Presumably you want them to go to that school so they can produce the best music possible with their abilities, for the benefit of society. The other option I see is that you might want them to go to that school because they have "earned it", but this is silly, especially considering a case where they haven't done anything, and are just naturally talented!
You now have two problems- first, one school might be excellent at training good musicians but not so great at training savants. Second...
Suppose you have one spot in a magical "savant school", which is able to develop somebody's skills better than anywhere else in the world. You'd want to assign the student who would benefit most to this spot- the one who has greatest potential.
This is NOT the student who currently writes the best music- this is the one who will write the best music after attending the school.
You don't care about ability now- you care about ability later. Predicting the latter from the former alone has an obvious flaw- training and practice improve ability.
Because of this, it's a good idea to consider measures of how much training somebody has had, in addition to their current ability, for admissions decisions.
Unfortunately, quantifying that is hard- so other metrics are used as proxies. In considering admission to an Olympic swimmer training program, for example, perhaps one might consider how early somebody learned to swim, or how often they visited a swimming pool.
No?
