30 Wilmott magazine

t was unfair really, but perhaps not alto-gether unexpected. On a late October day in 2001, the English weather unwilling to commit to any-

thing discernibly seasonal, Nassim Nicholas Taleb walks into the relative comfort of the Cambridge Union bar

I and informs those nearest to the door that the proposi-tion for debate just finished should have been “This house believes that most city hot-shots are lucky fools.” Taleb had been expected to argue from the less nuanced posi-tion that all city hotshots are lucky fools. He deserved his drink.

Wilmott magazine 31

TALES OF THE UNEXPECTED

thing about pure forms. As a lover of wisdom, a philosopher, one should turn away from the distraction that is the empirical world. In the empirical world we only see particu-lar cases of a universal, the beauty of a woman or that of a statue of a woman are instances of a universal beauty, our reference in physical aes-thetics. We have an understanding of ‘beauty’ prior to our reference to it in a particular case. Furthermore we are able to qualify comparatively between two objects we find beauti-ful – the Platonic argument is that things in the physical world are all compounds of various qualities – which do not in themselves exhibit perfect beauty but tend towards an ideal beauty to some degree. Therefore, since we have a sense of what ‘beauty’ is prior to applying it to a particular object, and we are able to discern between the levels of beauty inherent in two objects then there must be an a priori beauty which we have an understanding of in order to make that aesthetic judg-ment in the particular case. In order to make a classification there must be a universal class under which

these objects would be subsumed. The problem is that we are all

Platonists, we automatically look towards an ideal state, one that makes sense (always in hindsight), one that ignores the aberrations, the one that can stand up with us when we sing the national anthem of Mediocristan.

Mediocristan, Taleb explains, is the land of the bell curve. That is, no doubt, what they have embossed on their license plates there. This place is a ‘province dominated by the mediocre, with few extreme success-es or failures. No single observation can meaningfully affect the aggre-gate.’ So human height is a perfect example, because out of a sample of 1,000 people, even if one person were twenty feet tall it would make no impact whatsoever on that com-forting undulation. Randomness here is mild.

Leptokristan displays a wild randomness, ‘a province where the total is impacted by a single observation.’ So here you will find the distribution curve for, let’s say, personal wealth of 1,000 people, and one of them happens to be Bill Gates.

The problem is that we are all Platonists, we automatically look towards an ideal state, one that makes sense (always in hindsight)

worth discussing any further. And the final count? Abstention was the order of the day; a protest against globalization. Go figure.

Taleb takes a hammer to the lib-erty bell. This is an enjoyable specta-tor sport. He desecrates the bell curve and the entire business of statistical inference; the sanctum sanctorum of the priesthood of modern orthodox finance. Who do we see turning the tables outside the temple? Your understanding of randomness is wrong! The Gaussian only fits with non-scalable, usually physical phe-nomena. And it goes deeper than that. The reverberation from that big ding-dong shakes us to the founda-tions of modern finance.

In his latest essay, The Black Swan, Taleb takes us on a trip to the lands of Mediocristan and Leptokristan, but at the heart of darkness he finds Plato, or at least the band of Dennis Hoppers he inspired.

You will recall from your brush with philosophy that Plato had a

Misunderstood then, misunder-stood now. Wall Street’s principal dissident. Heretic! Calvin to finance’s Catholic Church. His second book, the treatise Fooled By Randomness, drew praise and derision in equal quantities. ‘Taleb is a breath of fresh air and tells it as it is!’ ‘Taleb either states the obvious or makes gener-alizations to the point of ludicrous-ness!’ Either way, people were quite breathless.

The book is accessible. What a double-edged sword that is. It brings with it the consequence that anyone literate in any of the 17 languages it has been published in feels able to formulate an objective opinion on it. Because it’s all about dumb luck, isn’t it? It leaves you with the conclu-sion that modern finance has gone long on snake oil. Boo hoo! The other conclusion that we draw is that it’s all just so much noise. Calm down!

So, the Cambridge Union invites Taleb to debate, given the exact word-ing of the proposition it’s barely

Taleb (with Espen Haug, and Benoit Mandelbrot) on Mandelbrot: “I will fight to the end to promote his method as a framework to look at deviations”

32 Wilmott magazine

TALES OF THE UNEXPECTED

There is scalability here, and the ride can be much, much more volatile. Deviations from the mean are not exponential.

