Buy the Book: The book offers a crash course on the concept of ergodicity. Through simple examples, the author shows the importance of recognizing debilitating risks from recoverable risks and why survival matters more than performance for long-term success.
The Notes
- Understanding ergodicity leads to decisions that minimize regret but maximize long-term potential.
- Time and variance have a big impact on decision outcomes. Ignoring that leads down a path where decisions must be exact to work out.
- Survival matters more than performance in determining long-term success. To be the greatest at anything, you must survive the trials, tribulations, injuries, mistakes, drawdowns, etc. that knock others out of the “race” before it starts.
- Performance matters most for success in a single race.
- Survival matters most for long-term success over many races.
- For example, a single injury can end a runner’s chance of racing again. Avoiding serious injury is key to long-term racing success.
- “In any endeavor in which success depends on you accumulating some kind of resource (money, skill, connections, trust, etc.), do not maximize growth regardless of survival. Instead, maximize growth that conserves survival.
- Ergodicity is the study of how short-term consequences affect the long-term.
- The chance of irreversible consequences must be considered in any repeatable activity. Ex: bankruptcy, injury, depression, burnout, break-ups, trust, etc.
- The goal is to separate risks into debilitating risks and recoverable risks. Avoid debilitating risks at all costs or reduce the exposure to debilitating risks to the point of being recoverable.
- Russian Roulette
- Is an example of how people misunderstand the long-term odds of success.
- The game is played with a 6-cylinder gun. A player has a 5-in-6 chance of winning (surviving) any one round.
- If, in addition, you win $6,000, for every round you win, you get an expected value of $5,000 ($6,000 x 5/6).
- How much do you win if you play 100 times? The correct answer is $0, not $500,000.
- The odds change dramatically when considering successive rounds.
- The expected return over time peaks in round 6 and drops dramatically after, eventually hitting zero.
- A player’s odds of survival drop to practically zero at about round 30 (assuming they survive that long). Death is certain.
- It’s an example of how odds or averages matter differently in the short and long term.
- Law of Large Numbers: suggests that fewer trials increase the chance of deviating from the average outcome but as the number of trials increases, the results converge on its average. This is only true if the activity allows for infinite repeatable trials. If it only allows for a limited number of trials, the average should not be expected.
- Population Outcome = expected outcome if a single event was performed by an infinite number of participants once. Ex: the expected outcome of 1,000 people flipping a coin once.
- Lifetime Outcome = expected outcome if a single participant repeats an event an infinite number of times. Ex: 1 person flipping a coin 1,000 times.
- Ergodic Systems = systems where the population outcome equals the lifetime outcome. The population outcome can be used to make optimal decisions.
- Non-Ergodic Systems = systems where the population outcome differs from the lifetime outcome. The lifetime outcome must be considered when making decisions.
- Non-ergodicity might explain why people are rationally risk-averse in games where they appear to have an advantage.
- Ex: A coin flipping game where heads you win $1,000 but tails you lose $950.
- Expected return of $25 = ($1000 x 50%) – ($950 x 50%).
- You’re expected to win $25 on each coin flip yet research shows that most people decline. Behavioral economists saw this as “people are irrationally risk-averse.”
- The expected return only works on average if the players have unlimited cash to bet. Unlimited cash means unlimited coin flips where the population outcome converges on the lifetime outcome.
- Without unlimited cash, going broke is possible. A limited amount of cash limits the coin flips, and introduces the possibility of losing everything.
- In two tosses, a player with $1,000 can:
- Win both tosses: win $2,000, or
- Win the first toss, lose the second toss: win $50, or
- Lose the first toss, no second toss: lose $950, or
- Lose the first toss, no second toss: lose $950
- Expected outcome: $37.50 (compared to $50 with infinite cash)
- A 25% chance to win a lot, a 25% chance to win a little, but a 50% chance of losing almost everything.
- Irrational behavior in an ergodic system may be rational in a non-ergodic system.
- “The part of the person that we envy doesn’t exist without the rest of that person. If we aren’t willing to trade places with them completely — their life, their body, their thoughts — then there is nothing to be envious about.” — Naval Ravikant
- “It is…pointless to envy someone with whom you wouldn’t trade places in all parallel universes — include those in which his gambles didn’t pay off… Do you desire to take his gambles, or do you only desire the winning outcome?”
- Non-ergodicity exists in time and space. Averages at the population level can hide wide deviations at the local or individual level.
- Ex: for the average person, the local economy is much more important than the nationwide economic average. Differences in gas prices, healthcare, or education are more important locally. People do not care if the system works on average, they care that it works for them.
- Centralized organizations make decisions for the whole that improve things on average. But something that works on average, does not guarantee it works everywhere. The solution is to use localized data and bring decisions down to the local level.
