Taleb Lessons

Nassim Taleb is a famous author who consolidated and communicated many useful ideas. His books are worth reading but you’ll need to endure his admonishing tone.

These ideas are from an interview about his book Antifragile:

  • “The general principle of antifragility, it is much better to do things you cannot explain than explain things you cannot do.” – Nassim Taleb

  • In opaque systems where cause and effect are difficult to associate, trial and error will yield the greatest outcomes. The trial and error is defined by defined cost or downside and the ‘option’ to keep what works. “We have the option, not the obligation to keep the result, which allows us to retain the upper bound and be unaffected by adverse outcomes”. We do not understand how to create great cuisine from chemical formulas but thru trial and error our culinary repertoire evolves. In the world of research, another complex system, we rely on the methodical practice of trial and error as a strategy. By doing this we are not only capable of discovering cures before even understanding how they work, but we can proceed in many directions at once informed by the discovery of the most recent node.

  • Systems that exhibit an asymmetry between gains and losses for a given level of stimulus are described as having a ‘convexity bias’. In math, this is curvature, in financial options it is known as gamma, and in physics, it is the second derivative or acceleration. Antifragile represents a class of phenomena that possess the ‘convexity bias’ property and therefore benefit from disorder or stressors. Fragile phenomenona  hate disorder. A cup of coffee will not tolerate much disorder before spilling or breaking. In an antifragile system, we maximize the result via optionality and the value of the optionality increases with the volatility in the system since the upside expands while the downside remains fixed. Recognizing whether the system you are operating in is fragile or antifragile should guide your approach to maximizing results and/or minimizing risk.

    1. Convexity can often be attained much more cheaply than knowledge. For example, lowering your cost per trial allows you to stay in a game longer which can maximize your chance to be exposed to a convex outcome. Starting 5 low cost businesses will expose you to much greater upside than becoming even a highly paid hourly worker like a dentist. The option to shed the businesses that fail and double down on the businesses that thrive derives its value from your worst case being a zero but unlimited upside. Likewise, a person with a highly marketable resume can derive more value from the fact that they can always ‘fall back’ on a job while they pursue a more speculative endeavor. The resume is valuable not because of the outcome of the job it maps to but because it has lowered the opportunity cost of taking a chance elsewhere.

    2. A complex system favors a “1/N” strategy…ie equal weighted small bets to benefit from ‘fat tails’ of outcome distributions. Venture capital model. Minimize cost per trial, maximize trials. The final outcome is accretive but any single bet is unpredictable.

    3. Maximize serial optionality. Stay flexible and avoid a “long highway with no exits”. 5 consecutive 1-year options are more valuable than a single 5-year option (a baseball player’s career is fragile because of injury risk…they will take a haircut to receive a 5-year guaranteed contract).

    4. Tinkering > fundamental/theoretical research for generating breakthroughs. The less linear the domain the more this is true.

These ideas are from an interview about his book Skin In The Game:

The concept of Skin In The Game (SITG) mirrors evolution in nature.

  • SITG as a filtering tool not just to remark on one-sided incentive/disincentive but its central attribute of symmetry is desirable to conserve on a collective level because:
    • When an agent’s outcomes are not appropriately aligned with their risks, we throw a wrench into the gears of evolution since we fail to weed out poor behavior. Systems learn at the collective level not the level of the individual. They learn from a repeated game of interactions.
      •  A restaurant faced with intense competition will likely be much better than a monopolistic food stand in a theme park or cafeteria in a school.
      • Why can we create 2 line highways without a divider when it is so simple for any individual driver to simply turn the wheel into oncoming traffic. Bad drivers have their licenses suspended and also kill themselves. Drivers have SITG
      • The selection pressures of SITG apply to tradespeople, engineers, and ‘hard’ scientists but we do not enjoy the benefit of this filter in fields where signal-to-noise is low. This allows for a higher degree of bs and opportunity to profit disproportionately if the system without the filter achieves large scale or influence (politicians, bureaucrats, too big to fail bankers, and tenured soft-science academics being his favorite targets). “The incentive is to be published on the right topic in the right journals, with well sounding arguments, under easily some contrived empiricism, in order to beat the metrics.” When architects judge architects the building aesthetics decline.

  • The prescription is less scale and more decentralization. The more micro the more apparent the skill (or lack of). Artisans eat what they kill and incentives are properly aligned with satisfying their clients. The more scale we introduce the more layers of principal-agent problems emerge.

  • Courage cannot be faked; the warrior bore the risk of his deserved glory in the service of his countryman. The ‘primacy of the risk-taker’ has been a feature of nearly all human civilization. When we reward leaders who did not bear commensurate risks we undermine virtue. Society frays as the truly virtuous/courageous bristles as they watch.

  • Don’t pick the Hollywood version of a surgeon. [Kris: This is reminiscent of Berkson’s Paradox which says the correlation of 2 related properties can flip when the “range is restricted”. For example, being good at basketball and being tall are correlated. But when you narrow the range to the NBA, then it becomes clear that all else equal a shorter player likely has more skill than a taller player because for them to have made it to the “restricted range” or universe of the NBA they had to overcome their height disadvantage]

Leave a Reply