Pasquale Cirillo
@DrCirillo
Followers
17,279
Following
198
Media
438
Statuses
3,482
Professor of Data Science at @sml_zhaw . Numquam periculum sine periculo vincitur. Risk, probability and statistics.
The Logic of Risk
Joined August 2015
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Labour
• 930483 Tweets
Reform
• 687602 Tweets
Argentina
• 625011 Tweets
Dibu
• 606455 Tweets
Messi
• 429791 Tweets
Ecuador
• 301419 Tweets
Kendrick
• 286020 Tweets
MyGO
• 271430 Tweets
Keir Starmer
• 222341 Tweets
Corbyn
• 144816 Tweets
水分補給
• 131760 Tweets
Rishi Sunak
• 97993 Tweets
Greens
• 76288 Tweets
Enner
• 64813 Tweets
Mogg
• 59574 Tweets
Clacton
• 57743 Tweets
Lib Dems
• 48168 Tweets
Scaloni
• 39504 Tweets
Liz Truss
• 38441 Tweets
クーラー
• 36631 Tweets
ميسي
• 35346 Tweets
Hayırlı Cumalar
• 25239 Tweets
#يوم_Iلجمعه
• 24913 Tweets
#UKElection2024
• 20212 Tweets
KERENBANGETTT
• 17932 Tweets
KerjaCERDAS KerjaKOMPAK
• 17241 Tweets
災害級の暑さ
• 14364 Tweets
Pinned Tweet
The fence paradox: we feel safe, we are not.
Financial regulations are like fences: they just hide risk.
72
1K
3K
I have decided to share my risk management classes at TU Delft.
Later today, or tomorrow, I will post the first class. Just recorded.
You will have access to most of the materials I share in class.
They are at the bachelor level, only basic statistical knowledge required.
32
92
691
One of the best books on the philosophy of probability and statistics, together with de Finetti's The invention of truth.
Almost no formula, beautiful examples.
17
73
699
If it happened, it can happen again.
If it didn't happen, it can still happen.
This is a simple rule in risk management.
Exceptions are very limited.
10
95
480
The death toll in Gaza (even with all the uncertainties of estimation) is now above the 60th percentile of all armed conflicts by number of casualties since the 16th century.
24
153
372
An informal conversation with
@nntaleb
.
Tomorrow, Monday 25, at 5pm (CET), in sync with the paper becoming openly available.
10
72
345
Once created, risk can be hidden, transferred, hedged, but no longer destroyed.
22
73
328
The concept of I.Q. is flawed.
Wearing a mask signals caring not only for yourself but also for others.
It is more a measure of empathy and altruism.
8
47
323
Every day (today in the Middle East, yesterday and today in Ukraine, and in many other places), we see how "The Better Angels of Our Nature" was/is a bluff.
A dangerous bluff, because if you deny risk, you cannot prepare for it.
With
@nntaleb
we have been saying this for years.
5
47
317
We all desire peace, but we cannot live under the spell of illusion, as Europe is re-discovering.
Data are there to remind us that, unfortunately, we humans are as belligerent as we have always been.
Here a non-technical note
@nntaleb
and I wrote in 2016.
9
49
284
Machine Learning for Risk Managers.
The first lesson is out.
The second is just behind the corner.
Enjoy!
11
46
290
Partisan science is no science.
And "a matter of life and death" sounds very much like millenarianism.
Scientific American has never endorsed a presidential candidate in our 175-year history—until now.
The 2020 election is literally a matter of life and death. We urge you to vote for health, science and Joe Biden for President.
13K
82K
183K
15
50
287
Ottimo italiano, Nassim.
Bravo.
14
17
257
One of my PhD students sent me this picture, which I find terribly representative of what happens very often out there.
6
61
255
Today Italy celebrates Liberation Day.
On April 25 1945 we finally got rid of fascism.
For too many years we had been on the wrong side of history-we should never forget!-but we also had a chance to redeem ourselves, and to become a democracy.
Viva l'Italia! Viva la libertà!
5
15
253
Discussing with colleagues, I often discover that even many statisticians are not aware of the different types of "Heavy Tails".
Fat tails are heavy, but not all heavy tails are fat.
The Lognormal distribution does not have fat tails (yet it can be very tricky and dangerous).
8
86
249
Yesterday I was asked about the difference between fat and heavy tails.
They are often used as synonyms, and for a non-technical discussion it is fine.
But from a technical point of view a fat tail is heavy, but not all heavy tails are fat.
