In his book Thinking Fast and Slow he talks about the way we believe things.
When encountering an idea our system 1 believes it automatically and than our system 2 considers the consequences of believing it and decides if you should UnBelieve it.
They found that when system 2 is tied up or exhausted you are more likely to believe something thats false because your system 2 didnt consider unbelieving it. This is why late night commericials and infomercials worked so well.
The danger you describe is in a case of uncertainty having confidence...Daniel Kahneman is talking about a case where you could know forsure but because of a cognitive bias you are overconfident without relying on statistical evidence which may be to the contrary.
And in fact, I agree with that rationale. I still can't believe I went through with it, much less succeeded.
But without foolhardy people to take on these gigantic risks, we'd be FAR less developed than we are.
Statistics cant prove anything! Thats his point. Dont be so confident that your mind is too closed to explore other possibilities.
Scientific method can only eliminate possibilities and show you the most probable option. It is by no means proof. But thats a bad reason to not consider competing hypothesis instead of just trying to confirm an existing bias.
Your point about incubator success is different. Those statistics are exactly the ones we want to avoid. Success or failure of a startup in an incubator is not diagnostic evidence as i can think of a dozen other hypothesis to explain those results. Its like a doctor saying you have strep because you have fever. Since fever is consistent with other diagnosis it has no diagnostic value beyond the point of saying you are sick.
In other situations it might be more useful to be able to make less biased guesses about outcomes, about starting wars or investing or stopping climate change. It's like you can't stop bluffing in a high stakes game even when you've lost your credibility.