This governs what you do when the thing that will actually determine the outcome is not on anybody's list. Taleb's claim is that a small number of events with three properties, rarity, extreme impact, and an explanation invented afterward, drive most of history, most markets, and most careers, and that our whole apparatus for handling uncertainty is built for the wrong kind. The practical answer is never a better forecast; it is changing your exposure so that being surprised is survivable, and much of the book is about what to believe rather than what to perform, which is why the Hold section here is long.

Do

Establish which quadrant you are in before you trust any model. Cross two questions: is the payoff simple and binary, or open-ended and dependent on magnitude, and is the randomness thin-tailed or fat-tailed? Three of the four combinations are survivable with ordinary tools. The fourth, open-ended payoffs in fat-tailed randomness, is where blowups live, because there you must estimate not just whether the rare event happens but how large it gets, which is exactly what cannot be estimated. [Source: "Postscript VI: The Fourth Quadrant, the Solution to that Most Useful of Problems"]

Split your exposure into two extremes instead of settling on a middle. Roughly 85 to 90% in maximally safe instruments no rare event can damage, and the remaining 10 to 15% in maximally speculative bets whose downside is capped at what you put in and whose upside has no ceiling. A single "medium risk" position depends on somebody's flawed model to even define what medium means; the barbell has a defined floor and an open top. [Source: "Chapter 13: Appelles the Painter, or What Do You Do if You Cannot Predict?"]

Sort every exposure by whether a surprise would help or hurt you, and take the opposite posture in each. Movies, publishing, research, and venture bets cap your loss and leave the gain open; banking and insurance cap the gain and leave a single rare event able to erase you. Be maximally aggressive where an error would benefit you and maximally guarded where an error would hurt, rather than carrying one appetite for risk everywhere. [Source: "Chapter 19: Half and Half, or How to Get Even with the Black Swan"]

Maximize the number of accidents that can find you. Serendipity cannot be aimed at, so the move is surface area: take the meetings, go to the parties, live where things happen, say yes to the unplanned, and seize an opportunity when it appears rather than assuming it will recur. Most people fail to even recognize a lucky break in the moment it shows up. [Source: "Chapter 13: Appelles the Painter, or What Do You Do if You Cannot Predict?"]

Look for the observation that would break your belief, not the one that would support it. Told that a rule produced the sequence 2, 4, 6, almost every subject in Wason's experiment tested only sequences designed to confirm their guess, and almost nobody found the actual rule. Chess grandmasters do the reverse by training: they hunt for the weakness in their own move rather than confirmation of its strength. [Source: "Chapter 5: Confirmation Shmonfirmation!"]

Question the stated assumption rather than computing perfectly inside it. Asked the odds of tails after ninety-nine heads in a row, the actuary says fifty percent and the street operator says no more than one percent, because a run like that means the premise of a fair coin is almost certainly wrong. In an exam, reasoning inside the box wins. Outside one, doubting the box usually does. [Source: "Chapter 9: The Ludic Fallacy, or The Uncertainty of the Nerd"]

Keep redundancy you would be embarrassed to justify. Two kidneys and two lungs look wasteful under naive optimization and exist to survive an unpredictable shock. The financial equivalent is cash reserves held rather than deployed, plus refusing to overspecialize, so that a second skill is there when the first one stops paying. [Source: "Postscript I: Learning from Mother Nature, the Oldest and the Wisest"]

Keep the parts small enough to fail without taking the rest down. Mother Nature caps the size of an individual organism without capping how many interact. The institutional version is letting fragile things break early while they are still small and refusing to let anything grow to where its failure has to be prevented, because ordinary competition otherwise rewards whoever has accumulated the most hidden risk. [Source: "Postscript VIII: The Ten Principles for a Black-Swan-Robust Society"]

Put the variability back into the routines you have smoothed out. Suppressing small stressors manufactures a false calm that makes the eventual shock worse, which is why preventing small forest fires sets up larger ones. Taleb's own regimen is deliberately unscheduled: long slow walks with rare random bursts of intense effort, feasting followed by skipped meals, cold exposure, irregular sleep. [Source: "Postscript II: Why I Do All This Walking, or How Systems Become Fragile"]

Don't

Don't take a long calm run as evidence of safety. A turkey fed every day grows more confident in human benevolence with each feeding, and its confidence peaks the afternoon before Thanksgiving, precisely when its risk is highest. Taleb's real-world version is the long post-Napoleonic peace that preceded the deadliest war to that point. [Source: "Chapter 4: One Thousand and One Days, or How Not to Be a Sucker"]

Don't accept a "because." On the day Saddam Hussein was captured, Bloomberg reported Treasury bonds rising because his capture reduced the terrorism threat, then reported roughly half an hour later, as bonds fell, that the same capture had boosted appetite for riskier assets. Split-brain patients invent a reason for an action they cannot know the cause of rather than admit they do not know. [Source: "Chapter 6: The Narrative Fallacy"]

Don't build a lesson out of the survivors. Diagoras, shown paintings of worshippers who prayed and survived a shipwreck, asked where the paintings of those who prayed and drowned were. Studies of successful people's traits, courage, optimism, risk-taking, are worthless alone, because the graveyard is full of people who had exactly the same traits. [Source: "Chapter 8: Giacomo Casanova's Unfailing Luck: The Problem of Silent Evidence"]

