## Highlights - all our decisions are underpinned by an implicit assumption of a time horizon to optimize for. Whether to do unpaid work is influenced by how long we think a work relationship will last. Whether to marry our romantic partner is influenced by how long we believe our love will last. ([Location 100](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=100)) - Time horizons are the bases of our decisions – and of our regrets. Live as if you would die tomorrow, and you will regret not having pursued long-term engagements ([Location 103](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=103)) - Getting the time horizon right is fundamental for a life with few regrets. Sadly, there’s too much variance and too many unknowns for us to get it reliably right. However, we can improve our outcomes by understanding the effect of time and variance on our choices – and taking choices that don’t need our guess to be exact to work out. ([Location 106](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=106)) - Ergodicity explains that; ([Location 110](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=110)) - In skiing, and life in general, it is not the best ones who succeed. It is the best ones of those who survive. ([Location 150](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=150)) - In practice, performance is subordinate to survival. It is the fastest racer of those who survive that wins races, it is the highest performing employee who doesn’t burn out that becomes the most successful, and so on. ([Location 154](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=154)) - in a single instant of time, pure performance is all that matters. Instead, over a prolonged period of time, survival dwarfs performance. ([Location 173](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=173)) - Too often, we observe a snapshot of someone’s life and believe that we witnessed a piece of short-term performance. But if practice is required to get to a place where one can attempt that, then what we are really observing is long-term performance. ([Location 198](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=198)) - Sadly, relationships, careers, investments, and sports often share this undesirable property. ([Location 259](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=259)) - I cannot keep picking risky stocks until I get rich – a few bad results in a row, and I am broke. ([Location 275](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=275)) - Whenever an activity cannot be assumed repeatable at infinity, we should be wary of expecting to achieve its average outcome. ([Location 276](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=276)) - Game-overs are common. They include bankruptcies, injuries, severe depressions, burnouts, and break-ups of all kinds (between romantic partners, business partners, or friends). ([Location 289](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=289)) - This expectation remains correct if we increase the number of gamblers. With 600 gamblers, you will have an average of 100 deaths and $3M. Divided by 600 people, it still makes 1/6 of a death and $5,000. Therefore, the expected outcome is also called the population outcome ([Location 322](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=322)) - The chart shows how your expected financial gains from playing Russian Roulette many times are roughly zero. I call this expectation, measured across one person and many repetitions, the lifetime outcome. Its technical name is time average or time probability. ([Location 329](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=329)) - A system is ergodic[21] if its population outcome coincides with the lifetime outcome of each of its components. Otherwise, it is non-ergodic. ([Location 335](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=335)) - The practical implication is that in ergodic systems, you can use the population outcome to make optimal decisions. In non-ergodic systems, you cannot. ([Location 338](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=338)) - - Irreversibility absorbs future gains. - We call “population outcome” the outcome of many people performing an action once, and “lifetime outcome” the outcome of one person performing an action many times. If they differ, the system that produces them is non-ergodic. - You can only rely on expected outcomes if you are guaranteed a large number of repetitions. Otherwise, they are misleading. (The law of large numbers requires a large number of repetitions). - Risk aversion is rational in the presence of non-ergodicity. ([Location 417](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=417)) - in non-ergodic contexts, irreversibility limits the number of trials over time, preventing the law of large numbers from applying. ([Location 445](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=445)) - As an individual, you do not care whether the system works on average. You care if it works for you. ([Location 458](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=458)) - averages might hide local spikes in irreversibility. This is problematic when survival is based on the local. In such cases, aggregate evaluations cannot substitute local ones. ([Location 462](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=462)) - Central organizations tend to receive more input and feedback from people geographically, socially, or culturally close to them. ([Location 480](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=480)) - The closer geographically, socially, and culturally your governors are to you, the more likely they are to work on what you care.[32] Hence the saying, centralization is only efficient to the central observer. The closer someone lives to the capital, and the more his culture and social status match those of the governors, the more he will care about what the central organization is good at, and the less he will care about what it’s bad at. When evaluating centralization, make sure you ask everyone, not just those close to you. ([Location 484](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=484)) - Be wary of centralization, especially when there is no skin in the game[33] ensuring that the incentives of the center coincide with those of the peripheries. ([Location 491](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=491)) - The same happens in all large systems: change can only be driven by people who are close to those they want to change, and some problems and needs can only be collected by those who personally know those having them. Hence the need in large countries and organizations to have a capillary structure where information isn’t collected nor pushed at scale, directly linking the center to the peripheries, but rather step-by-step, layer-by-layer, between people who know each other’s context well. ([Location 506](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=506)) - Averages are a meaningful metric only when future gains offset past losses. It is often not the case. Continuing the previous example, I do not care if, next time, the train arrives 30 minutes in advance. I do not get a free flight when the train arrives in advance. Arriving in advance doesn’t offset a delay if it doesn’t offset its consequences. As a rule of thumb, we cannot rely on averages whenever there is a possibility for irreversible damage[35]. This is because the possibility of irreversible damage makes a context non-ergodic. ([Location 527](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=527)) - The point is, 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 (in this example, the founder), you do not care about the overall outcome of all gambles but only yours. ([Location 555](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=555)) - This pattern is everywhere in nature and society. For an employee, his career and health are irreplaceable, but for the company, most employees are replaceable. ([Location 571](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=571)) - In general, the survival of a population does not coincide with the survival of its members, not all of them. ([Location 578](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=578)) - As an individual, you cannot blindly rely on membership of a group for your survival. Instead, become an irreplaceable part of it. ([Location 605](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=605)) - The trick is to become a population yourself. This is easier than it seems – after all, you already are a population: of habits and beliefs. If you let the feedback you receive from your environment hurt the habits and beliefs that are holding you back, you are effectively making natural selection act inside you rather than on you.[47] You become stronger rather than extinct. You are not your habits nor your beliefs. You are their container. Their survival is not yours. What’s best for them might or might not be what’s best for you. ([Location 612](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=612)) - 1) It is not performance that is the most important factor for long-term performance but survival. 