đ 'Fooled by Randomness' by Nassim Taleb
Incerto Book 1
âThis book is the synthesis of, on one hand, the no-nonsense practitioner of uncertainty who spent his professional life trying to resist being fooled by randomness and, on the other, the aesthetically obsessed, literature-loving human being willing to be fooled by any form of nonsense that is polished, refined, original, and tasteful. I am not capable of avoiding being the fool of randomness; what I can do is confine it to where it brings some aesthetic gratificationâ
âWe favor the visible, the embedded, the personal, the narrated, and the tangible; we scorn the abstract. Everything good (aesthetics, ethics) and wrong (Fooled by Randomness) with us seem to flow from itâ
âRecall that epic heroes were judged by their actions, not by the results. No matter how sophisticated our choices, how good we are at dominating the odds, randomness will have the last word. We are left only with dignity as a solution - dignity defined as the execution of a protocol of behavior that does not depend on the immediate circumstanceâ
C.P. Cavafy in âThe God Abandons Antonyâ: Antony is defeated and betrayed, and is told not to mourn his luck, not to enter denial, not to believe his ears and eyes are deceiving him.
âJust listen while shaken by emotion but not with the cowardâs imploration and complaintsâ
While shaken by emotion: there is nothing wrong with emotions - what is wrong is not following the heroic, dignified path
âStart stressing personal elegance at your next misfortuneâ
âTry not to play victim when diagnosed with cancer (hide it from others and only share the information with the doctor - it will avert the platitudes and nobody will treat you like a victim worthy of their pity; in addition, the dignified attitude will make both defeat and victory feel equally heroic). Be extremely courteous to your assistant when you lose money. Try not to blame others for your fate, even if they deserve blame. Never exhibit any self-pity, even if your significant other bolts with the handsome ski instructor or the younger aspiring model. Do not complainâŚ.The only article Lady Fortuna has no control over is your behaviorâ
âDonât do to others what you donât want them to do to you; the rest is just commentaryâ
Do not assume positive events were due to skill - assume they could have been luck and proceed accordingly
Consider all possible outcomes (both up and down) - donât just assume that because something happened that it was likely to
e.g., âIf I did this 1 million times, what is the outcome that would happen the most?â
Focus on inputs, not outputs (since outputs can be affected by randomness)
$10 million earned through Russian roulette does not have the same value as $10 million earned through the diligent and artful practice of dentistry
Oneâs dependence on randomness is greater than the other
Reality is far more vicious than Russian roulette
It delivers the fatal bullet rather infrequently, like a revolver that would have hundreds or thousands of chambers instead of six
After a few dozen tries, one forgets about the existence of a bullet, under a numbing false sense of security
The risk is undefined (unlike Russian roulette)
We struggle with appreciation of preparing for an abstract downside that doesnât end up happening
Our tendency to make and unmake prophets based on the fate of the roulette wheel is symptomatic of our ingrained inability to cope with the complex structure of randomness prevailing in the modern world
It is not natural for us to learn from history - our tendency is to have to experience it for ourselves
e.g., Children will cease to touch a burning stove only when they are themselves burned; no possible warning from others can lead to cautiousness
We also struggle to learn from our own history
We donât learn that our emotions from past events are short-lived (positive or negative)
Yet, we continually retain bias that purchases will bring long-lasting happiness or setback will cause severe or prolonged distress when these feelings were short-lived in the past
There is nothing wrong with a risk-taker taking a hit, provided that one declares that they are a risk-taker rather than that the risk being taken is non-existent
Those that âblow up,â or those that believe the risk taken is non-existent, think that they know enough about the world to reject the possibility of the adverse event taking place; thereâs no courage in taking the risk, just ignorance
They often make claims that âthese times are differentâ
Theyâre unable to see that these âcrashâ experiences have been lived and documented many times before
A mistake isnât to be determined by the outcome, but in light of the information to that point
We often judge by hindsight bias (overestimation of what one knew at the time, due to being able to see the actual result)
Unlike hard science, history canât lend itself to experimentation
But, over the long run, it delivers most of the possible scenarios
Ergodicity: long sample paths end up resembling each other; they revert to their long-term properties
Unlucky but skillful people eventually rise
Lucky fools converge to the unlucky fool
Some new information is better than past information, but the average of distilled past information is better
Focus on distilled past information (first principles) (this especially applies to reading)
The opportunity cost of missing an extraordinary new thing is low vs. the toxicity of all the garbage one has to go through to get to the jewels
âIf there is anything better than noise in the mass of âurgentâ news pounding us, it would be like a needle in a haystackâ
The ratio of undistilled information is rising; the need to capture attention > signal
If an event is important enough, you will hear about it
Over short time increments, you observe variability (noise), not returns (signal)
The higher the frequency of observation, the more volatility (noise) you consume
Plus, the volatility results in you feeling emotional highs and lows (which are 2.5x worse physiologically) every step of the way
Higher frequency exacerbates lows > highs
It is less unpleasant to lose $100 once than lose $1 one hundred times
So, making $1 per day for a long time then losing it all is actually pleasant from a hedonic standpoint, although nonsensical economically
âWe do not need to be rational and scientific when it comes to the details of our daily life - only in those that can harm us and threaten our survival. Modern life seems to invite us to do the exact opposite - become extremely realistic and intellectual when it comes to such matters as religion and personal behavior, yet as irrational as possible when it comes to matters ruled by randomness (say, a portfolio or real estate investments).â
Yiddish saying: âIf I am going to be forced to eat pork, it better be of the best kindâ
If Iâm going to be fooled by randomness, it better be of the beautiful (and harmless) kind
How to operate as humans with all this uncertainty?
