[orijing]

Thanks for the correction. That link worked when I checked...

Well, Chebyshev inequality always applies. It just might not be the tightest bound possible. Unless we know the distribution better, this is the best estimate (tightest bound that's provably correct given the single assumption about standard deviation)