Here an old picture about this.
4
60
230
Evidence that coronavirus superspreading is fat-tailed (with α parameter lower than 2, definitely scary yet important for countermeasures!).
8
68
228
I am starting a new YT playlist, where I will collect some brain teasers and exercises taken from one of my courses.
We will solve the exercises using important results in maths, stats and probability.
Hope you enjoy it.
3
23
229
In (badly) applied statistics, a common bias is apophenia, that is the tendency to see connections between unrelated things.
Interestingly, in risk management, it is the opposite: independence is assumed and perceived everywhere!
4
54
222
Another nice reading, which reminds me of some of de Finetti's views.
6
14
207
Nice work, which further confirms and extends the point
@nntaleb
and I raised a few months ago.
"Evidence that coronavirus superspreading is fat-tailed"
5
37
199
Homemade culaccia, matured 18 months in our cellar, from home-grown pork.
Culaccia is a very high quality type of ham from Northern Italy.
10
6
191
When speaking about risk, there are some common logical and argumentative fallacies. Here attached a few examples. Can you see what is wrong with each of them?
11
63
195
If you kill variability, you kill selection.
If you kill selection, you kill the markets.
5
106
189
Stay safe and have a nice Sunday.
Here I'm having some homemade pici (a Tuscan type of thick spaghetti) cooked in red wine.
Staying at home doesn't mean eating bad food. Actually the opposite!
11
5
183
July 31st will be my last day at
@tudelft
.
I am grateful to my colleagues and to this wonderful university.
But now it is time to move on, never stay too long in the same place, always take control of your life and put some skin in the game.
Incipit nova vita.
18
3
178
Since I start seeing the same errors we saw just a few months ago, with naive extrapolations, fittings and curves, here we are again.
With
@nntaleb
4
36
181
Isolation in the woods, with good books and homemade pizza.
4
4
175
Lesson 1 is out:
3
23
174
For an Italian, it is not really Sunday without homemade lasagne.
Today no probability, just mouthwatering certainties.
12
13
164
I have been living abroad since 2007, but my connections with Italy are strong, and I love my country.
Yet I am very indignant at the way the Italian Government has managed the coronavirus problem.
A mix of incompetence, stupid improv and underestimation.
A shame.
16
28
170
When a Government starts saying "we will use big data for...", run.
6
30
167
Every time we estimate something we make some error, and this error has its own error, with its own error, with its own error...
10
21
172
19th Century
Estimated nr. of deaths*: 4'300mio
Estimated nr. of deaths in wars: 45.8mio
% of deaths due to wars: 1%
% including El Niño Famines: 2%
20th Century
Estimated nr. of deaths*: 5'500mio
Estimated nr. of deaths in wars: 214mio
% of deaths due to wars: 4%
* All causes
23
49
161
As far as Covid data are concerned, I have criticised Excel amateur statisticians for months.
Now we have the professional league of Covid Excel statisticians.
And sadly, this is no surprise to me.
In the UK the number of cases rose rapidly.
But the public – and authorities – are only learning this now because these cases were only published now as a backlog.
The reason was apparently that the database is managed in Excel and the number of columns had reached the maximum.
1K
8K
18K
8
19
171
Central banks want to step into the battle against climate change, against the "green swan."
Prepare for new absurd regulations and dangerous market distortions then.
The road to hell is paved with good intentions.
7
42
169
Heavy tails, long tails or fat tails?
Now, you have no excuse left to mix or confound them.
Welcome to a bestiary of tails.
9
34
169
I have been cooking all day all the traditional dishes of my family.
For a couple of days no probability nor risk.
Only my loved ones and good food.
My best wishes to all of you.
14
2
162
In studying war casualties and other historical data, you have to treat them as imprecise random observations. If you don't, all your subsequent analyses will be flawed.
cc
@nntaleb
6
18
153
Dataism or Big Dada: a rising movement made up of self-declared experts and "data artists" who blindly and naively apply the "latest tools", without actually knowing what they're doing.
Trivial and nonsense results are then announced as revolutionary.
6
38
164
This is grandeur.
3
13
161
On Tue 21, Quantitative Risk Management will start.
These are the topics of the different lessons (minor changes possible), which will appear here:
Lesson 1: The Philosophy of Risk 1
Lesson 2: Risk Measures and Functionals
Lesson 3: Extreme Value Theory 1
4
23
161
As I did for Risk Management, I share online the videos of my live course in Financial Mathematics.