Taleb introduces us to the Ludic Fallacy, he asks us to stop talking about casinos and games when we want to discuss uncertainty. Why? Well, more of that in the interview, but suffice to say it’s because the house always wins. “The Leptokristan /Bell Curve gap is a simple application of this Ludic Fallacy.” Taleb says, but don’t get him wrong.“My problem is noncom-putable odds, whether bell-curve or scalable. It is just that I prefer the

scalable as a statistical paradigm.”The epistemological ground-

ing of finance the ory is unsound. Because just as men are from Mars and women are from Venus finance is from Leptokristan and mundane matters such as distributions of height, weight and calorie con-sumption are from Mediokristan. In a way there are a lot of people on Wall Street or in the city of London who go to work every day in drag. You have often read the following words: “Assuming a Black-Scholes world …” We believe we operate in Mediocristan but in fact it is the wild shores of Leptokristan upon

which we have been washed up. If we lived in a Black-Scholes world, LTCM would not have happened.

This was a viewpoint that many experienced traders did not find difficult to accept when Taleb first developed it in Fooled by Randomness. The Black Swan is not about finance, it is about the philosophy of history. But it carries the same warning. Wherever you encounter people saying something is statistically sig-nificant, it’s a sure bet that they have looked at their observation errors and assumed the big bouncy bell curve is in play. But the way that we interpret those samples? We ignore

the unseen, the unknown and the unexpected. Recall Taleb at the end of Fooled By Randomness, watching Odysseus battle the pull of the sirens song, and taking away the lesson that one should not attempt to emulate the man tied to the mast, but accept that at best we could hope to plug our ears with wax and live with randomitis. Here in the Black Swan we are at sea once more, Diagoras, when shown the offerings made to the gods by those who survived their sea voyages asked, what about the offerings made by those now in the bosom of the briny deep?

Platonicity GlossaryNassim Nicholas Taleb provides a brief guide to his key themes

Platonicity: the focus on those pure, well-defined, and easily discernible, objects, like triangles, or more social notions, like friendship or love – at the cost of ignoring those objects of seemingly messier and less tractable structures. Nerd knowledge: the belief and custom that what can-not be Platonized and studied does not exist at all, or is not worth considering. There even exists a form of skepti-cism practiced by the nerd.Ludic fallacy (or uncertainty of the nerd): Manifestation of the Platonic fallacy in the study of uncer-tainty; basing studies of chance on the narrow world of games and dice. Aplatonic randomness has an additional layer of uncertainty concerning the rules.Epistemic arrogance: Take a measure of the difference between what someone actually knows and how much he thinks he knows. An excess will imply arrogance, a deficit humility. An epistemocrat is someone of epistemic humil-ity, who holds his own knowledge in greatest suspicion.Epistemic libertarian: Someone (like myself) who con-siders that knowledge is subjected to strict rules, but not institutional authority as the interests of organized knowl-edge is self-perpetuation, not necessarily truth (just like governments). Narrative fallacy: our need to fit a story, or pattern to

series of connected or disconnected facts. The statistical application is data mining.Narrative discipline: discipline that consists in fitting a convincing and well-sounding story to the past. Opposed to experimental discipline.Platonic confirmation: You look for instances that con-firm your construction (or model) –and find them. Platonic fold: The place where our Platonic representa-tion enters in contact with reality and you can see the side-effect of models. Scandal of prediction: the poor prediction record in some forecasting entities (particularly narrative disci-plines) mixed with verbose commentary and lack of aware-ness of their own dire past record.Epistemic opacity: Randomness is the result of incom-plete information at some layer. It is functionally indistin-guishable from “true” or “physical” randomness.Black-Swan blindness: underestimation of the role of the Black Swan, and occasional overestimation of some specific one.Round-trip fallacy: the confusion of absence of evi-dence of Black Swans (or something else) for evidence of absence of Black Swans (or something else). It affects statisticians and other people who lost part of their rea-soning by solving too many equations.Fallacy of silent evidence looking at history, we do not see the full story, only the rosier parts of the process.Future blindness: Our natural inability to take into account the properties of the future –like autism does not allow the patient to take into account the existence of the

minds of others.Circularity of Statistics (the statistical regress Argument): We need data to discover a probability dis-tribution. How do we know if we have enough data? From the probability distribution. If it is a Gaussian, then a few points will suffice. How do you know it is a Gaussian? from the data. So we need the data to tell us what is the prob-ability distribution, and a probability distribution to tell us how much data we need. This causes a severe regress argument.Scorn of the abstract: Favoring contextualized thinking over more abstract matters. “The death of one child is a tragedy; the death of a million is a statistic”.Retrospective distortion: Examining past events with-out adjusting for the forward passage of time. leads to illu-sion of posterior predictability.Mediocristan: province dominated by the mediocre, with few extreme successes or failures. No single observation can meaningfully affect the aggregate.Leptokristan: province where the total is impacted by a single observation.Black Swan ethical problem: asymmetry between the rewards of those who prevent compared to those who cure.Problem of induction: Logical-philosophical extension of the Black Swan ProblemBell curve (Gaussian): or GIF, great intellectual fraud. Application of the ludic fallacy to randomness. There is a qualitative difference between Gaussians and scalable laws, much like gas and water.