2) Something can work on average and still fail with permanent consequences. 3) What’s best for the survival of the individual isn’t necessarily what’s best for the survival of the population, and the other way around. All three assertions only apply to non-ergodic contexts. Hence the importance of understanding ergodicity. Whether you are in an ergodic world or not determines what is rational and what isn’t. ([Location 620](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=620)) - A system[49] is ergodic if, for all its components, the lifetime outcome corresponds to the population outcome. Otherwise, it is non-ergodic. ([Location 637](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=637)) - Do not trust advice about a non-ergodic context that assumes it ergodic. While evaluating advice, always ask yourself, is it assuming ergodicity? If so, how would things change removing that assumption? Fully relaying on averages and expected outcomes makes sense only in ergodic contexts. In non-ergodic ones, you want to ensure that you make decisions based on an estimate of your lifetime outcome.[50] Similarly, when we see a behavior that we deem illogical or irrational, it is often worth asking ourselves, “is it irrational for an ergodic world, but rational for a non-ergodic one?” ([Location 641](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=641)) - Ergodicity is verified if, more or less, every time that you compute a statistical measurement (across space or time), you find the same result. It means that randomness is “well shared.” Another way to see it is that, when you have ergodicity, doing N random experiments in parallel will give you the same result as doing N experiments one after the other. ([Location 657](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=657)) - Irreversibility brings non-ergodicity. After a game-over, you cannot keep playing. Game-overs are a form of irreversibility, ([Location 673](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=673)) - As Nassim Nicholas Taleb noted, in some societies, if you’re poor, you’re likely to stay poor, and if you’re rich, you’re likely to stay rich. In these societies, one cannot assume that the lifetime outcome of a person born poor will coincide with the average of the population. He is likely to stay below average. Such a society in which wealth is at least partially irreversible is non-ergodic. ([Location 676](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=676)) - Here is a simple test to estimate whether a situation is ergodic. “Are there irreversible consequences?” If yes, it is non-ergodic. ([Location 681](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=681)) - Therefore, ergodicity is not just the property of an activity. Instead, ergodicity is a property of an activity, its participants, and the exposure of the latter to the former – in short, of a system. ([Location 692](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=692)) - “My death at Russian Roulette is not ergodic for me, but it is ergodic for the system.” ([Location 695](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=695)) - Also, in his Antifragile he pointed out how, for a species to be antifragile (adapt and therefore resist change), its members must be fragile. Natural selection is a mechanism that benefits a species[56] but requires the death of some of its members. If the members were nearly invincible, then the species would lose the capacity to adapt. Ergodicity is scale-dependent. A system observed at some scale may show a different ergodicity than the same system observed at a different scale. ([Location 697](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=697)) - it means that people neglect events that don’t create a meaningful gap between lifetime and population outcomes when measured over the time horizon of their concern. ([Location 736](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=736)) - To summarize the previous page, most people do not care if an activity has a source of irreversibility. Instead, they care whether irreversibility has a non-negligible chance of manifesting within the period they see themselves performing the activity. In other words, what matters is not whether the lifetime and the population outcomes diverge over infinite time-frames – what matters is if they diverge significantly over the medium term. ([Location 742](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=742)) - In investing, the barbell strategy consists of allocating part of one’s wealth in non-risky assets and part in risky ones with high upside ([Location 798](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=798)) - The barbell strategy consists of exposing most of yourself to safe activities and a tiny bit of yourself to risky ones with high upside. This is distinct from exposing all of yourself to safe activities[64] most of the time and seldomly to risky ones with high upside. ([Location 804](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=804)) - this strategy acknowledges that, in most contexts, a small exposure to high upside brings higher returns than a moderate exposure to moderate upside. ([Location 816](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=816)) - risk management is not about prudence but about removing the risks of “game-over” so that you can be aggressive with other risks. Similarly, the barbell strategy is not about reducing risk in general. Instead, it is about limiting the part of your assets that are exposed to irreversibility. ([Location 822](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=822)) - These two intuitions, “don’t go all-in” and “payoffs determine the relative size of the bet,” summarize a betting strategy known as the Kelly Criterion, ([Location 861](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=861)) - that risk exposure must depend on payoffs and, anyway, be limited. ([Location 863](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=863)) - Moods are adaptations to environments with clustered resources. As the hunter-gatherers’ example showed, moderately moody people tend to be more efficient than moodless ones. (The keyword being “moderately” – as with almost anything, excesses are bad.) ([Location 888](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=888)) - This theory[69] of moods is interesting. It shows how moods improve effectiveness by making people spend more time on what looks promising and less time on what is likely to be a waste of time. Moreover, they reduce exposition to what is dangerous (wasting time and energy), thereby improving survival. They allow us to adapt to a world where threats and resources are not uniformly distributed. A must, in a non-ergodic world. ([Location 892](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=892)) - Moods alone would not be enough to approximate the Kelly Criterion, though. Modulating bets based on payoffs is only one of its two components. The other is to avoid going all-in. Fear is an adaptation that helps us with that. If we didn’t feel fear, a few successes in a row would cause us to be in an ecstatic mood and go all-in. Fear helps us maintaining some prudence. Similarly, a few failures in a row might cause us to abandon any will to do anything. Fear of missing out, fear of starvation, fear of being labeled as lazy, and similar fears keep us at least somehow motivated and active. ([Location 905](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=905)) - The Precautionary Principle holds that we should not take risks that endanger the whole,[72] no matter how unlikely. If we keep taking them, we are guaranteed to blow up ([Location 950](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=950)) - The lesson is: do not expose the whole to irreversibility, no matter how attractive the payoffs. ([Location 980](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=980)) - The solution is to realize that natural selection, while inevitable, can be further pushed down to lower levels. Just like companies can protect themselves by firing their unfit employees, employees can protect themselves by firing their unfit mental patterns. ([Location 1005](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1005)) - To protect the individual, natural selection must be free to act within the individual. There, it can select the beliefs and habits of his or hers that are fit (i.e., beneficial) and terminate those that are unfit (i.e., hinder his or her performance). ([Location 1008](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1008)) - I call the process described above fractalization. The name comes from the realization that natural selection acts across levels with a fractal-like pattern. ([Location 1012](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1012)) - The first strategy to manage non-ergodicity is to reduce your exposure to irreversibility. In particular, we saw how: - You can only optimize in reversible domains. Otherwise, you should first think about avoiding irreversibility. - You can reduce exposure to irreversibility