Would they describe highly skewed distributions as random?
Natural selection would prbably mean that the number of reproducing offspring per parent would follow the
Poisson distribution.
This can be very highly skewed.
ETA: picture added
You can make simple assumptions and calculations, even at this level of abstraction (hey I am an engineer, this is what I do)*.
In a stable population, the average number of reproducing offspring per parent would be one (i.e. lambda would be one).
An advantageous trait might have a lambda of 1.1, i.e. a 10% advantage compared to the population without that trait.
If there are 1 million offspring/parent (e.g. from a fish spawing) then this 10% advantage translates into a raised chance of reproduction for an individual fry of 10% to 1.1 per million. Still not good odds, but if you have 1 million fish, all with this trait...
*Am I right that Meadmaker, mijo, and Schneibester also have an engineering/physical sciences background too, and is this significant?
ETA:
I am thinking that as rough estimations, I try to state my assumptions, and work out fthe implications form these to get a ball-park figure. These rough calculations would involve calculations. If I would try to do this with natural selection, and
don't assume something akin to the Poisson distribution, the numbers are obviously nonsensical, populations would rocket, and there would only be three states: disadvantage, neutral and advantage. The concept of a
slight advantage
requires a probabilistic treatment of natural selection.
In artificial selection there is no such thing as a slight advantage, one either trys to breed as many offspring from an organism as possible or none. (There is still natural selection limiting this too, of course).