Wilmott magazine 33

34 Wilmott magazine

TALES OF THE UNEXPECTED

Nassim Nicholas Taleb: He desecrates the bell curve and the entire business of statistical inference

TALES OF THE UNEXPECTED

I just finished the first draft of the Black Swan – spent a year hiding in cafés and libraries working on it. I wanted to get closer to my idea of a pure epis-

temologist, get rid of what trading was still hampering me, going deep-er into matters, as deep as I could. But unfortunately I am now back... (laughing).

How did you use quant finance as a base towards philosophy?NNT Quant finance proved to be a wonderful place to learn the stuff, a place to generalize because you can see the explosive “Platonic fold”, where reality enters in contact with what mental representation we have of it, where things are uncharted but people are deceiving themselves with constructions and calling them science. It is better to start with markets than the other way around. Many people study, say, physics, mathematics, probability, or psy-chology to try to understand finance. Then they show you that superim-posing these ideas works beauti-fully. A few might get rich for some random reason, and then we impart their success to their understanding of psychology or socio-politics. So people come to finance blinded by some idea, looking for – and finding – confirmation. I think that because of the messy nature of knowledge we have in finance, it is much prefer-able to go the other way around. We learn to think in an aplatonic way

– away from what I call Platonicity.Now I climbed up to the mother

of all the problems, the generator of Fooled by Randomness. It was also the problem with my earlier work Dynamic Hedging where I examined the contact between theory and real-ity. Platonicity is the generator of these problems.

Can you explain Platonicity?It is the focus on those pure and well-defined, and easily discernible, objects, whether triangles, or more social notions, like friendship or love – at the cost of ignoring those objects of seemingly messier and less tracta-ble structures. Seeing patterns is a form of Platonicity as we refuse the mess of randomness. This is also the product of our need for reduction – we need to reduce, and Platonicity has side effects.

Can you give us an example?Take a look at a schoolworm picking up a new language. He will learn, say, Serbo-Croatian or !Kung by reading a grammar book cover to cover, and memorizing rules; he will operate under the impression that some higher-up grammatical authority set the linguistic regulations so that non-learned ordinary people could subse-quently speak the language. In reality, grammar is something that comes after language, just as in any form of scientific model. It is something peo-ple without anything more exciting to do in their lives codify into a book.

While the scholastic-minded will memorize declensions, the aplatonic non-nerd will acquire, say, Serbo-Croatian by picking up potential girlfriends in bars on the outskirts of Sarajevo, or talking to cab drivers, then fitting (if needed) grammatical rules.

The same applies to science. It can be generalized to the manner that some people have a personal inclina-tion to dip into textbooks while oth-ers pick up their knowledge on the streets (there may be a few rare com-binations). Consider the economic planner. As with language, there is no grammatical authority codifying the economy; but try to convince one of those economists that the world might not want to follow his scientif-ic equations. Indeed Austrian econo-mists used the notion “tacit knowl-edge”, about economic matters, precisely for that part of knowledge that precisely cannot be codified, but that we should avoid repressing. It just comes out on its own.

I learned from looking at finance professors with some model taking possession of their head. But the best manifestation of Platonicity is with the ludic fallacy

The Ludic Fallacy?We are basing studies of chance on the narrow world of games and dice. Aplatonic randomness has an addi-tional layer of uncertainty concern-ing the rules themselves.

I called Ludic fallacy (after the latin ludus, play) the misuse of games as the wrong epistemologi-cal ground. Let me show here how randomness ends up disappearing in these games. Just consider that you know the probability, and that the

payoff does not change throughout. The casino never surprises you by announcing that it will be paying you 100 times more, or a tenth of your take. Furthermore, the dice average out so quickly that I can say with certainty that the casino will beat me in the very near long run at, say, roulette, as the noise will cancel out though not the skills (here, the casino’s advantage). The more you extend the period (or reduce the size of the bets) the more randomness, by virtue of averaging, drops out of these gambling constructs.

Alas, we suffer from the general tendency by those who put their noses too much into maps to mistake the map for the territory. Go buy a history of probability and probabil-istic thinking: you will be showered with a series of names you see recur-ring in all these books, like Cardano, Pascal, Fermat, Huygens, De Moivre, Laplace, Gauss, De Finetti, or Kolmogorov.

Contrary to what we were taught in all these probability volumes, and in the misguided books on the history of probability and “risk”, gambling could not offer us lessons about real randomness, nor can it be a laboratory where you could get actual training for the messy real life. As just as we tend to underes-timate the role of chance in life in general, we tend to overestimate it in these games. It is the only place I know where we have mild Gaussian-style randomness!

Furthermore I found it infuriat-ing to listen to people who, upon being informed that I specialize in problems of Chance, immediately jump to references to the dice.