by exposing only part of yourself (or of your assets, reputation, etc.). This is safer and more effective than going all-in on a medium-risk activity. - It is safer to keep a bit of yourself exposed to risk (to grab some upside) than not doing it. Otherwise, we involuntarily expose ourselves to the risks of inaction and obsolescence. - An effective tactic consists in fractal systems and decreasing the ergodicity of the lower layers to increase that of the higher ones. - Nature gave us tools to manage non-ergodicity, in the form of moods and fear. Here are a few ways in which you can apply the first strategy: - Invest your money as a barbell (most in low-downside assets, a bit in high-upside ones). - Invest your time as a barbell (most in activities that strengthen your bedrock, a bit in explorative ones). - Purposefully test your habits and beliefs to get rid of those that are negatively impacting your life or making you weak, while keeping an eye out for overoptimization to temporary trends. ([Location 1044](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1044)) - someone has skin in the game if, when wrong, he becomes the victim of his mistakes. ([Location 1070](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1070)) - Moral hazard happens when someone has incentives to increase an entity’s exposure to risk because he won’t bear the full cost of that risk. Other examples of moral hazards are parents imposing stressful career choices on their kids, ([Location 1078](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1078)) - Populations want their members to have skin in the game. This way, over time, they get rid of those that expose themselves and others to risks. ([Location 1101](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1101)) - a population can decrease its exposure to irreversibility by exposing its members to it. Whether irreversibility has a positive or negative effect depends on the layer observed. ([Location 1107](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1107)) - We saw how: - Skin in the game acts as an incentive and as a filter that removes dangerous individuals from the community and prevents dangerous behaviors from spreading. - Sociality helps aligning incentives within a community and conserving precious variance. It makes one feel like he cannot act against his community and then “jump ships.” Here are some practical applications: - Be wary of advice from people without skin in the game. They might be incompetent or not have your best interest at heart. - Be wary of envying people without skin in the game. Imitating them might expose you to problems you didn’t consider. - Careful of removing social bonding from your life or organization in the name of efficiency: it has a purpose. ([Location 1134](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1134)) - The faster a system can redistribute load the more ergodic it is, all other things equal. ([Location 1170](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1170)) - The point is, to avoid breaking down, don’t be brittle. Deform to accommodate your loads. More importantly, do not take too many commitments that prevent you from deforming. ([Location 1222](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1222)) - Conversely, localism (the opposite of centralization) wants decision-making to happen at the lowest level possible – at the community level, if possible; if not, at the town level; if not, at the province level, and so on. This allows the decision-makers to know the topic of the decision closely, to have skin in the game so that they do not take excessive or unethical risks, and to make very tailored policies. These are very valuable benefits, in particular in the long-term, because they make mistakes less likely. ([Location 1340](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1340)) - Non-ergodicity is undesirable when the lifetime outcome is lower than the population one but desirable when it’s higher. ([Location 1370](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1370)) - Examples of positive non-ergodicity Learning. If you ensure that past errors won’t happen again (i.e., you introduced positive irreversibility), good for you! Relationships. Going on a date with the same person ten times produces different outcomes than going on ten dates with ten people.[103] Customer acquisition. In some businesses, showing your product once to a thousand customers will convert few of them into clients, but showing it ten times to a hundred customers will convert more. Skin in the game. As saw in Chapter 5, in some domains, society relies on irreversible damage at the individual level. For example, bad drivers dying in car incidents or losing their driving license keeps other drivers safe. In this case, negative non-ergodicity at the individual level translates into positive ergodicity at the collective level. ([Location 1372](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1372)) - The easiest way to increase performance is to restrict the scope of its definition. For example, we can restrict the definition of fast from “fast during a whole championship” to “fast in this particular slope.” Or, we can restrict the definition of happiness from “a fulfilling career, a happy family, a good social life, and a healthy body” to “a prestigious job title” or “a coveted spouse.” ([Location 1461](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1461)) - the easiest way to hide problems is to increase the scope of measurement. ([Location 1471](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1471)) - The difference between efficiency and effectiveness is that the former is a snapshot in time, and the latter describes a lifetime. Or, to be more precise, efficiency is restricted by time (in that case, the duration of the race). In contrast, effectiveness is expanded across time, encompassing an entire lifetime. What is efficient in the short term might or might not be effective in the long term. ([Location 1495](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1495))  ## Highlights - all our decisions are underpinned by an implicit assumption of a time horizon to optimize for. Whether to do unpaid work is influenced by how long we think a work relationship will last. Whether to marry our romantic partner is influenced by how long we believe our love will last. ([Location 100](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=100)) - Time horizons are the bases of our decisions – and of our regrets. Live as if you would die tomorrow, and you will regret not having pursued long-term engagements ([Location 103](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=103)) - Getting the time horizon right is fundamental for a life with few regrets. Sadly, there’s too much variance and too many unknowns for us to get it reliably right. However, we can improve our outcomes by understanding the effect of time and variance on our choices – and taking choices that don’t need our guess to be exact to work out. ([Location 106](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=106)) - Ergodicity explains that; ([Location 110](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=110)) - In skiing, and life in general, it is not the best ones who succeed. It is the best ones of those who survive. ([Location 150](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=150)) - In practice, performance is subordinate to survival. It is the fastest racer of those who survive that wins races, it is the highest performing employee who doesn’t burn out that becomes the most successful, and so on. ([Location 154](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=154)) - in a single instant of time, pure performance is all that matters. Instead, over a prolonged period of time, survival dwarfs performance. ([Location 173](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=173)) - Too often, we observe a snapshot of someone’s life and believe that we witnessed a piece of short-term performance. But if practice is required to get to a place where one can attempt that, then what we are really observing is long-term performance. ([Location 198](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=198)) - Sadly, relationships, careers, investments, and sports often share this undesirable property. ([Location 259](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=259)) - I cannot keep picking risky stocks until I get rich – a few bad results in a row, and I am broke. ([Location 275](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=275)) - Whenever an activity cannot be assumed repeatable at infinity, we should be wary of expecting to achieve its average outcome. ([Location 276](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=276)) - Game-overs are common. They include bankruptcies, injuries, severe depressions, burnouts, and break-ups of all kinds (between romantic partners, business partners, or friends). ([Location 289](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=289)) - This expectation remains correct if we increase the number of gamblers. With 600 gamblers, you