A key purpose of the original paper is to inform policy to improve public funding of research [1]. Thus, to a first-order approximation, it should reflect realities. However, ... The number of lucky and unlucky events in the paper (figure 5, page 8) are each around 5-7 events per agent on average, where 'lucky' is defined as doubling the capital (or success) accumulated so far and 'unlucky' as halving it. In the real…
A significant event every half decade or so does not seem unrealistically high to me.
> that Log-Normal Distribution could fit data on real-world success as well
Power laws (eg Pareto distribution) and log-normal are extremely different though in the tails. The log of a log-normally distributed RV is normally distributed, and thus effectively restricted to a fairly small support area (+/- 5 std devs), as the tail falls off extremely fast (exponentially). The log of a Pareto distributed RV is exponentially distributed, and can get pretty high (the tail of the pdf decays only with some power, ie polynomially).
So, I'd expect the properties and conclusion to be quite different (more extreme wealth under power law/Pareto than under lognormal).
> Research has shown that Log-Normal Distribution could fit data on real-world success as well as Power Law and sometimes better. [4]
This surprised me, and I don't think the paper [4] supports your claim. It examines only the super rich (Forbes Billionaires), not the entire population. Log-normal might model a few hundred rich people reasonably well, but if you fit to the entire population, I'd think log-normal is basically ruled out by the rich (while I'd expect Pareto still to sort of work).