The ludic fallacy is present in the following chance setups: random walk, dice throwing, coin tosses, the infamous digital “heads or tails”

Standing at the Platonic Fold Nassim Nicholas Taleb brings us up to date on the thinking that is behind The Black Swan

Wilmott magazine 35

TALES OF THE UNEXPECTED

expressed into 0 or 1, the “Brownian motion” corresponding to the move-ment of pollen particles in water, and similar examples. These generate a quality of randomness that cannot be even qualified of randomness – pro-torandomness is a more appropriate designation. At the core, all these theories ignore a layer of uncertainty. Worse, they do not know it!

And the ludic fallacy leads to the Gaussian, the Great Intellectual Fraud, GIF.

Beware the Gaussian?The ludic fallacy leads to the use of that monstrosity called the Gaussian and its siblings. And fixing it is not easy. The problem is that you cannot naively fatten the tails: you are in a situation of uncertainty about the distribution. NonGaussian randomness generally implies that you do not know much about the tails.

Consider the following. We need data to discover a probability distri-bution. How do we know if we have enough data? From the probability distribution. If it is a Gaussian, then a few points will suffice – and we know how many. How do you know it is a Gaussian? from the data. So we need the data to tell us what is the probability distribution, and a probability distribution to tell us how much data we need. This causes a severe regress argument.

The other one is to start ranking results based on their dependence on models.

Platonicity has an important extension: categorization.

Categories?We need to extend our uncertainty to categories because categories are fuzzier than we think and we tend to want to see crisp boundaries. There are many things we cannot separate because they cannot be isolated. For instance, utility and probability can-not be dissociated.

Wilmott magazine is a second home for me, simply because I found credible like-minded people and social friends who can either agree with me, say Haug, or heavily disa-gree, like Elie Ayache. Ayache by chal-lenging my ideas has taught me a lot. He is the only critic I have felt com-pelled to answer. The others don’t tell me anything I didn’t think about.

Ayache argues that it is the philo-sophical foundation that needs to be overhauled in order that quantita-tive finance can be recognized as a science. That to accept your com-bination of skepticism and classical empiricism leads in two directions only, blind faith or skeptical resigna-tion. He proposes to find a third way through a rethinking of what a mar-ket actually is. What is your reaction to Ayache’s metaphysical interpreta-tion of derivatives as being the path toward this new approach?This is where I disagree. The skeptic in me is easily satisfied by my rank-ing situations in accordance to their robustness to model error, or their

Platonicity. For instance a portfolio that depends on tail probabilities is much less robust than a long-only venture capital investment or a long out-of-the-money calls. An arbitrage with cash and carry is more robust than a probabilistic one. You can generalize into soft and hard knowl-edge in real life, outside of finance.

As to metaphysics, I will leave it for another discussion. My branch of philosophy is that of statistical inference, a supervision of scientific claims, nothing abstract. But I agree that we can escape problems by rephrasing them, except that I do not want to escape these problems. I am both a radical skeptic and an empiricist and I do not find the two incompatible at all. The school of empirical medicine in the second and third centuries was grounded in radical skepticism.

The real divorce for me is between aplatonic view of the world and aes-thetics. I want the world to be more aesthetic yet I want to acknowledge that it is considerably less so than we think – or not where we think.

On the basis that our sampling is wrong to start with when it comes to assessing the value of a given approach to trading and investment, and that our programming doesn’t allow us to (easily) accommodate black swans prior to the fact it is highly unlikely that a shift is going to occur in finance toward the basis in skeptical empiricism that you repre-sent. Do you agree with this?

Finance is moving very quickly into firmer grounds. People know of the severe “Nobel side effects” and pseu-doknowledge. We already know that much of portfolio theory and modern finance provide negative knowledge because of their side effects, that many hedge fund managers have LTCM-style cosmetic alpha.

The next LTCM-style blowup will help a lot.

Can you say something about how an approach like Mandelbrot’s would disabuse the market of this dependency on the Gaussian?Mandelbrot changed geometry. He will change much of social science – his model starts with the tails as a central element, not some append-age. It is a completely different way of thinking –if you are going to be Platonic, this is the way to do it. I will fight to the end to promote his meth-od as a framework to look at devia-tions. He is the only person who ever understood what I was saying.

Can you envisage a time when trading and investment in general takes a more realistic view on out-liers and how to incorporate them into models? If this occurs what will the markets look like, won’t this cut out so much noise that sustains the industry? In such a world without the GIF holding sway will everyone be able to go home at the end of the day less anxious because markets have become less volatile as a result?I do not know what it would do to markets. But one day we will under-stand the toxic side of beliefs (and models). One day we will understand Shackle’s point that economics and human affairs are about epistemics, the study of unknowledge. It will make the world a better, and certainly safer, place. Ciao.

The aplatonic non-nerd will acquire, say, Serbo-Croatian by picking up potential girlfriends in bars on the outskirts of Sarajevo

36 Wilmott magazine

W