Don't trust a forecast whose error rate nobody has ever counted. Tetlock studied roughly 300 political and economic experts across about 27,000 predictions and found error rates far above the experts' own estimates, no advantage for PhDs, and a negative correlation between reputation and accuracy. Asked for a range they are 98% confident contains a true value, people miss 15 to 45% of the time instead of 2%, and Harvard MBAs are no exception. [Source: "Chapter 10: The Scandal of Prediction"]

Don't add information expecting accuracy to follow. More data reliably raises confidence and does not reliably raise accuracy. Clinical psychologists given more patient data grew more confident in diagnoses that did not get better; bookmakers given more variables did the same on horse races; subjects shown a blurry image resolving in more incremental steps recognized it more slowly, because the intermediate steps generated false hypotheses that stuck. [Source: "Chapter 10: The Scandal of Prediction"]

Don't run bell-curve tools on a scalable quantity. Over fifty years of U.S. market history the ten single most extreme days account for roughly half of all cumulative returns. Locke defined a madman as someone reasoning correctly from erroneous premises, and Taleb applies it directly: a fund built by the authors of the option-pricing formula collapsed in 1998 on a combination of events its models treated as essentially impossible. [Source: "Chapter 17: Locke's Madmen, or Bell Curves in the Wrong Places"]

Don't read a quiet stretch as a low-risk stretch. The shift into fat-tailed territory is often marked by falling short-term volatility even as the chance of a large jump rises, which is how the "great moderation" reading of pre-crisis calm went wrong. Distrust stress tests for the same reason: they anchor on the worst past deviation, and the same method would have failed to see that anchor event the day before it happened. [Source: "Postscript VII: What to Do with the Fourth Quadrant"]

Don't build your defense around the last specific disaster. After September 11 people learned narrow rules about airport security rather than the general lesson that unforeseen high-impact events happen. The Maginot Line was built on excellent lessons from the exact prior war and was rendered useless because the attack simply went around it. [Source: "Prologue"]

Hold

A Black Swan is defined relative to an observer, never as a property of the event. The September 11 attacks were a Black Swan to the victims and not to the planners; a Black Swan for the turkey is not one for the butcher. Holding this stops the pointless argument about whether an event "really counts" and turns it into the useful question, which is who was in a position to know. [Source: "Postscript IV: Asperger and the Ontological Black Swan"]

Reconstructing the past is harder than forecasting the future, not easier. Predicting how an ice cube melts into a puddle over two hours is straightforward; given only the puddle, recovering the shape of ice that produced it is vastly harder, because countless starting shapes produce the same puddle. Held, this drains authority from confident causal accounts of history, including your own account of your own life. [Source: "Chapter 12: Epistemocracy, a Dream"]

Almost nothing that changed things was aimed at. Penicillin came from mold contaminating an abandoned experiment. The cosmic microwave background came from two astronomers trying to eliminate a hiss they first blamed on bird droppings. The laser was built with no application in mind. Viagra began as a failed hypertension drug. Taleb's phrase for the pattern is a solution waiting for a problem. [Source: "Chapter 11: How to Look for Bird Poop"]

Prediction fails structurally, not just from incompetence. To predict a future discovery with any content you would already have to possess it. Poincare showed that adding a third body to an orbital system makes long-run prediction explosively sensitive to tiny errors, and Berry's extension is that predicting the fifty-sixth collision on a pool table would require knowing the position of every particle in the universe. Better analysts do not close that gap. [Source: "Chapter 11: How to Look for Bird Poop"]

A number attached to a rare event came from a theory, not from data. Because the event is rare, its probability cannot be estimated from frequency, so the figure necessarily rests on an unverifiable prior assumption. A Lehman employee told the Wall Street Journal that the August 2007 market events should happen once every ten thousand years, and three similar events then happened on three consecutive days. [Source: "Postscript V: (Perhaps) the Most Useful Problem in the History of Modern Philosophy"]

What matters is the payoff attached to being wrong, not whether the belief is true. Taleb's charge against centuries of epistemology is that it stayed in two dimensions, true and false, and left out the third: the size of the consequence. Held, "how likely is this" stops being the interesting question and "what does it cost me if I am wrong" becomes it. [Source: "Postscript V: (Perhaps) the Most Useful Problem in the History of Modern Philosophy"]

Nobody at the top of a winner-take-all field is safe there. Only 74 of the 500 largest U.S. companies of 1957 were still in the index forty years later. Because position in these domains owes so much to compounding early advantage rather than to durable superiority, a newcomer can displace an entrenched giant essentially overnight, and Taleb treats that churn as fairer than a world where fixed ability decided permanently. [Source: "Chapter 14: From Mediocristan to Extremistan, and Back"]

A lumpy payoff has to be paid for in waiting, and the waiting is the hard part. Our reward systems evolved for linear cause and effect, so long flat stretches register as failure well before any breakthrough. Taleb's image is the officer in The Tartar Steppe who spends thirty-five years at a remote fortress anticipating the attack that would justify the wait, and dies at a roadside inn as it finally begins. Holding this makes the flat stretch legible instead of damning. [Source: "Chapter 7: Living in the Antechamber of Hope"]

Being prepared to lose everything is the last defense available. Seneca stayed one of the richest men of his era and kept himself ready to lose it all daily, which is what Taleb says gives the Stoic argument its credibility, since it is not the counsel of someone with nothing at stake. Stilbo, asked what he had lost after his city was sacked and his family killed, said he had lost nothing, because his goods were all with him. [Source: "Postscript IX: Amor Fati: How to Become Indestructible"]

The one line

You cannot know which rare event is coming or when, so stop buying forecasts and start buying a shape of exposure where the surprise that arrives can only cost you a little and might pay you a lot.