Why confidence peaks just before disaster strikes
Nassim Nicholas Taleb (2007) — Based on Bertrand Russell's chicken (1912)
A turkey is fed by a farmer for 1,000 consecutive days. Each day, the turkey's confidence grows that the farmer is kind and benevolent. By day 1,000, the turkey's certainty has never been higher. The economics department, risk management division, and analytics team of the turkey all concur: humans are wonderful creatures who exist to feed turkeys.
Then comes Day 1,001—the Wednesday before Thanksgiving. The farmer arrives not with food, but with an axe.
The turkey's confidence reached its maximum precisely when its risk was highest. The feeling of safety peaked the night before death. Every additional day of feeding strengthened the belief that feeding would continue forever—but brought the turkey one day closer to Thanksgiving.
The turkey relied on inductive reasoning: observing specific instances (daily feedings) to derive a general rule ("I will always be fed"). This is how most of us navigate life—and it usually works. But inductive reasoning has a fatal flaw: no number of observations can prove a universal law.
The turkey had access to all the data available within its world—1,000 days of consistent feeding. What it couldn't access was the hidden structure of reality: the existence of Thanksgiving, the farmer's economic incentives, and the very concept of being "fattened up." The most important piece of information was invisible to the turkey's data collection.
Taleb notes that what is a "Black Swan" (unpredictable event) for the turkey is not a Black Swan for the butcher. The butcher knew all along. The lesson: ask yourself whether you're the turkey or the butcher in any given situation.
September 15, 2008
158 years of successful banking. "Too big to fail." Employees confident until the final weekend. Bankruptcy came on a Monday morning.
August 1998
Nobel laureates. 40% annual returns for 4 years. Their models said disaster was a 10-sigma event (virtually impossible). It happened anyway.
January 19, 2012
130 years dominating photography. Invented the digital camera in 1975. Filed for bankruptcy when digital killed film—their own invention.
September 23, 2010
9,000 stores worldwide. Declined to buy Netflix for $50 million in 2000. Netflix is now worth $150+ billion.
2007-2008
"Housing prices always go up." Decades of data confirmed it. Then they didn't. $8 trillion in home value evaporated.
January 15, 2015
Swiss National Bank promised to maintain 1.20 EUR/CHF floor "with utmost determination." Abandoned without warning. 30% move in minutes.
Taleb offers these strategies to avoid the turkey's fate:
Instead of trying to predict Black Swans (impossible), identify where you're fragile—where a rare event would devastate you. Reduce fragility before the event occurs.
The turkey's data was complete within its observable world. But the unknown unknowns—Thanksgiving, the farmer's motives—were the real determinants. Always question what's outside your data.
Each feeding confirmed the turkey's belief. But confirmation is cheap—any theory can find confirming evidence. Look for what could disconfirm your beliefs.
In every transaction, someone knows more. Ask yourself: am I the turkey or the butcher? Is my counterparty acting on information I don't have?
The turkey had no options—no escape route, no plan B. Create situations where you have limited downside and unlimited upside (what Taleb calls "antifragility").
Past performance is not just a "poor predictor" of future results—it can be actively misleading. The longer a pattern persists, the more confident we become, even as the probability of a pattern-breaking event may be increasing. The turkey's tragedy is not that it was stupid, but that it was rational given its available information.
The problem isn't inductive reasoning itself—we'd be paralyzed without it. The problem is overconfidence in induction, especially in domains where rare, catastrophic events are possible. Some turkeys are destined for Thanksgiving. The question is whether you know which kind you are.