[alexpetralia]

It might be useful to think of "luck" as an error term on a regression model, where the covariates are well-known factors (e.g. wealth, ambition, connections) and the error term is everything else.

Given that plenty had privilege and money (i.e. those are the covariates), and many didn't succeed, we'd expect those weren't significant variables.

Many likely had drive, connections, know-how as well. Many of them probably failed too. Again, insignificant covariates.

So what does that leave? The error term: luck.

In other words, during a tech boom like at the turn of the century, there are so many winners (often outsized, which distort many basic statistical assumptions, such as a normal distribution) that is almost futile to try and identify significant covariates.

To speculate here, often with the pretense of certainty, is more often reflective of post-hoc reasoning than actual science.

That, in my opinion, is why "luck" is an adequate answer. We can be fairly sure that certain covariates contribute to success over the long-term in life (i.e. we have a large sample size of being alive, and we can extrapolate from many other people who have lived). It is far harder to do this with nonce hype cycles (infrequent, low-sample size).

Chalking up more success to the error term, "luck", seems perfectly appropriate during such unusual times.

[dang]

If I understand you correctly, then "luck" is another word for "we don't know". That's reasonable, but then we shouldn't make strong claims to knowing.

There's a cruder version of this argument, according to which all success is luck. I hear that, or things that sound like it, a lot, but it's too simplistic and usually too self-serving to be plausible. Though many successful people will be the first to tell you they were lucky. (I always remember https://news.ycombinator.com/item?id=1621845)

[alexpetralia]

I think that's a fair opinion.

The only tweak I'd make is really around sample sizes and statistical significance.

All success _may_ be due to luck (the error term), but we can be fairly confident (say p < .001) it isn't. Said otherwise, we can be reasonably sure that some covariates (e.g. hard work, wealth, education) are significant, given a large enough sample size and we define what qualifies as success. One of course could quibble about these - do we really have enough statistical significance for each trait? - but the fact is that every person everyday operates with some intuitive understanding that these things matter. In other words, not all outcomes are due to luck.

This is much harder to do with small sample sizes or unusual occurrences. Statistics is based on frequencies, and if we have low frequencies (such as a tech boom), we should be much less confident in the significance of each variable.

This is, presumably, why people intuitively chalk up much success during these times as survivorship bias. It's not that it certainly, absolutely is; rather, it's that we're far less confident on which variables are significant and which aren't. The error term remains.

Again, attributing it to the error term, implies mostly that we shouldn't have too much confidence in our speculations on which traits are significant during such one-off events.