A 2011 dissertation conducted in Italy, involving over 1750 children from 30 schools. Each school was randomly allocated a control class and test class, which received a 30-hour chess course. The test classes showed a statistically significant improvement in maths scores, but schools were not randomly selected. Read more here.