will have an average of 100 deaths and $3M. Divided by 600 people, it still makes 1/6 of a death and $5,000. Therefore, the expected outcome is also called the population outcome ([Location 322](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=322)) - The chart shows how your expected financial gains from playing Russian Roulette many times are roughly zero. I call this expectation, measured across one person and many repetitions, the lifetime outcome. Its technical name is time average or time probability. ([Location 329](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=329)) - A system is ergodic[21] if its population outcome coincides with the lifetime outcome of each of its components. Otherwise, it is non-ergodic. ([Location 335](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=335)) - The practical implication is that in ergodic systems, you can use the population outcome to make optimal decisions. In non-ergodic systems, you cannot. ([Location 338](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=338)) - - Irreversibility absorbs future gains. - We call “population outcome” the outcome of many people performing an action once, and “lifetime outcome” the outcome of one person performing an action many times. If they differ, the system that produces them is non-ergodic. - You can only rely on expected outcomes if you are guaranteed a large number of repetitions. Otherwise, they are misleading. (The law of large numbers requires a large number of repetitions). - Risk aversion is rational in the presence of non-ergodicity. ([Location 417](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=417)) - in non-ergodic contexts, irreversibility limits the number of trials over time, preventing the law of large numbers from applying. ([Location 445](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=445)) - As an individual, you do not care whether the system works on average. You care if it works for you. ([Location 458](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=458)) - averages might hide local spikes in irreversibility. This is problematic when survival is based on the local. In such cases, aggregate evaluations cannot substitute local ones. ([Location 462](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=462)) - Central organizations tend to receive more input and feedback from people geographically, socially, or culturally close to them. ([Location 480](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=480)) - The closer geographically, socially, and culturally your governors are to you, the more likely they are to work on what you care.[32] Hence the saying, centralization is only efficient to the central observer. The closer someone lives to the capital, and the more his culture and social status match those of the governors, the more he will care about what the central organization is good at, and the less he will care about what it’s bad at. When evaluating centralization, make sure you ask everyone, not just those close to you. ([Location 484](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=484)) - Be wary of centralization, especially when there is no skin in the game[33] ensuring that the incentives of the center coincide with those of the peripheries. ([Location 491](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=491)) - The same happens in all large systems: change can only be driven by people who are close to those they want to change, and some problems and needs can only be collected by those who personally know those having them. Hence the need in large countries and organizations to have a capillary structure where information isn’t collected nor pushed at scale, directly linking the center to the peripheries, but rather step-by-step, layer-by-layer, between people who know each other’s context well. ([Location 506](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=506)) - Averages are a meaningful metric only when future gains offset past losses. It is often not the case. Continuing the previous example, I do not care if, next time, the train arrives 30 minutes in advance. I do not get a free flight when the train arrives in advance. Arriving in advance doesn’t offset a delay if it doesn’t offset its consequences. As a rule of thumb, we cannot rely on averages whenever there is a possibility for irreversible damage[35]. This is because the possibility of irreversible damage makes a context non-ergodic. ([Location 527](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=527)) - The point is, 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 (in this example, the founder), you do not care about the overall outcome of all gambles but only yours. ([Location 555](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=555)) - This pattern is everywhere in nature and society. For an employee, his career and health are irreplaceable, but for the company, most employees are replaceable. ([Location 571](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=571)) - In general, the survival of a population does not coincide with the survival of its members, not all of them. ([Location 578](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=578)) - As an individual, you cannot blindly rely on membership of a group for your survival. Instead, become an irreplaceable part of it. ([Location 605](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=605)) - The trick is to become a population yourself. This is easier than it seems – after all, you already are a population: of habits and beliefs. If you let the feedback you receive from your environment hurt the habits and beliefs that are holding you back, you are effectively making natural selection act inside you rather than on you.[47] You become stronger rather than extinct. You are not your habits nor your beliefs. You are their container. Their survival is not yours. What’s best for them might or might not be what’s best for you. ([Location 612](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=612)) - 1) It is not performance that is the most important factor for long-term performance but survival. 2) Something can work on average and still fail with permanent consequences. 3) What’s best for the survival of the individual isn’t necessarily what’s best for the survival of the population, and the other way around. All three assertions only apply to non-ergodic contexts. Hence the importance of understanding ergodicity. Whether you are in an ergodic world or not determines what is rational and what isn’t. ([Location 620](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=620)) - A system[49] is ergodic if, for all its components, the lifetime outcome corresponds to the population outcome. Otherwise, it is non-ergodic. ([Location 637](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=637)) - Do not trust advice about a non-ergodic context that assumes it ergodic. While evaluating advice, always ask yourself, is it assuming ergodicity? If so, how would things change removing that assumption? Fully relaying on averages and expected outcomes makes sense only in ergodic contexts. In non-ergodic ones, you want to ensure that you make decisions based on an estimate of your lifetime outcome.[50] Similarly, when we see a behavior that we deem illogical or irrational, it is often worth asking ourselves, “is it irrational for an ergodic world, but rational for a non-ergodic one?” ([Location 641](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=641)) - Ergodicity is verified if, more or less, every time that you compute a statistical measurement (across space or time), you find the same result. It means that randomness is “well shared.” Another way to see it is that, when you have ergodicity, doing N random experiments in parallel will give you the same result as doing N experiments one after the other. ([Location 657](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=657)) - Irreversibility brings non-ergodicity. After a game-over, you cannot keep playing. Game-overs are a form of irreversibility, ([Location 673](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=673)) - As Nassim Nicholas Taleb noted, in some societies, if you’re poor, you’re likely to stay poor, and if you’re rich, you’re likely to stay rich. In these societies, one cannot assume that the lifetime outcome of a person born poor will coincide with the average of the population. He is likely to stay below average. Such a society in which wealth is at least partially irreversible is non-ergodic. ([Location 676](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=676)) - Here is a simple test to estimate whether a situation is ergodic. “Are there irreversible consequences?” If yes, it is non-ergodic. ([Location 681](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=681)) - Therefore, ergodicity is not just the property of an activity. Instead, ergodicity is a property of an activity, its participants, and the exposure of the latter to the former – in short, of a system. ([Location 692](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=692)) - “My death at Russian Roulette is not