That would be consistent for a consistent input though? (I'll admit the maths is beyond me)

Random wouldn't have that 'advantage'.

I'd go for a consistently bad algorithm over a randomly bad algorithm, although that is subjective.