[pfortuny]

No: he does not say that statistics are useless. He says that the assumption of normality does not hold in real life: fat tails and the absence of means are ubiquitous. That is all the “black swan” means and how he got rich: taking risks which are nonsense under normal distributions but not so under, for instance, the Cauchy distribution.

[Animats]

Taleb's basic claim is that options that are way out of the money are underpriced. So he set up a fund, Empirica, which bought options a long way from the current value. This strategy does extremely well in years of a sudden downturn, and loses money in all other years. He and his investors did really well in 2008. He doesn't talk about the other years. Solid numbers for Empirica are really hard to find. Whether this is a win as an ongoing strategy is a big question. If you'd bought into that strategy in 2009, after it worked once, you'd have lost money for a decade.

[pfortuny]

Not trying to argue (wp is not a source) but it says he closed down in 2004?

[Animats]

Right, Empirica was the first "black swan" fund, which caught the crash of 2000. Universa was the next round, starting in 2007. That caught the 2008 crash.

As a strategy, this depends on a rough prediction of when the next crash will occur.

[repsilat]

> this depends on a rough prediction of when the next crash will occur

If you think the tails are mispriced (and you have some kind of alternative distribution), and markets are liquid, can't you just keep on making Kelly bets? Maybe you need to model third-party investors bailing too, though...

[pfortuny]

OK thanks for the info.

[spenuke]

Thanks. Ok, so is he creating strawmen everywhere, then? I know people working in government, biotech, climate research, and none of them seem to think that the normal distribution is the only model out there, or that the mean is a very meaningful (no pun intended) statistic.

[pfortuny]

The problem is: the normal assumption is just an assumption with a very interesting asymptotic property. That makes it good for computations.

However, Nature and more specifically economic events have little asymptotics in them (mainly isolated things like crises, earthquakes, crashes, accidents...).

I tend to agree with him on this.

He is not against science. He is mostly against economic modellers.