ergodic for me, but it is ergodic for the system.” ([Location 695](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=695)) - Also, in his Antifragile he pointed out how, for a species to be antifragile (adapt and therefore resist change), its members must be fragile. Natural selection is a mechanism that benefits a species[56] but requires the death of some of its members. If the members were nearly invincible, then the species would lose the capacity to adapt. Ergodicity is scale-dependent. A system observed at some scale may show a different ergodicity than the same system observed at a different scale. ([Location 697](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=697)) - it means that people neglect events that don’t create a meaningful gap between lifetime and population outcomes when measured over the time horizon of their concern. ([Location 736](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=736)) - To summarize the previous page, most people do not care if an activity has a source of irreversibility. Instead, they care whether irreversibility has a non-negligible chance of manifesting within the period they see themselves performing the activity. In other words, what matters is not whether the lifetime and the population outcomes diverge over infinite time-frames – what matters is if they diverge significantly over the medium term. ([Location 742](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=742)) - In investing, the barbell strategy consists of allocating part of one’s wealth in non-risky assets and part in risky ones with high upside ([Location 798](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=798)) - The barbell strategy consists of exposing most of yourself to safe activities and a tiny bit of yourself to risky ones with high upside. This is distinct from exposing all of yourself to safe activities[64] most of the time and seldomly to risky ones with high upside. ([Location 804](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=804)) - this strategy acknowledges that, in most contexts, a small exposure to high upside brings higher returns than a moderate exposure to moderate upside. ([Location 816](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=816)) - risk management is not about prudence but about removing the risks of “game-over” so that you can be aggressive with other risks. Similarly, the barbell strategy is not about reducing risk in general. Instead, it is about limiting the part of your assets that are exposed to irreversibility. ([Location 822](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=822)) - These two intuitions, “don’t go all-in” and “payoffs determine the relative size of the bet,” summarize a betting strategy known as the Kelly Criterion, ([Location 861](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=861)) - that risk exposure must depend on payoffs and, anyway, be limited. ([Location 863](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=863)) - Moods are adaptations to environments with clustered resources. As the hunter-gatherers’ example showed, moderately moody people tend to be more efficient than moodless ones. (The keyword being “moderately” – as with almost anything, excesses are bad.) ([Location 888](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=888)) - This theory[69] of moods is interesting. It shows how moods improve effectiveness by making people spend more time on what looks promising and less time on what is likely to be a waste of time. Moreover, they reduce exposition to what is dangerous (wasting time and energy), thereby improving survival. They allow us to adapt to a world where threats and resources are not uniformly distributed. A must, in a non-ergodic world. ([Location 892](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=892)) - Moods alone would not be enough to approximate the Kelly Criterion, though. Modulating bets based on payoffs is only one of its two components. The other is to avoid going all-in. Fear is an adaptation that helps us with that. If we didn’t feel fear, a few successes in a row would cause us to be in an ecstatic mood and go all-in. Fear helps us maintaining some prudence. Similarly, a few failures in a row might cause us to abandon any will to do anything. Fear of missing out, fear of starvation, fear of being labeled as lazy, and similar fears keep us at least somehow motivated and active. ([Location 905](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=905)) - The Precautionary Principle holds that we should not take risks that endanger the whole,[72] no matter how unlikely. If we keep taking them, we are guaranteed to blow up ([Location 950](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=950)) - The lesson is: do not expose the whole to irreversibility, no matter how attractive the payoffs. ([Location 980](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=980)) - The solution is to realize that natural selection, while inevitable, can be further pushed down to lower levels. Just like companies can protect themselves by firing their unfit employees, employees can protect themselves by firing their unfit mental patterns. ([Location 1005](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1005)) - To protect the individual, natural selection must be free to act within the individual. There, it can select the beliefs and habits of his or hers that are fit (i.e., beneficial) and terminate those that are unfit (i.e., hinder his or her performance). ([Location 1008](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1008)) - I call the process described above fractalization. The name comes from the realization that natural selection acts across levels with a fractal-like pattern. ([Location 1012](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1012)) - The first strategy to manage non-ergodicity is to reduce your exposure to irreversibility. In particular, we saw how: - You can only optimize in reversible domains. Otherwise, you should first think about avoiding irreversibility. - You can reduce exposure to irreversibility by exposing only part of yourself (or of your assets, reputation, etc.). This is safer and more effective than going all-in on a medium-risk activity. - It is safer to keep a bit of yourself exposed to risk (to grab some upside) than not doing it. Otherwise, we involuntarily expose ourselves to the risks of inaction and obsolescence. - An effective tactic consists in fractal systems and decreasing the ergodicity of the lower layers to increase that of the higher ones. - Nature gave us tools to manage non-ergodicity, in the form of moods and fear. Here are a few ways in which you can apply the first strategy: - Invest your money as a barbell (most in low-downside assets, a bit in high-upside ones). - Invest your time as a barbell (most in activities that strengthen your bedrock, a bit in explorative ones). - Purposefully test your habits and beliefs to get rid of those that are negatively impacting your life or making you weak, while keeping an eye out for overoptimization to temporary trends. ([Location 1044](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1044)) - someone has skin in the game if, when wrong, he becomes the victim of his mistakes. ([Location 1070](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1070)) - Moral hazard happens when someone has incentives to increase an entity’s exposure to risk because he won’t bear the full cost of that risk. Other examples of moral hazards are parents imposing stressful career choices on their kids, ([Location 1078](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1078)) - Populations want their members to have skin in the game. This way, over time, they get rid of those that expose themselves and others to risks. ([Location 1101](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1101)) - a population can decrease its exposure to irreversibility by exposing its members to it. Whether irreversibility has a positive or negative effect depends on the layer observed. ([Location 1107](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1107)) - We saw how: - Skin in the game acts as an incentive and as a filter that removes dangerous individuals from the community and prevents dangerous behaviors from spreading. - Sociality helps aligning incentives within a community and conserving precious variance. It makes one feel like he cannot act against his community and then “jump ships.” Here are some practical applications: - Be wary of advice from people without skin in the game. They might be incompetent or not have your best interest at heart. - Be wary of envying people without skin in the game. Imitating them might expose you to problems you didn’t consider. - Careful of removing social bonding from your life or organization in the name of efficiency: it has a purpose. ([Location 1134](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1134)) - The faster a system can redistribute load the more ergodic it is, all other things equal. ([Location 1170](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1170)) - The point is, to avoid breaking down, don’t be brittle. Deform to accommodate your loads. More importantly, do not take too many commitments that prevent you from deforming. ([Location 1222](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1222)) - Conversely, localism (the opposite of centralization) wants decision-making to happen at the lowest level possible – at the community level, if possible; if not, at the town level; if not, at the province level, and so on. This allows the decision-makers to know the topic of the decision closely, to have skin in the game so that they do not take excessive or unethical risks, and to make very tailored policies. These are very valuable benefits, in particular in the long-term, because they make mistakes less likely. ([Location 1340](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1340)) - Non-ergodicity is undesirable when the lifetime outcome is lower than the population one but desirable when it’s higher. ([Location 1370](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1370)) - Examples of positive non-ergodicity Learning. If you ensure that past errors won’t happen again (i.e., you introduced positive irreversibility), good for you! Relationships. Going on a date with the same person ten times produces different outcomes than going on ten dates with ten people.