Score on lowest performing subjects

It could be interesting to add a metric (or function to compute) the score on the n% lowest performing subjects.
In paradigms where we can get nearly perfect scores such as ERP, SSVEP or cVEP, the lowest performing subjects are where improvements can/should be made.

The F1@20% metric (F1 score of the worse 20% subjects) was used for example in [Gnassounou 2025](http://arxiv.org/abs/2503.04582).