A book about luck disguised as non-luck (skills) and randomness disguised and perceived as nonrandomness (determinism).  

Far better to mistake skill for luck than luck for skill. A lucky fool; attributes his luck to a very precise reason, often skill. Probability applied to practice is a rare thing. 

Either we believe in a utopian vision – that we can control our nature; much like the cure for obesity is to tell people to be healthy. Or, we accept a flawed version of self; we don’t try to correct our flaws – we seek workarounds. 

This book, alongside many behavioural science books, suggests we are flawed beyond repair for this environment – we are predictably irrational – we might like to believe in reason and rationality but in truth we are idiosyncratic. We need wily tricks, not moralising help.

Delivering advice assumes that our cognitive apparatus rather than our emotional machinery exerts some meaningful control over our actions. Behavioural science shows this to be untrue.

Few claim art is a tool for the investigation of the truth; much better an escape from it or to make it more palatable.

We fail to think critically, confusing conjecture for truth. There are those who believe that there are easy, clear-cut answers and those who don’t think simplification is possible without distortion. Wittgenstein vs Descartes.  

The problem of induction – what we see is all there is (black swan) tied to the issue of skewness. 

When we look back at history only one observation is made. We take events that have occurred  – that is, an event that has happened – 100 % probability, i.e., certainty, and believe that to be the only possible outcome. There are of course alternative sample paths (invisible histories – events that could have happened) – The past is deterministic.

If studying the news has no impact on your ability to predict what happens next or your current knowledge then is it worth it? 

Taking into account observed and unobserved possible outcomes sounds like lunacy – probabilistic thinking –  based on all probable paths – some of which result in success and others in failure. 

On balance, you can judge success or failure in the long run. Ergodicity – overtime despite loss or success (sampling) we find our level 

Dentistry is stable; trading not so much – for evaluation of success, consider those who are in position and those who have left – not just the sample that has had success. This is survivor bias, a sampling error. That which arrived by luck can be taken away by luck. Things that come with little help from luck are more resistant to randomness. It doesn’t matter how frequently something succeeds if failure is too costly to bear (youth sport) 

“Common sense is nothing more than a collection of misconceptions acquired by age 18.” Einstein. [Plato, believed that common sense, or what the Greeks called “Doxa” was riddled with errors, and prejudice – and not to stand up to reason]. 

Alternative histories don’t need to be created to assess their attributes. Probability is a qualitative subject. In well-defined precise games, knowing the possibility of odds of a likely outcome is achievable, but calculating probability when we can’t see the forthcoming reality gives us meaningless answers about the odds. Beware of those offering slick easy to understand theories. 

Degradation of history- what happened to others won’t happen to you. Heroes are not heroic because they won or lost, but because of their behaviour – think If by Rudyard Kipling.  

A mistake is not to be determined after the fact but in light of information to that point. Hindsight bias fools you into thinking that if you are good at predicting the past the same will be true of the future. By taking the long view as opposed to sampling every day where noise is as much a part of it as signal, you get a greater sense of the direction of travel. 

Some forms of learning that we are not aware of – conscious and nonconscious, declarative and non-declarative. Much of risk avoidance comes from experience and is part of the second. A fascinating study by the Swiss doctor Claparede demonstrated that amnesic patients knew full well not to repeat the same mistake twice when shaking the doctor with a drawing pin stuck to their hand, a mistake they had made the week prior.  

Deductive reasoning comes from a well-established axiomatic framework like arithmetic, or inductive, verifiable through another system (like statistics). Some inductive statements are impossible to verify, and empiricism is likely to give people confidence if issued – however, a good start for making people responsible for their statements. Rhetoric can be randomly generated but not scientific reasoning.  

We do not need to be rational and scientific when it comes to the details of our daily lives – only those that can harm us and threaten our survival. Modern life appears to invite us to do the opposite.

Hegel-style verbiage-based philosophy from a scientific standpoint plain garbage, and from an artistic point of view; inferior to music. 

Is survival of the fittest a myth? At any given time, you can look like you know what you are doing – conditions can suit anyone for a while but not for the whole while – unless you do know what you are doing, much like a broken clock is right twice a day.

