[imgabe]

I'm not saying it's all luck, but if you have a million people rolling dice, some percentage of them will have a lucky streak doing "better than rolling dice would suggest". A random outcome doesn't mean every individual gets the average result every time.

[rmrfrmrf]

Is this actually true? I know it's definitely not true if a million people are rolling dice forever -- each die should land on 1-6 with equal frequency.

Given a limited number of rolls, though; say, 20 rolls per person; is there actually a consistent, measurable percentage of people that will outperform others? Somehow I doubt it.

[imgabe]

Let's say you have a million people. They each roll a 6 sided die. The odds of getting any one number are 1/6. The odds of getting any given number 5 times in a row are 1/6^5 or about .0001286. Out of your million people, 128 of them will probably roll e.g. 3 five times in a row. It doesn't mean they have a particular talent for rolling 3s. If you get the million people to roll 5 more times, probably a different set of a million people will roll all 3s.

I'm not saying there's no skill involved. There are certainly things you can do to guarantee failure. But if someone has 5 successful startups in a row, it's not easy to distinguish them from someone who rolled 5 3s in a row. The population of people who've done startups is pretty small, relatively speaking.

[voronoff]

This is reminding very much of the debate around the Efficient Market Hypothesis, as to whether traders can be systematically good vs. just repeatedly lucky. There is a lot of lit on this issue.

[sdenton4]

Go read 'Fooled by Randomness.' It makes the case that your wall street pundits are EXACTLY that measurable, consistent percentage of the rather large population of traders who haven't blown up after five years. Thus, every year, half+ of the pundits blow up themselves, never to be heard from again.

[eterm]

If just 1000 people all each toss a coin ten times you'd expect there to be a person who landed ten heads.

Still just luck though.