- “The best strategy depends on whether you are the gambler or the gamble. If you are the gambler, you do not care about each gamble making money. You care about the aggregate of all gambles making money. Conversely, if you are the gamble…you do not care about the overall outcome of all gambles but only yours.”
- The gambler = ergodic
- The gamble = non-ergodic
- Non-ergodicity exists at the societal and biological levels.
- What’s best for the survival of a species may not be the same for the survival of one of its members.
- What is best for a company, may not be the same as what’s best for you the employee.
- Being irreplaceable ensures survival.
- Ways to Ensure Survival
- Diversify bets instead of going all-in. Diversification reduces your exposure to permanent loss.
- Limit downside: hedging, insurance, etc.
- Set aside reserves: cash, time, etc.
- Look out for non-ergodicity (irreversibility) over shorter intermediate time frames as opposed to infinite time frames.
- Look for ways to increase the ergodicity of the systems, businesses, investments, etc. you’re in.
- Barbell Strategy:
- Expose the majority of your time, money, and work to safe activities with only a small amount of time, money, and work to risky activities. The risky activities create an opportunity for a high payoff but the lower exposure limits the downside.
- The reduced chance of permanent loss allows for a small exposure to more aggressive activities. So long as the small exposure of time, money, etc. is something you can afford to lose.
- The tradeoff is that by limiting risk you limit upside.
- Kelly Criterion
- A betting strategy that finds the optimal betting size based on the odds and bankroll in an attempt to maximize wealth over time.
- Risk exposure depends on the odds/payoffs of a bet.
- Ludic Fallacy: a mistake of applying the rules of a game to real-life situations. It ignores randomness.
- Precautionary Principle: risk management strategy that prioritizes caution whenever there is a lack of knowledge or uncertainty about potential harm. It’s a preemptive strategy to avoid destroying the whole. Ex: Viruses/epidemics are a reason to be overly cautious until more information is known.
- Natural Selection
- Is non-ergodic.
- What’s best for the survival of a species is not always what’s best for you.
- Natural selection is present in the business world. The best companies adapt, change, and grow, the worst companies fail.
- It’s also present inside businesses. The “best” employees rise to the top, the worst get fired.
- At all levels, natural selection is avoided by being too valuable/important to replace. At the individual level, it’s best done by removing bad habits and beliefs — constant improvement.
- Skin in the Game
- Sharing of risk.
- Reduces moral hazard.
- Incentivizes irreversability.
- Protects the individual and the population.
- Avoid advice from those without skin in the game.
- Load Redistribution
- Ensures stability in a system by redistributing demand, workload, money, time, etc. when one area experiences overload or failure.
- It prevents systemic risks from cascading failures.
- Manages non-ergodicity — differences between the local and systemwide averages.
- Flexibility prevents systemic risks.
- Financial Redistribution
- Works when all involved benefit and share the risk of failure.
- Reduces the chance of total and partial loss.
- Ex: social insurance programs (unemployment benefits, social security, etc.), taxes, insurance, portfolio rebalancing.
- Pre-Mortems: a study done before a project starts, that assumes it failed, to figure out how it failed and how to avoid it.
- Post-Mortems: a study after a project ends to find out what, if anything, went wrong and how to improve on it next time.
- Behavioral change is non-ergodic. Habits are formed based on the frequency of repetition in a short period.
- “The easiest way to increase short-term performance is to do so at the cost of long-term performance. It is hardly a tradeoff you want to make.”
- Problems can be hidden in broad averages or other measures. Those problems fester, grow, and become irreversible.
- Warren Buffett made over 99% of his wealth after age 52. His skill in avoiding permanent losses was more important to that achievement than his skill in generating returns.
- Enron
- Won Fortune magazine’s “Most Innovative Company” award for the 6 years from 1996 to 2001.
- It filed for bankruptcy in 2001.
- “Sustainability is often a larger obstacle to performance than talent.”
- Leading Indicators: measure what might influence future performance and may help spot problems that lead to irreversible damage. Leading indicators increase ergodicity.
- Lagging Indicators: measure performance that already happened. Lagging indicators decrease ergodicity.
- Tragedy of the Commons
- An idea that if people have unlimited access to a resource they will eventually overuse it to the point of destruction.
- Non-ergodic.
- Ex: A lake next to town will never run out of fish if 1 person fishes out 1 fish per day. But if the entire town does it, the fish population will be decimated at some point. The maximum sustainable amount of fish that can pulled out of the lake depends on the reproductive rate of the fish. Fishing at any rate higher than the reproductive rate leads to a shrinking fish population and eventually to no fish.
- Fatigue Failure: the point where the damage rate exceeds the recovery rate.
- Avoiding fatigue failure increases ergodicity.