You're being very frequentist (http://www.statisticalengineering.com/frequentists_and_bayes...).

You don't have to have a random sample to estimate a probability. In the Bayesian interpretation, you can have a reasonable prior that you update as you collect more data. For example, a meteorologist can look at the clouds and "guesstimate" that there's a 50% chance of rain tomorrow. A frequentist would object and say that you can't take lots of random samples of tomorrow to estimate that probability. The Bayesian is willing to take the meteorologist's guess as a prior, and update it as & when more data becomes available. In this case, I think Sachin's statement is reasonable in the absence of any real data.

Look, you get one bite at the apple per idea/situation. A random sample isn't an appropriate counter-factual. Your job as an entrepreneur is to maximize your chances for success given the situation at hand.

I suppose "chances for success" wouldn't have set you off, and I acknowledge that. But we're at the point where we mock people for using "beg the question" incorrectly and ignore the substance of the argument, which isn't really productive for others reading this thread.

The only people I have heard that say this, travelling around the world, meeting multiple 9+ digit exit entrepreneurs and tons of 7+, are the ones that started in SV.

I tend to think based on what I read that there are a lot more large exits inside SV than outside. But, there seem to be a lot more early deaths and me-too companies there as well.

Are you optimising for the wrong solution (insane exits)?

[edit: early deaths = companies, not people]

You've got a point, in the sense that a lot of the "probabilities" are just best guesses based on anecdotal evidence. But pragmatically, I don't see much of a choice; it would be paralyzing to try and gather all that data before making a decision. And, in my mind, it's usually better to make that decision on imperfect information- like the aphorism that "Your probability of success is higher in the Valley"- than no information at all.

I refer to the book: how to measure anything. Measurement is the reduction in uncertainty, not paralysis by analysis. Often you can get great reductions in uncertainty with simple methods if you think for a few minutes.

Some examples that the book may help you with could be looking at the governmental statistics for new businesses by geographical location (e.g. if they are better, then this must show in turnover right?). Geographical distributions of the last 10 years Fast 50/100, breaking the groups up by funding type.

Are you so certain that the probability of success is higher in the valley? It seems to me that the majority of the people who say this have something to gain, directly or indirectly from this being promoted as fact. This is usually an indicator of a falsehood.

That looks like it might be a worthwhile book. Thanks for the recommendation.

I'm not "so certain that the probability of success is higher in the valley," but there are two main reasons I think there might be something to it: firstly the fact that people whose opinions I respect are saying it, and secondly the anecdotal-evidence-based but sensible conventional wisdom about meeting more relevant people (whether for funding, employees, advisors, bouncing around ideas, etc.) in the valley.

I haven't noticed as much that "the majority of the people who say this have something to gain, directly or indirectly from this being promoted as fact," but maybe I haven't been paying enough attention. :-)

I am a full time skeptic and tend to notice these things :)