The main problem with the OP is that it was looking at broad averages while necessarily venture capital and ambitious start ups need to be quite exceptional. T

The main problem with the OP is that it was looking at broad averages while necessarily venture capital and ambitious start ups need to be quite exceptional. The broad averages say next to nothing about the exceptional cases. In particular, the probability of being exceptional is just irrelevant. Instead, what is just crucial is the conditional probability of being exceptional given the details of the specific project. So, it's just crucial to consider the details of a particular project. A sad part about venture capital is that it is looking for exceptional projects, necessarily ones that will in effect push forward the boundary of computing in our civilization, and doing so with in nearly all cases very poor qualifications in doing or evaluating such projects. And venture capital has another problem: Too few projects on the other side of the table have much potential of such pushing the boundary. For all concerned, only a tiny fraction of the venture general partners have any hope at all of evaluating projects that push the boundary. We should be making much faster progress.