Exams regulator Ofqual expected its GCSE and A-level grading algorithm to produce “unreliable” results for some students, papers published today show.

The revelation comes in minutes from an emergency Ofqual board meeting held the week before A-level results day and this summer’s momentous grading U-turn.

Revealed: How the summer exams grading fiasco unfolded

Grading fiasco: How DfE bounced Ofqual into mocks plan

GCSEs: Autumn grades boosted by summer ‘generosity’

“The board were very concerned about the prospect of some students, in particular so-called outliers, being awarded unreliable results, but accepted reluctantly that there was no valid and defensible way to deal with this pre-results, and that it would therefore have to be addressed via the appeals route,” the minutes state.

In the emergency meeting on 4 August, it was noted that “the data distribution of a small cohort or entry is particularly subject to volatility, and that apparent inconsistencies between CAGs [centre-assesed grades] and historical data could be due to either real differences in student performance or to instability related to the size of the available historical data set”.

Ofqual said that the cause of inconsistencies between teacher-assessed grades and historical data would be impossible to identify from analysing the data on its own.

“Those causes are indistinguishable when viewing or analysing the data set alone and would require additional information to differentiate,” the minutes state.

The regulator thought at the time that smaller schools and colleges would be most impacted by the algorithm.

“A technical panel would be established to hear appeals which Ofqual would chair. Ofqual would support exam boards with appeal cases to ensure there was a consistent approach. It was highly likely that small centres would be most affected due to the shape of the distribution of anomalous grades and the size of entry in a subject in a centre,” the minutes add.