Richard Murphy and Felix Weinhardt:

Conventional wisdom suggests that it is always best to place children with higher-performing peers.
Our research, which looks at their later outcomes, indicates that this is
not necessarily true.
Imagine two pupils of the same high ability: one is top of their class but the other is in the middle because their school attracts many high-ability children.
We find that the pupil who was top of the class becomes more confident and performs better in secondary school than the pupil who had the same test score in primary school but a lower rank.
These rankings are inferred by the pupils themselves as it is not standard practice for teachers to discuss rankings. We find that being highly ranked during primary school has a positive effect on later test scores that is equivalent to being taught by a highly effective teacher for one year. And being ranked in the top quarter of your primary school peers as opposed to the bottom quarter improves later test scores by twice as much as being taught by a highly effective teacher for one year.