Many personal achievements in life are a "complete mystery" to many or most observers, or to science, or to mainstream intellectual discourse, but they still draw on rare and special qualities of individuals. Even if you don't "control" the outcome, that doesn't mean it's not a function of who you are (in the broadest sense).
However, you could have just created a successful startup. You could create an incredible product. You can still fail if Google launches a competing product at the same time before yours can gain traction. You can still fail if someone else comes out with a better and cheaper way to do what you're doing. You can still fail if a key part of your marketing fails you for reasons out of your control (Facebook blocking access to their API).
It's simply a function of the amount of chaos. Nobody can tell you if your startup will succeed or fail. I'm very sure that a special forces instructor will be able to tell if you'll make it after just a few minutes of watching you perform.
Leadership and team-building abilities are a significant success factor due to peer evaluations. You could be a 0.1%'er in physical ability and still get peered out because no one trusts you or wants you in their squad. Like a startup, your ability to build relationships is key to success.
So yes, there's a high degree of chaos. This means to succeed you must be all the more resourceful and dogged and/or experienced. Maybe these things are harder to quantify than predictors of success in the special forces. I don't know anything about the special forces. The wording used above just strikes me as true for start-ups rather than being a contrast.
* Getting demotivated by statistics is the worst possible use of time and energy. It means nothing that the average reader of this comment has various qualities and abilities that, on average, are about average for a reader of this comment. Every reader is an individual, not an average. Start-ups, like people, and marriages, are so different that you just can't generalize across them in very illuminating ways. This is the same as saying there are many subtypes, and at best you could draw comparisons or generalizations within a subtype. Which is the same as saying if you scope your generalizations small enough and choose your points of comparison wisely, you have a chance of capturing a meaningful pattern that might actually tell you something about an individual specimen. Or you might just be over-analyzing.
* It's possible to create causality and attribute events to your actions, even retroactively -- it's called taking responsibility. You can even take responsibility for things you didn't know about, like an executive or official who resigns over misconduct below him. There is very little we have direct control over -- the positions of our limbs, for example, but only when we are paying attention to them -- but we are expected to exercise whatever indirect control we have. People in high positions of power and responsibility have very indirect control -- hiring the right people, putting policies and checks in place -- but they must accept this responsibility and milk it for all it's worth. When you take responsibility for something, you choose to account for (recognize, pay attention to) the causality between you and that thing. We exist in a vast interconnected web of causality, so it's all what you choose to focus on.
This isn't about disposition vs. circumstances, either. Your disposition and circumstances are both different from other people's, as are your goals, motivations, and strategies. It's about how much weight you should give a statistic about other people when trying to determine something about yourself. As you learn more about yourself and the world, and especially if you pursue areas where you are strong, statistics become meaningless.
If you're making decisions, evidence-based decision making can help you decide between different courses of action (A/B testing is an extremely simple example of statistics in action). Statistics can even give you some information about which factors can impact the likelihood of different outcomes. Maybe you think factor Y about your current situation (or your personality, or your product, or whatever) makes some outcome particularly likely. Is that actually true? Sometimes the honest answer is we don't have any way of knowing, because it's a unique situation. But I think people jump to that conclusion far too quickly. Often there is actually considerable information available, if you want to use it. I'd rather use existing data than cultivate some kind of "everything is my decision and I'm in control" mindset, when that would just be delusion in many cases: many things are beyond my control, but some are within it, and I think making effective decisions requires figuring out which is which.
If a certain statistical interpretation is simplistic and naive, the response I'd like to see is a more sophisticated but still evidence-based interpretation. I don't think that "I'm a unique snowflake and I don't need evidence or science because of GRIT" is a good substitute.
When life gets real, there is no comfort in figures. Real (and wonderful) commitments such as marriage and a challenging work environment have taught me that. And the more you excel, the more unusual you become and the more statistics will predict your failure.