4
17
157
Homemade lasagna with basil pesto and baccalà (salt cod).
4
1
156
Hi Nassim
@nntaleb
, I finally found a black swan in my home library.
It says it waits for you here in Switzerland.
3
1
157
Quiz.
If I tell you that I have two children
and that one of them is a girl, what
is the probability that I have two
girls?
You can assume that girls and boys
are equally likely to be born, and that genders are independent.
73
13
154
It is highly surprising (and scary) to see how many data analysts ignore the problem of hidden events in their data. Reporting delay is rarely treated, and this is particularly dangerous when blindly applying machine learning techniques.
8
21
149
The paper with
@nntaleb
and
@yaneerbaryam
is finally online.
Once completely processed, it will be open access.
2
35
145
I'm working on a risk management diploma on a freemium basis.
Stay tuned.
@IFFunic
What an interesting content the QRM (Quantitative Risk Management) Lessons by
@DrCirillo
on YouTube!
Search and follow him @ channel "The Logic of Risk".
I'm loving these lessons: I can review some topics as well as I can learn a lot!
Grazie mille Prof ✅
0
5
24
6
11
144
There is only one possible answer.
DON'T YOU EVEN DARE.
Do we all agree that spaghetti noodles are too long?
I see no disadvantage in them being half as long
292
32
651
14
8
137
I am becoming increasingly impatient with those who blather on about artificial intelligence and machine learning and then don't know what a logistic regression is or how to interpret its parameters, and with those who think they can safely use code they don't understand.
9
16
143
Starting from Tue 21, when the new quarter starts, I will share online my new Quantitative Risk Management lessons (master level), and continue sharing the Financial Mathematics ones (you already find the first part of the course here: ).
4
20
139
Visiting my parents in Italy.
Today is the day of the "Norcino", the highly specialized pork butcher of the Italian tradition.
We are preparing our sausages, culatellos and meat for the next months.
Everything is home-made here.
My parents are definitely self-sufficient foodwise.
9
7
137
A short 48-hour travel back home to Italy.
It does not matter if you have grown there, this kind of beauty is always astonishing, nectar for your soul.
2
6
133
I am finalising the last pictures and cleaning the example codes, then it is ready for the new academic year.
11
7
138
Brain Teasers, Episode 11.
Another exercise with Markov chains.
How many tosses of a fair coin we need on average to get HHH or THH?
12
15
133
The Gini index (or coefficient) is a statistic, therefore it is neither good nor bad. It is just a tool.
It has nice properties but also important limitations.
Unfortunately most of the times I see it used or cited, those limitations are completely ignored.
5
20
127
What makes probability so beautiful to me is the fact that even in its definition there is no certainty.
And you have to accept this to grasp the inner meaning and beauty of many probabilistic results.
6
17
128
Two extremely nice readings on ergodicity and economics:
The former also has a beautiful picture :-)
4
36
124
Pinker is very good with sophisms, writing ambiguous statements and pivoting on words he can ex post interpret as needed.
In Italian we would say "rigirare la frittata," meaning "to turn an argument around," "to turn someone's words on him/her."
What struck me about this blurb:
a) There was a section explicitly on the question of whether the world is getting more peaceful, with two papers, one by me and one by
@nntaleb
and
@drcirillo
.
b) Our answer was "no." We did not disagree at all.
1
15
91
2
18
122
Ossobuco alla milanese, a traditional dish from Milan.
Because, you know, Italian cuisine does not exist, it is an invention for foreigners.
In Italy we have local dishes from Rome, Bologna, Florence, Venice and all our beautiful villages and towns, from the mountains to the sea.
9
5
123
Forthcoming in Nature Physics.
Sir, don't trust SIR (and friends).
Paper with
@DrCirillo
"Tail Risk of Contagious Diseases" using Extreme Value Theory.
Advice by non EVT-trained epidemiologists concerning COVID-19th tail risks is no different from advice by a carpenter concerning complex odds at roulette tables.
19
118
400
4
18
124
"Governments are good at doing things that look important but are not necessarily effective."
4
22
129
It may be difficult to accept initially, but then it is a real epiphany: given our incomplete understanding, we need risk-survival heuristics (paraphrasing
@nntaleb
).
We need to eat humble pie and stop thinking we are able to fully grasp, compute and mitigate risk.
If politicians and regulators understood point 1) alone, it would be enough to solve many problems of today.
1
2
32
3
18
119
At
#RWRI
,
@nntaleb
,
@financequant
and
@normonics
are fighting over the best definition of entropy...
