The poorest pupils have made more progress than their better-off peers at just three per cent of secondary schools, new analysis shows.

About 100 schools posted a positive Progress 8 score for their pupils on free school meals, or who were on free school meals at some point in the last six year, which beat the positive score of more affluent pupils – and nearly a third are selective.

Researchers at Impetus-PEF, an education consultancy, said “very few schools” are ensuring these pupils, whom the government classes as ‘disadvantaged’, make enough progress to catch up with their peers.

Andy Ratcliffe, its chief executive, said the school system should want to “bottle up” the work of these schools and emulate it, pointing to “brilliant leadership” and “constant focus on success”.

But experts in value-added measures insist that Progress 8 is fundamentally flawed, and said the data could favour grammars.

Just 102 schools out of the 3,526 secondary schools for which data was available posted a Progress 8 score that was better for disadvantaged pupils than non-disadvantaged meal pupils, and in which both groups also scored above 0.

The pupils with top scores were those with rich relationships with staff

The figure, from government performance data released last week, is a slight drop on last year, when 107 schools improved the poorest pupils the most. Overall this year 687 schools posted positive progress scores for both disadvantaged and non-disadvantaged pupils, but the majority, bar the magic 102, still recorded better progress scores for wealthier pupils.

Headteachers have pointed to a culture of “high expectations” for achieving top scores.

Dixons Trinity Academy in Bradford came top for the progress of its poorest pupils, with a score of 1.35 for disadvantaged pupils compared with 1.16 for the rest.

Luke Sparkes, the executive principal, said the result was due to the “whole school being designed around our most vulnerable student”, with strategies such as morning and afternoon feedback sessions every day, and double-staffing core subjects so one teacher could always give dedicated support.

An “achievement-oriented culture” and a “razor-sharp focus on data” is also vital, he said.

His words were echoed by Ed Vainker, principal at Reach Academy Feltham in west London, which came third with a score of 1.24 for disadvantaged pupils compared with 0.95 for other pupils.

Both schools opened in 2012 and had their first GCSE cohort last year.

“Starting from scratch” allowed the school to focus on “powerful relationships” between teachers, pupils and their families. Home visits before the start of term and using teachers’ first names, as well as a team of counsellors available daily for mental health issues, also helped.

The pupils with top scores were “those with rich relationships with staff”, he said.

Progress 8 is a flawed measure because it relies on error-prone assessments of pupil attainment

A spokesperson for Harris Academy Chafford Hundred in east London, which posted a 0.99 score for disadvantaged pupils and 0.8 for non-disadvantaged school meal pupils, said the higher score was down to the school’s intervention strategies “tending to benefit” the poorest pupils most.

But the data also showed many grammar schools made better progress with the poorest pupils than their more affluent peers. Schools Week analysis revealed 27 per cent of the 102 schools are selective, and 29 per cent are single-sex.

Professor Stephen Gorard of Durham University, who has studied value-added measures, said it would be better to use a measure which showed the proportion of years a pupil has been eligible for free school meals.

He said “this measure of disadvantage recognises the chronic nature of poverty in some areas and some types of school.”

Meanwhile Progress 8 is a “flawed” measure because it relies on “error-prone” assessments of pupil attainment, and so tends to produce “unstable” results over time, he added. This is backed by Schools Week analysis which shows 15 schools cropped up in last year’s data set as this year’s. All other schools were new.

The data should be run for at least five years before the schools which work best with pupils on free school meals can be determined, said Gorard.