Even then, high impact very low likelihood is still often worth avoiding.

Maybe (impact * likelihood) - cost to mitigate