Warren Buffet once said, “It took us 37 years to build trust, and we could lose it all in 37 minutes.” 

Darwinian ideas are about reproductive fitness, not survival. A winner in the short term, but a loser in the long term, is just as likely to reproduce as anyone else. 

We tend to think traders survive because they are good – when they could just be lucky – traders can benefit from a market cycle as much as they can their ability –  right time, right place – fooled by randomness.

Continuous betterment might not be as important as leaps in progress. 

Paying attention to skewness: Where there is asymmetry the average (expected) and the median are unrelated.

More than 50% of people can be wealthier than average: 

9 have a net worth of £30k

1 has a net worth of £1k

The average is £27,100

9 out of 10 have an above-average wealth. 

Table 1: Frequency (probability) is irrelevant if not linked to the outcome probability, pay attention to the expectation – probability x payoff. 

Event	Probability	Outcome	Expectation
A	999/1000	$1	$.999
B	1/1000	-$10,000	-$10
		Total	-$9.001

In this example, the bet is a poor one with the expectation of -$9.001. Lots of small wins don’t make up for the big loss. Rare events exist because they are unexpected – their effects are disproportionate and unsettling, sometimes even career-ending.

Investing in something that provides small losses that you can tolerate to wait out for the big win is better than lots of small wins and a big unexpected wipeout.  

The problem with past performance: Past performance is no indicator for the future – we can’t jump from has never gone down to will never go down.  

Statements such as: 

The market never goes down 20% in a given three-month period can be tested but is completely meaningless.  

Think of it like Russian roulette: there’s a small chance of a large loss and a large chance of a small win. You’re likely to show up as a winner in almost all samples – except in the year you are dead. Don’t approach investing as a game to win, except, of course, if it is a game to win.  

Deal with randomness with a critical, open mind and change opinions with minimal shame. We need to offer conditions under which our theory can be proved wrong – that’s science. Newton moved on by Einstein, and so on.

Popperism suggests avoiding verificationism – the desire to be right or fit the model – when in reality, we’re waiting for the black swan. Compression reduces randomness making it easier for us to handle. 

An optional strategy for humans is to believe in the existence of god. If He exists we win; if not we have nothing to lose.  

Survivor bias illustrated: 10,000 fund managers are selected by the flick of a coin – half make $10k, and the other half lose $10k. The losing half gets fired. By the end of 5 years, 313 fund managers are left, not on ability but on luck. Of course, we hear from the successful how they achieved this success. 

Ergodicity and regression to the mean: In a large sample, it’s possible to have a “hot hand,” like getting 8 heads in a coin flip, for example. However, in the long term, this run will return to the mean, and the run will end – ergodicity.

You need to know the size of the population (not the sample) from where they came. In a large population, it’s much more likely that it’s luck, and that streak will end. Not all deviations come from this, but many do.  

Data mining – looking for data to fit: Consider the birthday paradox. The chance of randomly meeting someone who shares the same birthday as you 365/1. However, the chance of meeting someone at a birthday party of 23 people is 50% – any pairing will work. 

If you look for a fit between two sets of data – a specific relation, say stock price and election news that’s likely to be believable (1/365 example). But when you are looking for a fit (birthday party) then you are going to find something – small world – not really, much like throwing monkeys at a typewriter but not telling them what type of book to write.  

We follow the crowd – and if by chance something or someone is picked out of a crowd – like an actor in a room full of actors during an audition – then we believe fame is because of a particular attribute or skill due to their rise to fame – perhaps we don’t have such an attribute – when truly it’s a positive feedback loop – people follow people. 

Markets are dominated by a few – network effects – fat tails. Modeling this type of success is difficult. Modeling relies on an assumption of independence – the next decision is not dependent on the last choice – when skewed, it is. 

Tipping points might not exist in network effect, too unstable or impossible until after the fact. Important to know about them and less important to try to model them.

Success is often nonlinear but not random; we can learn something every day and get no payback until one day it clicks. Nonlinear – the straw that broke the camel’s back – a small input creates a disproportionate response. 