I don't see why, given that the best definition is mine 😅: a pseudo-metric of distortion, whose interpretation varies with its generator.
7
11
122
Pre-print worth reading.
Another reason why what
@normonics
@nntaleb
and
@yaneerbaryam
said one year ago was and is so important.
4
27
121
The basic Risk Management course is approaching its end, but in February I will also record my live Financial Mathematics lessons. It will be a course (master level in maths) about the use of stochastic calculus and arbitrage theory for pricing and financial modeling.
The second last video of this introductory course to risk management is available.
In the second part of Lesson 9 we discuss the basics of the model building approach (i.e. var-cov) to Market Risk.
3
3
27
7
11
121
I was born in Tuscany, Italy, and one of our old sayings tells that "Chi non risica non rosica," meaning that "if you do not risk, you will never get anything."
Simple and powerful.
15
8
124
Homemade truffle polenta with venison and red wine ragu.
10
0
127
What counts is not having the perfect model, which is impossible, but rather knowing how much model risk you can tolerate.
This is what makes you a good statistician and risk manager.
1
31
122
One of the exercises of my forthcoming course (starting next August 5) for
@NYUFRE
about statistics, probability and brain teasers for quants.
23
11
121
The problem is never the measure, here the p-value, the problem is that people don't know statistics, which is a subtle and difficult discipline.
1
37
121
@nntaleb
@normonics
@tonyennis
@michaelshermer
It's funny to read claims about changes in the number and magnitude of wars, when the average time span between two large wars is larger than the observation window used to support the claim.
It's like: since it has been raining for three days, the sun does not exist any longer!
4
32
112
@nntaleb
@fractalharry
@Forbes
Some progressive narrative likes to describe the past as something darker, worse and more violent.
Data simply show that the world changes, but our violence and belligerence stay the same. They only adapt to the new technologies.
Think of cybercrime.
7
35
116
Qui, se interessa, parlo del rischio di coda nelle pandemie, sulla base del recente lavoro con
@nntaleb
.
(the video is in Italian)
6
22
111
In my opinion, the future of university education will be a mix of good-quality cheap or free technologically-enabled "homeschooling", and very old-fashioned but maximally effective bottegas, especially for high-level research-driven students.
8
17
114
I've been challenged to cook some Italian American classics.
And I really enjoyed it.
Here we are: Spaghetti with meatballs (following an original Staten Island's recipe).
14
3
110
I'm very happy to say that I received a
@DIZH_ZH
fellowship for 2024-26.
I will work on new risk measures for fat tails, with applications in climate finance.
19
2
110
The Causes of Peace is out.
Is humanity becoming more peaceful?
5
22
105
As usual, no probability on Sunday.
Offal is an important part of Italian cuisine, which has its roots in poor dishes for the people, not for aristocrats. The best Italian (regional) dishes are simple, uncomplicated and genuine.
A good example is Trippa (tripes) alla fiorentina.
9
6
105
Tomorrow I will record the second class of Financial Mathematics.
The first one (divided into 3 videos) is here.
(always check the description of the videos, you could find the lecture notes and extra materials)
5
10
106
Today I had a discussion with a risk manager about operational risk.
He could not see the paradox of the Basel framework in dealing with such an erratic object.
Here is a joint work with
@nntaleb
on the matter.
4
20
101
If you wait for all the fantastic & reliable data, before taking your risk decisions:
1) you fall into the delay fallacy
2) you probably die while waiting
Ioannidis got the reasoning completely BACKWARD.
I mean reallllllly backward.
In the real world, one must REDUCE RISK in the absence of reliable data, via the MOST ROBUST (model resistant) method.
That's the message of the INCERTO.
94
517
2K
0
30
101
The distinction between risk and uncertainty is just utterly academic.
8
26
101
I am glad and proud to announce that, together with colleagues from Spain, Italy, Belgium and the Netherlands, I have won a big EU H2020 grant.
I will supervise the subproject related to the use of urn models and machine learning in credit risk.
7
8
100
Apparently, if you ignore it, operational risk disappears.
They think.
6
11
100
In the
#AppliedForecasting
course (
@M_O_F_C
), we will soon discuss tails, uncertainty and many horoscope-like predictions.
If you want to forecast, you need to know your model and its/your limits very very very well...
3
18
102
The textbook we use is the one you see in the picture, but it is not strictly necessary to follow the course in its basics.
My slides should be enough as a guide. Then you can complement them as you prefer.
3
5
100