[103] Customer acquisition. In some businesses, showing your product once to a thousand customers will convert few of them into clients, but showing it ten times to a hundred customers will convert more. Skin in the game. As saw in Chapter 5, in some domains, society relies on irreversible damage at the individual level. For example, bad drivers dying in car incidents or losing their driving license keeps other drivers safe. In this case, negative non-ergodicity at the individual level translates into positive ergodicity at the collective level. ([Location 1372](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1372)) - The easiest way to increase performance is to restrict the scope of its definition. For example, we can restrict the definition of fast from “fast during a whole championship” to “fast in this particular slope.” Or, we can restrict the definition of happiness from “a fulfilling career, a happy family, a good social life, and a healthy body” to “a prestigious job title” or “a coveted spouse.” ([Location 1461](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1461)) - the easiest way to hide problems is to increase the scope of measurement. ([Location 1471](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1471)) - The difference between efficiency and effectiveness is that the former is a snapshot in time, and the latter describes a lifetime. Or, to be more precise, efficiency is restricted by time (in that case, the duration of the race). In contrast, effectiveness is expanded across time, encompassing an entire lifetime. What is efficient in the short term might or might not be effective in the long term. ([Location 1495](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1495))  ## Highlights - all our decisions are underpinned by an implicit assumption of a time horizon to optimize for. Whether to do unpaid work is influenced by how long we think a work relationship will last. Whether to marry our romantic partner is influenced by how long we believe our love will last. ([Location 100](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=100)) - Time horizons are the bases of our decisions – and of our regrets. Live as if you would die tomorrow, and you will regret not having pursued long-term engagements ([Location 103](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=103)) - Getting the time horizon right is fundamental for a life with few regrets. Sadly, there’s too much variance and too many unknowns for us to get it reliably right. However, we can improve our outcomes by understanding the effect of time and variance on our choices – and taking choices that don’t need our guess to be exact to work out. ([Location 106](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=106)) - Ergodicity explains that; ([Location 110](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=110)) - In skiing, and life in general, it is not the best ones who succeed. It is the best ones of those who survive. ([Location 150](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=150)) - In practice, performance is subordinate to survival. It is the fastest racer of those who survive that wins races, it is the highest performing employee who doesn’t burn out that becomes the most successful, and so on. ([Location 154](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=154)) - in a single instant of time, pure performance is all that matters. Instead, over a prolonged period of time, survival dwarfs performance. ([Location 173](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=173)) - Too often, we observe a snapshot of someone’s life and believe that we witnessed a piece of short-term performance. But if practice is required to get to a place where one can attempt that, then what we are really observing is long-term performance. ([Location 198](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=198)) - Sadly, relationships, careers, investments, and sports often share this undesirable property. ([Location 259](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=259)) - I cannot keep picking risky stocks until I get rich – a few bad results in a row, and I am broke. ([Location 275](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=275)) - Whenever an activity cannot be assumed repeatable at infinity, we should be wary of expecting to achieve its average outcome. ([Location 276](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=276)) - Game-overs are common. They include bankruptcies, injuries, severe depressions, burnouts, and break-ups of all kinds (between romantic partners, business partners, or friends). ([Location 289](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=289)) - This expectation remains correct if we increase the number of gamblers. With 600 gamblers, you will have an average of 100 deaths and $3M. Divided by 600 people, it still makes 1/6 of a death and $5,000. Therefore, the expected outcome is also called the population outcome ([Location 322](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=322)) - The chart shows how your expected financial gains from playing Russian Roulette many times are roughly zero. I call this expectation, measured across one person and many repetitions, the lifetime outcome. Its technical name is time average or time probability. ([Location 329](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=329)) - A system is ergodic[21] if its population outcome coincides with the lifetime outcome of each of its components. Otherwise, it is non-ergodic. ([Location 335](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=335)) - The practical implication is that in ergodic systems, you can use the population outcome to make optimal decisions. In non-ergodic systems, you cannot. ([Location 338](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=338)) - - Irreversibility absorbs future gains. - We call “population outcome” the outcome of many people performing an action once, and “lifetime outcome” the outcome of one person performing an action many times. If they differ, the system that produces them is non-ergodic. - You can only rely on expected outcomes if you are guaranteed a large number of repetitions. Otherwise, they are misleading. (The law of large numbers requires a large number of repetitions). - Risk aversion is rational in the presence of non-ergodicity. ([Location 417](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=417)) - in non-ergodic contexts, irreversibility limits the number of trials over time, preventing the law of large numbers from applying. ([Location 445](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=445)) - As an individual, you do not care whether the system works on average. You care if it works for you. ([Location 458](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=458)) - averages might hide local spikes in irreversibility. This is problematic when survival is based on the local. In such cases, aggregate evaluations cannot substitute local ones. ([Location 462](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=462)) - Central organizations tend to receive more input and feedback from people geographically, socially, or culturally close to them. ([Location 480](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=480)) - The closer geographically, socially, and culturally your governors are to you, the more likely they are to work on what you care.