Evidence
Chess providers quote the same handful of numbers, and some of those numbers are wrong. Here is the whole picture instead: what the research found, what it failed to find, and the one finding this programme is built around. Every source is linked.
Part one
This part needs no study. It is simply a description of the game.
Every move begins with reading a position that has just changed. Nothing is highlighted for the child.
A plan has to survive the opponent doing something unexpected, or be replaced on the spot.
Chess builds a natural pause into every exchange. The child cannot act until it is their turn.
In a game being played out, each move changes the position, and the next decision starts from the position the child has just made.
In a taught session a child is regularly asked why, and has to put a reason into words.
Part two
931 primary pupils were split three ways: chess taught by instructors trained to teach problem-solving strategies, chess taught by school teachers without that training, and no chess at all. Only the first group improved. The group taught chess without the problem-solving instruction did no better than the children who never touched a chessboard.
This is the finding that decides whether a chess programme is worth a school's afternoon. Chess is not the active ingredient. The teaching is.
482 children in five Danish schools, first to third class. 323 of them had one of their four weekly maths lessons replaced with chess for around 30 lessons. Maths scores rose by 0.10 to 0.18 standard deviations, which the authors describe as roughly a third of a school year of extra learning by third class. Among the pupils who reported being most bored at school, the effect was 0.56. Among pupils who were never bored, there was no gain at all.
This was a quasi-experiment rather than a randomised trial, so it is weaker evidence than a trial would be. The pattern is worth knowing anyway: the children it helped were the ones who had already switched off.
Reported average effect sizes were 0.38 for maths, 0.34 for cognitive ability and 0.25 for literacy. The authors also estimated that around 25 to 30 hours of instruction is the minimum likely to produce a measurable effect, which is roughly one lesson a week across a school year.
The same authors flagged the central weakness themselves: almost none of the studies they reviewed used an active control group, so a placebo effect cannot be ruled out.
A randomised controlled trial across 100 schools in England with 4,009 Year 5 pupils reported an effect size of 0.01 on maths, equivalent to zero months of additional progress, one year after the programme ended. The same was true for reading and science. The Education Endowment Foundation rated the security of the finding as high.
This is the strongest single piece of evidence available and it is a null result. Any provider quoting the positive research without quoting this is giving you half the picture. Read alongside the 931 pupil study above, the two fit together: that programme was delivered by tutors whose teaching the evaluators went on to criticise, and it produced nothing.
Eight per cent of pupils reported that they did not like the chess lessons. Teachers and headteachers were generally positive about the programme and about its effect on pupils' skills and behaviour.
Enjoyment was measured. Attainment gains were not found. Those are two different things and are reported here as two different things.
The evaluators concluded that tutors themselves need strong class management and teaching skills, and that without them it was difficult to keep every child engaged. Improving tutors' teaching skills was named as one of two priority areas for improving the programme. What pupils liked least was tutors talking too much.
This is the finding that shaped how Little Chess Explorers is built. It is also the reason a chess qualification on its own is not enough for a primary classroom.
Part three
The Education Endowment Foundation report says half of pupils liked the lessons a lot, and that 8% did not like them. The 92% figure appears to come from subtracting 8 from 100, which is not what was measured. The real answer, half, is still a good answer.
Those are two separate pieces of research with different designs and different conclusions. The 24 studies belong to a 2016 meta-analysis. The 100 schools and 4,009 pupils belong to a 2016 randomised trial that found no effect. Adding them together makes the evidence look bigger and more positive than it is.
