[marcus_holmes]

We do have some scientific insight because we have several control groups, aka accelerators.

If it wasn't mostly luck, accelerators would have an improved success rate, because they are selected by experienced people based on judgement calls about their likelihood of succeeding [0].

Accelerators actually have, at best, a very minor improvement on the base success rate.

Therefore we can safely conclude that startup success is mostly luck.

[0] Of course, this is using a definition of success that is more than just getting investment. I would define it as "becoming a sustainable business". But that would exclude things like Uber and Twitter that are not sustainable (yet), but are undeniably successful.