How well generated music matches the statistical distribution of real music across acoustic and structural features.