[32] Hence the saying, centralization is only efficient to the central observer. The closer someone lives to the capital, and the more his culture and social status match those of the governors, the more he will care about what the central organization is good at, and the less he will care about what it’s bad at. When evaluating centralization, make sure you ask everyone, not just those close to you. ([Location 484](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=484)) - Be wary of centralization, especially when there is no skin in the game[33] ensuring that the incentives of the center coincide with those of the peripheries. ([Location 491](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=491)) - The same happens in all large systems: change can only be driven by people who are close to those they want to change, and some problems and needs can only be collected by those who personally know those having them. Hence the need in large countries and organizations to have a capillary structure where information isn’t collected nor pushed at scale, directly linking the center to the peripheries, but rather step-by-step, layer-by-layer, between people who know each other’s context well. ([Location 506](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=506)) - Averages are a meaningful metric only when future gains offset past losses. It is often not the case. Continuing the previous example, I do not care if, next time, the train arrives 30 minutes in advance. I do not get a free flight when the train arrives in advance. Arriving in advance doesn’t offset a delay if it doesn’t offset its consequences. As a rule of thumb, we cannot rely on averages whenever there is a possibility for irreversible damage[35]. This is because the possibility of irreversible damage makes a context non-ergodic. ([Location 527](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=527)) - The point is, 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 (in this example, the founder), you do not care about the overall outcome of all gambles but only yours. ([Location 555](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=555)) - This pattern is everywhere in nature and society. For an employee, his career and health are irreplaceable, but for the company, most employees are replaceable. ([Location 571](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=571)) - In general, the survival of a population does not coincide with the survival of its members, not all of them. ([Location 578](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=578)) - As an individual, you cannot blindly rely on membership of a group for your survival. Instead, become an irreplaceable part of it. ([Location 605](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=605)) - The trick is to become a population yourself. This is easier than it seems – after all, you already are a population: of habits and beliefs. If you let the feedback you receive from your environment hurt the habits and beliefs that are holding you back, you are effectively making natural selection act inside you rather than on you.[47] You become stronger rather than extinct. You are not your habits nor your beliefs. You are their container. Their survival is not yours. What’s best for them might or might not be what’s best for you. ([Location 612](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=612)) - 1) It is not performance that is the most important factor for long-term performance but survival. 2) Something can work on average and still fail with permanent consequences. 3) What’s best for the survival of the individual isn’t necessarily what’s best for the survival of the population, and the other way around. All three assertions only apply to non-ergodic contexts. Hence the importance of understanding ergodicity. Whether you are in an ergodic world or not determines what is rational and what isn’t. ([Location 620](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=620)) - A system[49] is ergodic if, for all its components, the lifetime outcome corresponds to the population outcome. Otherwise, it is non-ergodic. ([Location 637](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=637)) - Do not trust advice about a non-ergodic context that assumes it ergodic. While evaluating advice, always ask yourself, is it assuming ergodicity? If so, how would things change removing that assumption? Fully relaying on averages and expected outcomes makes sense only in ergodic contexts. In non-ergodic ones, you want to ensure that you make decisions based on an estimate of your lifetime outcome.[50] Similarly, when we see a behavior that we deem illogical or irrational, it is often worth asking ourselves, “is it irrational for an ergodic world, but rational for a non-ergodic one?” ([Location 641](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=641)) - Ergodicity is verified if, more or less, every time that you compute a statistical measurement (across space or time), you find the same result. It means that randomness is “well shared.” Another way to see it is that, when you have ergodicity, doing N random experiments in parallel will give you the same result as doing N experiments one after the other. ([Location 657](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=657)) - Irreversibility brings non-ergodicity. After a game-over, you cannot keep playing. Game-overs are a form of irreversibility, ([Location 673](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=673)) - As Nassim Nicholas Taleb noted, in some societies, if you’re poor, you’re likely to stay poor, and if you’re rich, you’re likely to stay rich. In these societies, one cannot assume that the lifetime outcome of a person born poor will coincide with the average of the population. He is likely to stay below average. Such a society in which wealth is at least partially irreversible is non-ergodic. ([Location 676](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=676)) - Here is a simple test to estimate whether a situation is ergodic. “Are there irreversible consequences?” If yes, it is non-ergodic. ([Location 681](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=681)) - Therefore, ergodicity is not just the property of an activity. Instead, ergodicity is a property of an activity, its participants, and the exposure of the latter to the former – in short, of a system. ([Location 692](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=692)) - “My death at Russian Roulette is not ergodic for me, but it is ergodic for the system.” ([Location 695](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=695)) - Also, in his Antifragile he pointed out how, for a species to be antifragile (adapt and therefore resist change), its members must be fragile. Natural selection is a mechanism that benefits a species[56] but requires the death of some of its members. If the members were nearly invincible, then the species would lose the capacity to adapt. Ergodicity is scale-dependent. A system observed at some scale may show a different ergodicity than the same system observed at a different scale. ([Location 697](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=697)) - it means that people neglect events that don’t create a meaningful gap between lifetime and population outcomes when measured over the time horizon of their concern. ([Location 736](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=736)) - To summarize the previous page, most people do not care if an activity has a source of irreversibility. Instead, they care whether irreversibility has a non-negligible chance of manifesting within the period they see themselves performing the activity. In other words, what matters is not whether the lifetime and the population outcomes diverge over infinite time-frames – what matters is if they diverge significantly over the medium term. ([Location 742](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=742)) - In investing, the barbell strategy consists of allocating part of one’s wealth in non-risky assets and part in risky ones with high upside ([Location 798](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=798)) - The barbell strategy consists of exposing most of yourself to safe activities and a tiny bit of yourself to risky ones with high upside. This is distinct from exposing all of yourself to safe activities[64] most of the time and seldomly to risky ones with high upside. ([Location 804](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=804)) - this strategy acknowledges that, in most contexts, a small exposure to high upside brings higher returns than a moderate exposure to moderate upside. ([Location 816](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=816)) - risk management is not about prudence but about removing the risks of “game-over” so that you can be aggressive with other risks. Similarly, the barbell strategy is not about reducing risk in general. Instead, it is about limiting the part of your assets that are exposed to irreversibility. ([Location 822](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=822)) - These two intuitions, “don’t go all-in” and “payoffs determine the relative size of the bet,” summarize a betting strategy known as the Kelly Criterion, ([Location 861](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=861)) - that risk exposure must depend on payoffs and, anyway, be limited. ([Location 863](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=863)) - Moods are adaptations to environments with clustered resources. As