Here I am at #RWRI13.
Day 1, arrived just in time, so I couldn't grab a front seat.

@NonMeek will be happy to learn that his latest Offshore mini-book is in the opening slide.

1/n

There were some jokes about Raphael almost inaudible. Audio fixes now. NNT saying some instructors hate each other. Attendance:
2 MD, 25 in finance (1 VC, 1 academic), 10 traders, no one in insurance, 5 consultants, 1 electrical engineer, 2 working for military, 2 unemployed

2/n

that identify as entrepreneurs, and almost half of the attendance dead lifts.

The heavy brochure is almost a present for those who dead lift. Many 2 days attendants (10 to 15), when I came here first we were just a couple doing the 2 days only.

Coffee breaks contain carbs.

3/n

In the past, some hackers from the Czech Republic attended it. I remember a Burlgarian professional internet poker player from my first visit at #RWRI 6.

Seriuous stuff starts.

4/n

Puts in Mathematica.

What is the correlation of 2 random vectors containing 18 numbers each?

5/n

It is something like 42%.

NNT said this is what FBR is about. Journalists/academics will look for such correlations until they will find one.

Why 18 numbers?

6/n

There is a guy called Tyler W. (could not grasp) who is an expert in finding correlations such of number of chickens living in your state with number of strangulated people, and similar spurious correlations.

The "FBR" effect.

7/n

It is easy hence to hack correlation.

Let us now look at fake clusters.

A matrix is generating random 0, 1. It is almost impossible to generate a series of more than 7 0s or 1s. But if we clump from top or lower rows we can get clumps of up to 11 such "series" easily.

8/n

Raphael explaining some of his adventures hacking Côte d'Azur casinos using some of these tricks.

Many of these random "clumps" referred if they are positives for cancer will sue their water company, NNT explains. People experience a sampling error, we will cover this...

9/n

when we will see ergodicity.

Let us talk "Mediocristan" and "Extremistan". How many have read TBS?

Almost all the attendants have.

We will start seeing also the randomness of many of the "slopes" in the linear regression graphs.

10/n

NNT saying most of those slopes would need to be 0.

He (NNT) has made his life detecting ecomomists BS. How many professional economists in the room?

A couple rise their hands, timidly.

11/n

Logic of risk taking:

"Better take risks that you can measure than measure risks that you are taking".

It is often easier to behave under uncertainty than under measure risks.

12/n

He gives an example cherished by your truly.

You do not know if someone has coronavirus. Do you shake his hand, yes or no?

13/n https://twitter.com/NachoOliveras/status/1228555864979771398?s=20

Obviously not.

Everyone understands this except some academics, like some Alberto Alemanno NNT says he is after on Twitter on these later days.

Often it is better to have 1 lawyer than 10 statisticians. Lawyers clip tail risks, on occasions.

14/n

Some people in real estate (e.g. Donald Trump) became an expert in making money by collecting options that had more upside than downside, sometimes he executed these options.

15/n

Richard Thaler cares about his reputation in Twitter.

Those "big guns" guys care deeply about their intellectual reputation, that is more fragile than it is apparent.

That is why NNT, Raphael go after these guys: Ruin is not the same type of risk as tinkering.

16/n

Path dependence matters: making money is OK, but to make money you first need to survive.

Scaling: we can vote for Netanyahu (right wing hawk) from a kibbutz (a communist setting). The scales of Israel and of the kibbutz are different.

17/n

Scales matter in risk management.

"If I die" NNT says, "that is a catastrophe different from humanity disappearing".

That is why we need to worry about coronavirus or GMOs: multiplicative risks.

Very different from traffic accidents. The worst threats are multiplicative.

18/n

Raphael: Did Chinese overreact with coronavirus?

No, they didn't, even if we ignore a lot of things about this virus, such as its mortality rates.

They did the right thing because the growth of the disease looked exponential.

19/n

Conditioned question: if 200 million were to die next year of traffic accidents or of coronavirus, what is more likely (even if many more people have died so far of traffic accidents)?

The latter is more likely. This is the reverse "stress test" we should apply often.

20/n

People like Pinker are unable to make a difference between "Mediocristan" type of risks and the "Extremistan" ones. They are reliable reverse indicators on how to think.
@trishankkarthik will visit us on day 4 to talk on this and the P/NP problem, so stay tuned.

21/n

Never use complicated solutions when an easy one is available.

