natural selection seems largely consistent with how other physical laws operate (and therefore we can model genetic evolution by using physical processes such as random walks, brownian motion, etc).

This is not a bad assumption.

But, by the same token, there's no reason to think that evolution will produce the same outcome everytime. Even if you have selection forces on a randomly moving particle the path and therefore outcome, will not be the same every time.

Is this guy's paper pretty much universally refuted by LWers?