the hunter-gatherers’ example showed, moderately moody people tend to be more efficient than moodless ones. (The keyword being “moderately” – as with almost anything, excesses are bad.) ([Location 888](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=888)) - This theory[69] of moods is interesting. It shows how moods improve effectiveness by making people spend more time on what looks promising and less time on what is likely to be a waste of time. Moreover, they reduce exposition to what is dangerous (wasting time and energy), thereby improving survival. They allow us to adapt to a world where threats and resources are not uniformly distributed. A must, in a non-ergodic world. ([Location 892](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=892)) - Moods alone would not be enough to approximate the Kelly Criterion, though. Modulating bets based on payoffs is only one of its two components. The other is to avoid going all-in. Fear is an adaptation that helps us with that. If we didn’t feel fear, a few successes in a row would cause us to be in an ecstatic mood and go all-in. Fear helps us maintaining some prudence. Similarly, a few failures in a row might cause us to abandon any will to do anything. Fear of missing out, fear of starvation, fear of being labeled as lazy, and similar fears keep us at least somehow motivated and active. ([Location 905](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=905)) - The Precautionary Principle holds that we should not take risks that endanger the whole,[72] no matter how unlikely. If we keep taking them, we are guaranteed to blow up ([Location 950](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=950)) - The lesson is: do not expose the whole to irreversibility, no matter how attractive the payoffs. ([Location 980](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=980)) - The solution is to realize that natural selection, while inevitable, can be further pushed down to lower levels. Just like companies can protect themselves by firing their unfit employees, employees can protect themselves by firing their unfit mental patterns. ([Location 1005](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1005)) - To protect the individual, natural selection must be free to act within the individual. There, it can select the beliefs and habits of his or hers that are fit (i.e., beneficial) and terminate those that are unfit (i.e., hinder his or her performance). ([Location 1008](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1008)) - I call the process described above fractalization. The name comes from the realization that natural selection acts across levels with a fractal-like pattern. ([Location 1012](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1012)) - The first strategy to manage non-ergodicity is to reduce your exposure to irreversibility. In particular, we saw how: - You can only optimize in reversible domains. Otherwise, you should first think about avoiding irreversibility. - You can reduce exposure to irreversibility by exposing only part of yourself (or of your assets, reputation, etc.). This is safer and more effective than going all-in on a medium-risk activity. - It is safer to keep a bit of yourself exposed to risk (to grab some upside) than not doing it. Otherwise, we involuntarily expose ourselves to the risks of inaction and obsolescence. - An effective tactic consists in fractal systems and decreasing the ergodicity of the lower layers to increase that of the higher ones. - Nature gave us tools to manage non-ergodicity, in the form of moods and fear. Here are a few ways in which you can apply the first strategy: - Invest your money as a barbell (most in low-downside assets, a bit in high-upside ones). - Invest your time as a barbell (most in activities that strengthen your bedrock, a bit in explorative ones). - Purposefully test your habits and beliefs to get rid of those that are negatively impacting your life or making you weak, while keeping an eye out for overoptimization to temporary trends. ([Location 1044](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1044)) - someone has skin in the game if, when wrong, he becomes the victim of his mistakes. ([Location 1070](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1070)) - Moral hazard happens when someone has incentives to increase an entity’s exposure to risk because he won’t bear the full cost of that risk. Other examples of moral hazards are parents imposing stressful career choices on their kids, ([Location 1078](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1078)) - Populations want their members to have skin in the game. This way, over time, they get rid of those that expose themselves and others to risks. ([Location 1101](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1101)) - a population can decrease its exposure to irreversibility by exposing its members to it. Whether irreversibility has a positive or negative effect depends on the layer observed. ([Location 1107](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1107)) - We saw how: - Skin in the game acts as an incentive and as a filter that removes dangerous individuals from the community and prevents dangerous behaviors from spreading. - Sociality helps aligning incentives within a community and conserving precious variance. It makes one feel like he cannot act against his community and then “jump ships.” Here are some practical applications: - Be wary of advice from people without skin in the game. They might be incompetent or not have your best interest at heart. - Be wary of envying people without skin in the game. Imitating them might expose you to problems you didn’t consider. - Careful of removing social bonding from your life or organization in the name of efficiency: it has a purpose. ([Location 1134](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1134)) - The faster a system can redistribute load the more ergodic it is, all other things equal. ([Location 1170](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1170)) - The point is, to avoid breaking down, don’t be brittle. Deform to accommodate your loads. More importantly, do not take too many commitments that prevent you from deforming. ([Location 1222](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1222)) - Conversely, localism (the opposite of centralization) wants decision-making to happen at the lowest level possible – at the community level, if possible; if not, at the town level; if not, at the province level, and so on. This allows the decision-makers to know the topic of the decision closely, to have skin in the game so that they do not take excessive or unethical risks, and to make very tailored policies. These are very valuable benefits, in particular in the long-term, because they make mistakes less likely. ([Location 1340](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1340)) - Non-ergodicity is undesirable when the lifetime outcome is lower than the population one but desirable when it’s higher. ([Location 1370](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1370)) - Examples of positive non-ergodicity Learning. If you ensure that past errors won’t happen again (i.e., you introduced positive irreversibility), good for you! Relationships. Going on a date with the same person ten times produces different outcomes than going on ten dates with ten people.[103] Customer acquisition. In some businesses, showing your product once to a thousand customers will convert few of them into clients, but showing it ten times to a hundred customers will convert more. Skin in the game. As saw in Chapter 5, in some domains, society relies on irreversible damage at the individual level. For example, bad drivers dying in car incidents or losing their driving license keeps other drivers safe. In this case, negative non-ergodicity at the individual level translates into positive ergodicity at the collective level. ([Location 1372](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1372)) - The easiest way to increase performance is to restrict the scope of its definition. For example, we can restrict the definition of fast from “fast during a whole championship” to “fast in this particular slope.” Or, we can restrict the definition of happiness from “a fulfilling career, a happy family, a good social life, and a healthy body” to “a prestigious job title” or “a coveted spouse.” ([Location 1461](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1461)) - the easiest way to hide problems is to increase the scope of measurement. ([Location 1471](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1471)) - The difference between efficiency and effectiveness is that the former is a snapshot in time, and the latter describes a lifetime. Or, to be more precise, efficiency is restricted by time (in that case, the duration of the race). In contrast, effectiveness is expanded across time, encompassing an entire lifetime. What is efficient in the short term might or might not be effective in the long term. ([Location 1495](https://readwise.io/to_kindle?action=open&asin=B09PLJPM95&location=1495))