Nassim Taleb: We should retire the notion of standard deviation

What is hilarious is that I have been scratching my head over the exact opposite phenomenon!

I actually love Taleb's books (Black Swan, Antifragile). And I work with "data scientists", most with Ph.D.s, many in statistics.

I always talk about his book, but I cannot for the life of me find a single person who's even read it. I think: these are some the ONLY book about stats on the NY Times best seller lists (up until Nate Silver). And yet all these professionals have not only not read it, but barely even heard of it?

My hypothesis was that it is more popular on the East Coast than the West Coast (of the US). Are you from either of those places? I am originally from the east coast but work on the west coast. To hand wave a bit, I feel like west coast people are less into "ideas" and more into actions and experiences. Taleb's ideas do have somewhat of a nytimes-ish new yorker-ish east coast culture flavor. And a lot of people working in Silicon Valley are not really that interested in philosophy.

On the subject of his writing, I can totally see that people can be turned off by his writing. He can be arrogant and insulting. I find it kind of funny, but that's a matter of taste.

A few things I remember from his books that I really liked:

- The story of Nobel prize winner Myron Scholes, and namesake of the Black-Scholes equation, which I learned about in computational finance in college. He started a company "Long Term Capital Management", to monetize this ideas, and promptly lost billions of dollars. http://en.wikipedia.org/wiki/Myron_Scholes

That's not interesting? I think the difference between theory and practice is intensely interesting, and Taleb has a lot to say about it.

- I largely agree with his philosophy that people who claim to know things cause more harm than good. The downfall of Alan Greenspan and Bernanke is their arrogance. They think they can control the economy. But they can't and caused millions of people real harm.

- Respect for the old. For all its virtues, Silicon Valley does have a severe case of "neomania". Taleb's ideas about things that last apply to software too. Unix and C are going to be around a lot longer than say Hadoop or Puppet/Chef.

- The philosophy of fragility also applies to software in a straightforward way. Most people know this now, but you should continually expose your software to users and the market, not build up grand ideas in your head.

- actions over knowledge, i.e. people who know how to do things but not explain them

I could list a half a dozen more important ideas but I'll stop there.

I did write in my comments about this book that he overreached on his "trilogy" idea for the Antifragile. But I do like how he draws together a lot of seemingly disparate ideas that are philosophically related.

In fact, most of Taleb's ideas are about how to not rely on predictions of events that are unpredictable - so faulting him for predicting or not predicting this and that means one does not really understands what Taleb is talking about.

Exactly. I'm astounded that this basic fact - that financial modelers have been using the wrong distribution for decades IS NOT BEING DENOUNCED. For example, the Capital Asset Pricing Model (CAPM) which assumes that returns fall on a normal distribution.

>>> I am pretty sure that if his stuff were posted without his name the conclusion would be obvious: it's just nonsense and very offensive at that.

This is not true, at least for me. When I first read Taleb's ideas, I had no idea who he was and what is his credentials, but his ideas were interesting enough and challenging enough to look him up and find out what else he wrote. Yes, his style is abrasive and sometimes outright combative. So what. His ideas are interesting, that's what matters. I'd rather have an abrasive person who has interesting ideas than a polite one who has nothing.

And speaking of Python vs. Java things, I'm reading those kind of things literally every day, frequently including here on HN. If idiot badges were distributed for each such occasions, there would be a very sizeable majority wearing them. But I am not convinced calling each other idiots is actually adds anything to the discussion.

The true case against the standard deviation is more or less impossible to make without going into the math. For a mathematically advanced Bayesian, the idea is straightforward if you've already been through the lawful derivation of Gaussians from their underlying assumptions: Using standard deviations in your thinking makes sense if you think that an increasingly large error has a decreasing log-probability that goes as the square of the size of that error. The distribution that you get from the sum of independent variables is the classic example. Basically, if you think in standard deviations, then all postulates of the form "The error was that size" will come with probability penalties that you trade off against each other in proportion to the square of the size of that error. Thinking in standard deviations similarly corresponds to models where exceptional events become unlikely in proportion to the square of the size of that exception, only by "unlikely" you have to read log probability rather than probability, and log probability is (always) the right way for Bayesians to think about it anyway for reasons I won't go into.

And then the critique of "standard deviation" is that people got taught SD as a statistical tool that you just pull out any time you feel like it, and they don't know what it means in underlying probability-theoretic terms as an assumption about either the world or our own uncertainty, and so it's misused horribly on all sorts of occasions.

I'd guess that SDs are appropriate 40-90% of the time depending on which field you work in, but without a lot of Bayesian background with fairly advanced math people will not be able to really appreciate what the other 60-10% of times look like. And the state of education is not anywhere near like that. It's just people being taught to calculate the standard deviation cause, like, that's something you do. They don't know what assumptions it corresponds to even in the cases where SD does apply.

Burning SDs to the ground and starting over would not be very much amiss in one of the fields where SDs only make sense 40% of the time, but the practitioners are using them all the time. (Machine learning is one of the fields where SDs make sense maybe 40% of the time, and if I found an ML practitioner who had been taught to think in terms of squared error I'd send them off to learn the underlying probability theory until their entire worldview had been translated into likelihood functions instead.)