[luckydata]

in the current state they are already plenty useful, I don't think it's worth proving mathematically that something can work 100% of the times when 80% is good enough.