âSatisficingâ: stop at near-satisfactory solutions; otherwise, you could spend eternity reaching small conclusions âperfectlyâ
Use heuristics: practical methods or rules
Beware that these are prone to bias
It doesnât matter how likely an event is to happen - what matters is how much is made in that scenario
Probability doesnât matter, probability x outcome matters
Nobody takes home a check linked to how often theyâre right - what you have is a profit or loss
An event, although rare, that brings large consequences cannot be ignored
You cannot remove the outliers if the impact is large - you have to take them into account
Rare events are always unexpected, otherwise they would not occur
If you expected them, you would prepare and avoid them
Because negative variations cause more emotional pain than positive, people are drawn to strategies that experience rare but large variations
âPushing randomness under the rugâ
They will aim for less frequency of variation, which by nature means the unexpected will happen
People are more sensitive to presence or absence of stimulus rather than its magnitude
Statistics fails us when distributions are not symmetric
Statistics = information
By nature of asymmetry, we collect more information on the more likely (i.e., frequent) event, so we get more confident in predicting the frequent event (confirmation bias)
This makes our understanding that the rare event is less likely than it is
Especially if this is greater in magnitude, the effect is severe
This is exacerbated by assuming that our knowledge of future events increases by identification of past events
But, things change, and past events are not indicative of future events
So, statistics fails us
It underestimates the likelihood of the rare event
It doesnât take magnitude into account
The problem with empiricism and extrapolation: the Black Swan problem
No amount of observations of white swans can allow the inference that all swans are white, but the observation of a single black swan is sufficient to refute that conclusion
You can use data (empiricism) to disprove a proposition, never to prove one
You canât make the leap easily from âthe market has never gone down 20% in any 3-month periodâ to saying âit will neverâ
Example:
A: âNo black swans exist because I looked at 4,000 swans and found noneâ
B: âNot all swans are whiteâ
You cannot logically say A, no matter how many swans seen (unless you know with certainty that youâve seen all available swans)
However, it is possible to say B by finding one single counterexample
You often hear âit has never happened beforeâ
But, when your last âworst caseâ scenario was experienced, it was a surprise
If your past didnât resemble your pastâs past, why should our future resemble our current past?