We find holding two ideas in equal weighting very difficult, if not impossible, let alone if it was weighted 85/15. We chose one. For a $1000 bet, we focus on the win or the loss, never both equally.  

We are likely to act irrationally because we focus on the win or the loss, never the probability. We don’t think of ourselves as 78% dead and 22% alive. 

Rules are useful and save time and effort. We can consider our brain like a rulebook; we just can’t be sure what page we are on. Biases don’t always disappear when there are incentives which means that some are not cost-savings. 

Here are some of the heuristics we fall for:  

Availability heuristic: estimating frequencies according to the ease with which frequency of events are recalled. 

Representative heuristic: Gauging the probability that a person belongs to a particular social group by assessing how similar a person’s character is to a typical group member. 

Affect heuristic: What emotions are elicited by events determine their probability in your mind.  

In Joseph LeDoux’s theory, we feel emotions (limbic) and then find an explanation for them (neocortex). Much of the opinions and assessments that we have concerning risk may be the simple result of emotions. 

Conditional probability refers to the chances that some outcome occurs given that another event has also occurred. For example, 10% of people who batter their wives go on to murder them. Of the men who murdered their wives 50% had battered them previously. Look at the outcome and its probability. 

We overvalue our knowledge and underestimate the probability of being wrong.  

Absence of evidence vs evidence of absence – not yet proven vs this does not work.  

Signal vs noise: If a cyclist is beaten by a second in a 3-week race there is no need to read anything into the result.

But if beaten by a week then take a look.

Multivariate analysis.

Look at all factors. 

Look at historical effects both in isolation and jointly.

Then look at the stability of such influence. 

Consult the test statistic. 

Isolate the factor if it’s possible to do so. 

The confidence level needs to be given to the factor itself. All this is rarely done. 

We commentate on noise – such noise does not warrant an explanation – note it and move on – one data point – not a trend or behaviour pattern just a data point. 

Percentage moved over baseline is helpful. But a 2% move is not necessarily twice as important as a 1 % move – it’s non-linear. A 6% move might be 1000 more significant. As the magnitude of change increases so too does the likelihood of it being signal rather than noise. 

Wittgenstein ruler unless you have confidence in the ruler’s reliability, if you use a ruler to measure a table you may also be using a table to measure a ruler.  

Full of superstitions, rituals, and habits we take life too seriously- imposing significance on events where there is none – or taking comfort in thinking that somehow our destiny has been crafted for us.

We are emotional, not rational. We need tricks not lectures to modulate our emotional behaviour. 

Probabilistic thinking: Most probable – view the possible contingencies as distinct and separable events with probabilities attached to each one of them – role of skepticism. 

Offer alternatives. Path dependencies – the ability to change your mind – free from last actions – every day is a clean slate. Beliefs are said to be path-dependent if the sequence of ideas is such that the first dominates. We are married to our ideas.

The emotion of attachment- purely rational behaviour – defect of the amygdala – psychopath. 

Attribution bias we attribute our success to our skills but our failures to randomness – the illusion that we are better than we are. 

Dignity as a solution – execution of a protocol of behaviour that does not depend on the immediate circumstances.  

Contribution-based on nonrandom factors – such as dentistry or cooking but contribution on the corporate ladder the higher you go the less clear the contribution. Upper management judged on results alone, and lower down the ladder, judged on process and results. Exception – risk-bearing entrepreneurs. 

Repetitiveness is key to the revelation of skill- Ergodicity- the detection of long-term properties. 

Use uncertainty to save yourself from routine – think of train times as being random versus working to a schedule and being deprived of the joys of life. Randomisation prevents us from optimising and being exceedingly efficient at the wrong things.

The richer we get the pickier we become. 

Lack of randomness makes you predictable and liable to people taking advantage of your routine – easy to read. Trying to be optimal in our enjoyment is a strain. Satisficing- a blend of satisfying and maximising. 

We favour the visible, the embedded, the personal, the narrated, and the tangible; we scorn the abstract.