- The idea is not that “it is all random”, but “it is more random than we think”. Not all or nothing. Not necessarily that Warren Buffet is not skilled, but that there can not be surety that he is skilled (there is a statistically valid alternative explanation). Do not fall for “affirming the consequent” fallacy. Necessary conditions do not prove sufficient conditions. All successful entrepreneurs are risk takers does not mean all risk takers are successful entrepreneurs!
- “Black swans” can be both bad and good rare events.
- Do not confuse luck with skill, randomness with determinism, noise with meaning, conjecture with truth, volatility with return. Literary intellectuals most confuse symbolism in coincidental occurrence of patterns (ex. ink blots resembling human figures). Economists often confuse cause and effect. Many biographies ascribe success to people who were merely “in the right place at the right time”.
- Bad information is worse than no information.
- Utopian Vision (Descartes, Paine)—implore reason and rationality, vs. Tragic Vision (Popper, Friedman, Kahneman)—people are humans and therefore flawed. Work around flaws; moralizing advice is ineffectual, it assumes humans are controlled by the cognitive instead of the emotional.
- Sometimes you can’t use KISS—simplification may not be possible without severe distortion.
- Skewness issue: it doesn’t matter how frequently something succeeds, if the failure is too costly to bear.
- On work ethics: mild success can be due to risk-conscious hard work and discipline; wild success is due to variance (luck). Think hard, not work hard—those who merely work hard eventually lose their focus and intellectual energy.
- Use probabilistic theory in average life outcomes. Take into account observed and unobserved possible paths. Over many sample paths of lives, aim for the highest expected value (steady eddie dentist over lucky rock star).
- “Mathematics is principally a tool to meditate, rather than to compute.”
- The quality of a decision should be judged by its alternative histories, not by the single outcome. Similarly, a mistake should not be judged by its outcome, but by the information and circumstances at the decision point.
- Risk detection and avoidance come from the emotional side of the brain, not the logical one. Therefore, people react to sensationalist and vivid fears more than real or probabilistic ones.
- Einstein: “common sense is nothing but a collection of misconceptions by the age of 18.” All radical discoveries were thought to be lunacies until proven by science.
- Risk managers must play politics and operate with a margin of error because their payoffs are asymmetric. Preventions of blowups are largely unseen, while prevention of profits are.
- Learn from history in two ways: 1) from the past, by reading elders, and 2) from the future, by building Monte Carlo simulations to examine alternative sample paths, unrealized outcomes. “Summing under histories.”
- People learn distinctively from non-declarative (non-conscious, experiential) memory versus declarative (conscious, textbook) memory. Some can only learn from own mistakes. The easy stuff should be learned from others’ mistakes. Amnesic patient recognizes pin in researcher hand and avoids shaking it, despite not remembering researcher.
- Hindsight bias, historical determinism: hard for people to imagine the preceding events surrounding historical events, that history while obvious in retrospect, was unpredictable a priori.
- Ergodicity: bad trades catch up with you. Over time (long sample lives), different sample paths resemble each other and are averages of shorter sample paths—mean reversion.
- The older the idea or investor, the better— it/ he has been distilled through time and demonstrated evolutionary fitness. When in doubt, it is optimal decision-making to reject any new idea, information, or method. Finding information in the noise of news is like looking for a needle in a haystack—not cost benefit positive.
- The cross-sectional problem: at any given point in time in the market, the most successful traders are often the worst traders, because their style or traits are best fit to only that market cycle.
- Bad traders share some common traits: over-reliance on some measure (economic or statistical), become married to a position, no stop loss game plan, denial or change story (short-term traders changing to “investors for the long haul”).
- Time (temporal) aggregation balances out over the long-term—noise cancels out. Darwinism means survival of the fittest on average, not for a single sample, and not for all sample paths, but fittest for the realized sample path.
- In investments and betting, it is not probability (frequency) that matters, but expectation (probability x magnitude). Do not confuse because payoffs may not be symmetrical, they may have large skews.
- The rare event is by definition unpredictable. Because if you can anticipate it, you would have already taken measures to mitigate it, and the impact would not be a black swan. Stability breeds turbulence—the peso problem.
