All of these have 10 different paths (in reality), any of which if you take the wrong one could end up in some kind of expensive dead end or spectacular failure.
You are zooming out, and when you look at an abstraction from a distance it's easy to miss the details and just boil it down to probablities.
The reality is that you measure and get information on all of these things to determine more accurately what odds you have. The act of measurement changes the probabilities, because if you do it right you will know more than you did before.
And sometimes you start in a place (or take directions into a place) which put you at such a disadvantage with competitors that you can’t ‘win’. It’s rare you can know that while it’s happening though.
In theory all these things are controllable and no problem is unsolvable. Sometimes it’s even true (post-facto however, more often than not). That view is often the best for succeeding (or more precisely, NOT having that view often puts one at a disadvantage when it comes to succeeding).
But the probabilities are often the more accurate model. That is often the best model for deciding if you want to take a bet, or for comforting yourself later if it doesn’t actually work out perfectly. It really doesn’t help though (most of the time) with actually doing successfully.
Even the Marxist analysis would say it’s about privilege, access, and exploitation. Not “probability”.
Considering well connected people with startup know-how and experience in the Bay Area, I would say they do pretty well with serious startup projects.