y11 taking core together so I would have expected to at a similar level of confidence to last year.  I decided to do a bit of digging on the internet – it is surprising how little published information there is, and how many teaching professionals say they can’t predict with any accuracy.

This paper by Cambridge Assessment would suggest that teachers are better at predicting at the top end of the grade range than the bottom end. I was interested to read that accuracy is around 45% for maths and the sciences overall.  Unfortunately there is no data for special schools (we are a small group) but accuracy in the 20% range seems reasonable for students in the same ability range as my school.  This could help explain why the accuracy of my predictions dropped from last year (and would rate my predictions as better than many!)

It will be interesting to see how this picture changes with the move to 1-9 GCSEs as most teachers I speak to seem to be converting old-money GCSE grades into new rather than working natively in the 1-9 grades.  This will be compounded by the change in grade boundaries as more grades are introduced at the top end – perhaps the accuracy of predictions will flip with those grades in the 1-3 range being higher?


I’d be interested to hear how you predict your (GCSE) grades and how accurate your predictions were (you can get a matrix here if you want to crunch the numbers yourself)




Published by Rob Butler

Ex-science teacher, ex-school leader and full-time geek.