All risk-mitigation strategies are actually variance-reduction strategies.^1
You can reduce the variance of a gaussian process without changing its mean results. If you’re working with a power law distribution or something similar, the majority of the results come from low-probability results in the tail of the distribution. If you mitigate risk by reducing the variance of the distribution, you will inevitably squash potential outlier positive results.
This is uncomfortable because people want things to be predictable (low-variance) especially things that they are putting money into.
^1: Is this true?