This is the problem of induction (empiricism of the past driving future expectations)
Popperâs response to induction
Induction is risky because black swans make empiricism of the past impossible to drive expected future events
Popper: only two types of theories
A: Theories known to be wrong, as they were tested and rejected (falsified)
B: Theories not yet known to be wrong, but exposed to be proved wrong
A theory can never be verified
So, the only impactful event is if youâre proven wrong (thatâs what you have to prepare for)
Knowledge doesnât always increase with more information; itâs also not about what we know, itâs about what we donât know
Taleb speculates in all activities on theories that represent some vision of the world, but stipulates that in carrying out the theory, no rare event should harm him; all conceivable rare events should help; in other words, if heâs proven wrong, it should help
Result: an open mind is necessary when dealing with randomness
Inductive inferences are easier for the mind
Causality (the reduction of the general occurrences to a particular cause) is easier to commit to memory than all the specifics
Talebâs strategy: will use statistics and inductive methods to make aggressive bets, but will not use them to manage risk and exposure
Trade on ideas based on some observation (which includes past history), but (like Popper), make sure that the costs of being wrong are limited (donât determine probability based on past data)
Know which events would prove you wrong and allow for them; prepare in this context
The extent to which you can translate past success to future success is based on:
The randomness of the profession
The number of people doing it
You donât see the full sample of all failures (survivorship bias), making things seem likelier than they are
The highest performing will be most visible, as the losers donât show up
Biases result in improper understanding of probabilities of historical outcomes
Since we extrapolate history to probabilities in the future, this sets us up poorly
Applicable biases
Survivorship bias: assuming the probability of success is higher than it is by looking at the survivors (and not those that tried and also failed)
You need to take into account the whole population (including those that tried and failed) to understand the validity of track record
If population is 10,000 vs. 10, higher chance that somebody will be successful due to randomness
Hindsight bias: overestimation of what one knew at the time, due to being able to see the actual result
Availability bias: over-estimating the frequency of an event if you can recall an occurrence
e.g., The likelihood of a terrorist attack seems greater than it is due to our memory of 9/11
Representative bias: gauging the probability that a person belongs to a particular social group by assessing how similar their characteristics are to the âtypicalâ member
Simulation bias: playing the alternative scenario; âwhat would have happened if I sold at the peakâ
Affect bias: emotions elicited by events determine our view of their probability
Avoid backtesting (trying to find any correlation to historical results)
You must look for a given relationship, not any
You want a narrow lens + time (to get benefits of ergodicity, of time lessening effects of randomness)
Path-dependent outcomes throw off probability; the world increasingly has this property - a core component of Extremistan, a world in which we increasingly live (vs. Mediocristan)
e.g., Winning makes it more likely to win in the future
Drives huge variance and winner-take-all effects
Network effects are an example (the more connected the network, the more useful, thus the more likely to succeed)
Two systems of reasoning
System 1 (fast thinking): effortless, automatic, emotional, social, personalized
System 2 (slow thinking): effortful, self-aware, slow, deductive, asocial, depersonalized
Regression to the mean: the larger the deviation from the norm, the bigger the probability of it coming from randomness vs. skill or causality
But, if it continues to deviate from the norm repeatedly (due to effects of ergodicity, may be due to a certain cause)
e.g., The largest dogs tend to product offspring smaller than them
The greater the variance in outcomes, the more causation matters if there is causation (for that, see above - look to ensure continued variance to benefit from ergodicity)
e.g., If a 1-month bike race and the winner wins by 3 weeks vs. 1 second, you might start to look for potential causes
Look at large percentage changes - unless something changes by more than its usual daily percentage change, the event is likely noise
A 7% move can be several billion times more relevant than a 1% move
When change in amplitude is small, itâs more likely to be from noise
The likelihood of an event being a signal increases exponentially as magnitude increases
Trading example
$100,000 profit - may assign 2% probability to the cause being a strategy, 98% to noise
If $1,000,000 profit - may assign 99% probability tot he cause being a strategy
âSince the heart doesnât agree with the brain, we need to take serious action to avoid making irrational trading decisions, namely, by denying access to a performance report unless it hits a predetermined thresholdâ
Enables you to filter out more of the noise and look for signals (longer time horizon + bigger moves)
âUnless you have confidence in the rulerâs reliability, if you use a ruler to measure a table you may also be using the table to measure the rulerâ
Our behavior is immediately to establish a causal link between two events, even when one may not exist - we are not made to view things as independent from each other
âProbability is not about the odds, but about the belief in the existence of an alternate outcome, cause, or motiveâ
Lower-ranking persons are judged on both process and results, while top management is paid on results only
The CEO makes a small number of large, non-repeatable decisions, more like a person walking into a casino with a single $1,000,000 bet
External factors play a considerably larger role
Subway riders are considerably freer than train-riders
Higher frequency of trains
Uncertainty of schedule
A slightly random schedule prevents us from optimizing and being exceedingly efficient, particularly in the wrong things
Research on happiness shows that those who live under self-imposed pressure to be optimal in enjoyment of things suffer distress
If you enjoyed this newsletter, forward it to a curious friend and tell them which idea they'd love.
Click here to read the archive.
Click here to follow on Twitter.
If you were forwarded this email, click the button below âŹď¸ and enter your email to subscribe.