- Statistics can confuse and fail us when the distributions are asymmetrical. The absence of an event may take very many observations to gain statistical confidence (still never gain confirmation), but the presence of an event gains confidence immediately once it occurs. The Black Swan problem: no amount of observations of white swans can prove all swans are white, but the observation of a single back swan can disprove it. In addition, distributions can have stationary problems, thereby rendering historical time series useless in forecasting.
- You cannot infer much from a single experiment in a random environment—an experiment needs a repeatability showing some causal relationship. Therefore, the success of one trade proves nothing.
- According to Karl Popper: a theory can never be verified, but it can be falsified. True scientists are always testing and trying to find refutations of their own theories. A theory that isn’t disprovable is not science at all (i.e. astrology). He disagreed that incremental information is always additive to knowledge (the basis for statistical inference).
- We are not genetically built to be rational, much less act rationally. Even Popper did not always practice what he preached.
- Use statistics to make bets, but use reasoning to manage risk. Using statistics to manage risk can expose you to black swan threats. Challenge your thesis and ensure you have covered yourself in the rare event you are wrong.
- Think independently, avoid the “firehouse effect”. By interacting and integrating too much with others, you risk resembling others in thoughts and actions.
- Survivorship bias problem: we are miscomputing the probabilities because we are not counting the disappearing losers. Being an accumulator is not sufficient (not even necessary, though higher probability) to being rich, you must accumulate the right things (currencies, asset types, etc.).
- Live in a poor neighborhood. By living in a wealthy neighborhood, you are 1) prone to miscompute the distribution of success in the entire population, 2) running the psychological wealth treadmill (lifestyle reverts to a higher set point), and 3) happiness plateaus after a certain level of wealth.
- There is no special heroism in money, particularly if the person is foolish enough to not even derive any tangible benefit from it. Becoming rich is a selfish act, not a social, moral, or even intellectual one (Buffett however, has pledged his wealth to charity).
- Counter-intuitively, a population of all bad managers would produce some fantastic track records (owing to volatility and large initial population). The larger the initial population and higher the volatility, the larger the expectation of number of excellent records.
- Adverse selection: an unsolicited investment manager coming to you has a much higher probability of being a spurious survivor (only good track records advertise) than one you analyze and choose from the entire pool of managers. Therefore, judging an unsolicited offer should be a lot more stringent than one you seek.
- Randomness does not look random! There are bound to be spurious patterns in true randomness. Ex: miracle cures of cancer due to spontaneous remission, the Da Vinci Code. Data mining and data snooping can produce spurious relationships.
- Path dependent, non-linear outcomes: winner takes all. Network externalities: chance events coupled with positive feedback rather than superiority can determine winner (ex. Microsoft). It is better to have a dozen strong supporters than a hundred admirers. Though these critical events/progressions exist, they are impossible to predict or model.
- Mathematical modeling techniques need to consider that the math may be too restrictive for finance problems, or the precision of the math can give a false sense of the solution.
- Herbert Simon believes humans are “boundedly rational”—we “satisfice”. This makes reaching decisions faster and easier.
- Normative science is prescriptive, how things should be (i.e. efficient markets); positive science is descriptive, how things are (empirical, i.e. behavioral finance).
- We should be free and not ashamed to contradict ourselves. Our decisions should reflect the best information at that given moment in time, not dependent on prior decisions or opinions.
- Beware conditional and joint probabilities when using statistics. Ex: unconditional life expectancy is 73 years; life expectancy at 75 years old, conditioned on still being alive, is nonetheless positive.
- Beware of non-linearity: a 2% market move is much more than twice as significant as a 1% move.
- Confidence levels: it is not so much the point estimate that matters so much as the degree of confidence in the forecast.
- Wittgenstein’s ruler: unless you have confidence in the accuracy of the measuring device, if you use a ruler to measure a table you may just be using the table to measure the ruler.
- Human beings like to have a causal link. It is emotionally harder to reject a hypothesis than to accept it.
- Skinner pigeon experiment: faced with random feedings, the famished birds developed sophisticated rain-dance behavior in trying to produce feedings.
Finished: Jun-2008