EEG is the usual way to read brain activity without surgery, and decoding motor imagery from it, the intention to move rather than the movement, is the classic brain-computer interface task. It is also stubbornly unreliable, because signals differ between people and even within the same person from one session to the next.

The field keeps producing claims that one decoding pipeline is broadly better, usually a spatial or Riemannian method. This work tests the weakest form of that claim under conditions as favourable as it could ask for, using the MOABB benchmark across 1,056 decoding configurations built from combinations of feature extractor, scaler and classifier, over more than 340,000 subject-level results.

The title carries the finding: average rankings mask per-subject optimality. A method can top the table overall while being the wrong choice for a great many individuals, in the same way an average shoe size fits nobody in particular. For a technology worn by one person at a time, that distinction is the whole game.