However, although these approaches often claim to avoid using probability distributions, the way that possible futures are sampled has a strong bearing on estimates of performance, trade-offs, and robustness.

Choices about how to sample possible futures are often unrealistic (eg, all values in some range are equally likely) or hard to identify. Of course, any assumptions will be wrong — we don’t have crystal balls! — but bad assumptions that are transparent are better than obscure ones

· · Metatext · 1 · 0 · 2