Easy solutions are often "via negativa". E.g. one of his father (NNTs) patients was a guy in the diplomatic corps who had a throat rush.

The easier remedy?

Stop smoking!

22/n

Raphael on an Antifragility paper. NNT wrote a paper on it, no one understood it as "science" and thus didn't want it accepted in academic press, so NNT and Raphael added some obscure thing/useless formula to make it possible to publish as an acedemic paper (successfully).

23/n

NNT: Look at this BS (on his own paper with Raphael Douady), reading out loud passages: "the inherent capacity of antifragility, blabla"...

Academics don't have skin in the game, I had to put theorems with no meaning to get the paper accepted.

24/n

Sometimes it is needed to "outmath" economists to beat them at their own game.

25/n

Surgeon heuristic: you need brain surgery.

There is a choice for 2 surgeons. One has a Brooklyn accent. He has thick fingers. The other has nice glasses, speaks with long sentences with oldish words in it.

Which one to pick?

26/n

The Brooklyn "butcher look like" would have gone through so much hostility in his career, that if he got there, it is likely that he has some good "surgeon" game.

Evolution has to be thought as filtering. Raphael: in academia this is often the exact opposite.

27/n

I didn't know this one: NNT has worked for McKinsey.

He said he will have done some 5 or more presentations for them, overcharging them.

28/n

Risk management is more than statistics.

You will not get rich by being a good forecaster. Forecasting and speculation are different business.

x is not f(x).

29/n

Is anybody in the attendance over 85?

No one.

Framing matters. If you get corona your risk of death might be close to 3%.

But let us frame it some other way: your probability of survival is greater than 97%. Isn't that great?

30/n

Raphael: butcher heuristic applies to money managers. You can choose between an obscure guy, with diction problems, that is difficult to understand. Another one is a HBS graduate, over confident, eloquent. Which one to pick?

The HBS graduate is possibly great explaining...

31/n

how he lost your money.

NNT: I did a lecture for the Bank of England. The Q&A session was not included in the minutes. The Head of Research couldn't understand Markov processes. He asked why he should worry about risks that might come 10 years down the line. NNT got mental.
32/n

Never accept risk advice from non practitioners.

Rationality needs to be defined bottom up, not top down.

Raphael: if you stay home, eat veggies, never met women, will you live longer? Maybe, but what is certain is that life will feel longer to you!

33/n

Minority rule: Coca-Cola is kosher.

It is easier to save some pennies doing it kosher than handling separate inventories for kosher and non kosher Coke.

Behavioral finance is BS: if we understand how humans work, do we understand markets?

No, obviously. Time for coffee.

34/n

As live tweeting is so exhausting had some of the carbs that were promised on the coffee break to us.

Maybe I have had too many of them in the past couple of years since @digvijoy_c, whom I follow on Twitter for years, did not recognize me.

35/n

Briefly shook hands with NNT.

I ignore what leads him to believe I don't have corona (read tweet 14).

Maybe the crazy-paranoid tweets...?

36/n

Starting again.

The Egyptians calculated risks of floods marking the higher levels the river had earlier got.

They got f* a couple of times, of course. Raphael interrupts him.

37/n – at Lebanese American University

Same as the Japanese in Fukushima (assume they predicted a 10m wave as the worst possible scenario).

The Dutch like @dwnhogendoorn are smarter than this, and they dimension the boulders of their dams as bigger than the worst ever experienced past event.

38/n

The problem of models is that models have errors.

If we are linear or convex to those errors, it is not big deal.

There is always a f*er and a f*ee. You don't want to be the f*ee: that is the problem you have if you are concave to these errors. Those terms NNT learned them
39/n

in his first summer job at Citybank.

Let us start with the "seminar" proper.

What are fat tails?

40/n

There are several tentative definitions by the attendance.

There are many definitions he didn't like.

NNT, the best is one given after some ten attempts "A few determine the outcome of the distribution."

41/n

Mediocristan: if I get random 1000 people then I add to the sample the heaviest human on Earth, say a 500 pound individual. He will just weight a bit more than 3x the average person. What if I take 10000 people? His effect will even be more irrelevant: law of large numbers.

42/n

Mediocristan: What would be the maximum number of calories a human can get without collapsing? 4000 or 5000?

Can we halve our weight also in one day? Nope.

Can we lose half our money in just one day? Yep: Extremistan.

43/n