A note on the expected minimum error probability in equientropic channels