The Fifteenth YearWas the Least Informative of All

1 · The openingThe year the streak was most convincing

Between 1991 and 2005, the Legg Mason Value Trust beat the S&P 500 in fifteen consecutive calendar years. No other mutual fund manager had done it. Bill Miller ran upwards of $70 billion at the peak, Morningstar named him fund manager of the decade, and the arithmetic looked unanswerable: if roughly three quarters of funds trail the index in any given year, a fifteen-year run by chance sits somewhere below one in a billion. Then 2008 came, the fund's financial holdings collapsed, and Value Trust fell 55 per cent against a 37 per cent fall in the index.¹

The useful question is not why 2008 went wrong. It is what the fifteenth year was actually evidence of, on the day it happened, before anybody knew what came next. Nassim Nicholas Taleb published a book on exactly that question in 2001, when Miller's streak was ten years old and being cited in every profile as proof that skill beats the market.

2 · The argumentWhat the book claims

Fooled by Randomness is about a single confusion with expensive consequences: mistaking an outcome for evidence of the process that produced it. Reality delivers one history. We observe that history, work backwards to a cause, and treat the cause as established. Taleb's counter-move is to insist that the realised path is one draw from a distribution of paths that could equally have run, and that a serious judgement about anyone's ability requires imagining the drawer rather than admiring the draw.

From that single move everything else in the book follows. If outcomes are draws, then track records are samples, and small samples from skewed distributions tell you almost nothing. Survivors are visible and the dead are not, so any population of the successful is pre-selected for luck. Frequency of being right is not the same as being right in expectation, so a strategy that wins nineteen times in twenty can still be a catastrophe. And the more often you look at a fluctuating quantity, the more of what you see is noise. Taleb states the asymmetry directly: “Mild success can be explainable by skills and labor. Wild success is attributable to variance.”

This is the first book of what became a sequence, and it is the quieter one. Where The Black Swan is about the events the record cannot contain, Fooled by Randomness is about the events the record does contain and what they fail to prove.

3 · The authorWhere Taleb is standing

Taleb wrote this as a working options trader rather than as an academic, and the book carries the texture of the trading floor: the colleague who blew up, the bonus season, the analyst on television explaining the day. He had lived through Lebanon's collapse from prosperity into civil war, which taught him early that a long stable record proves nothing about the next period. He was also running strategies that lost small amounts most days and made large amounts rarely, which is a specific vantage on the question of who looks competent.

That vantage is a genuine conflict, and worth naming. A trader whose book is short frequency and long magnitude has a professional interest in the world believing that frequent winners are lucky. Taleb does not hide this; his fictional stand-in, Nero Tulip, is transparently the author's ideal self, and the book's villains are drawn with more relish than fairness. Read the mechanism, which is rigorous, and discount the character sketches, which are advocacy. The prose in the 2001 edition is looser and more personal than his later work, and some readers find that the charm and some find it the flaw.

4 · Key insightsFive ideas that carry the book

Judge the generator, not the draw

Taleb's central instrument is what he calls alternative histories: the set of paths the world could have taken from the same starting conditions. A decision is good or bad in relation to the whole set, and history hands you one member of it, chosen at random. The Russian roulette illustration does the work in a sentence. Someone who plays for a million dollars and survives is a millionaire in this history and a corpse in one chamber out of six, and no amount of admiring the millionaire tells you the game was worth playing.

Applied to business, it dissolves most of what passes for evidence. The founder who bet the company and won is a case study; the identical founder in the branch where the market moved the other way is not written about at all. Once you accept that outcomes are draws, praise and blame both have to be re-attached to the decision rather than the result, which is an uncomfortable discipline because only the result is observable.

The dentist who checks too often

Taleb's most transferable piece of arithmetic concerns a retired dentist with a portfolio expected to return 15 per cent a year with 10 per cent volatility. Over a full year, the chance of showing a gain is about 93 per cent. Look at the same portfolio monthly and it falls to roughly 67 per cent. Daily, 54 per cent. By the minute, 50.2 per cent. The underlying investment has not changed at all; only the sampling rate has.

The reason is mechanical. Expected return scales with time while volatility scales with the square root of time, so shortening the window shrinks the signal faster than it shrinks the noise. At high frequency the dentist is watching a coin. He will nonetheless form opinions about what he sees, because human beings are not equipped to observe a random series without interpreting it, and each opinion is a fresh opportunity to act on nothing.

The survivors were always going to exist

Start ten thousand managers, give each a fair coin, and let a head count as beating the market. After five years, chance alone leaves roughly three hundred with unbroken records. Those three hundred are not frauds. They will be interviewed, and they will describe a philosophy, and the philosophy will be internally coherent, because human beings are excellent at constructing one. The cohort of 9,687 who flipped a tail somewhere is not available for comment, and the tails that occurred inside the winners' own runs have been forgotten.

Taleb pushes further than the familiar survivorship point. The size of the starting population determines how impressive the surviving record has to be before it means anything, and almost nobody adjusts for it. Add a second effect and the picture worsens: those who took the most reckless positions are over-represented among both the dead and the spectacular survivors, so the visible winners are drawn disproportionately from the people whose method was worst.

Being right often is not the same as being right

Taleb reports being asked, at a meeting, whether he was bullish or bearish, and answering that he was slightly bullish about the market and heavily short it. To most of the room this sounded like a contradiction. It is arithmetic. If the market is 70 per cent likely to rise by 1 per cent and 30 per cent likely to fall by 10 per cent, the most probable direction is up and the expectation is negative.

Once you separate frequency from magnitude, a great deal of ordinary business behaviour reads differently. A trading strategy, a product line, a hiring approach or a covenant structure can produce a long run of small confirmations and carry a negative expected value the whole time. The confirmations are what get reported, discussed and rewarded. The magnitude term shows up once, at the end, and is then explained as an unforeseeable shock.

Wittgenstein's ruler and the hindsight machine

The last idea is epistemic housekeeping. Taleb borrows an image from Wittgenstein: measuring a table with a ruler tells you about the table only if you already trust the ruler, and otherwise the table is measuring the ruler. Applied to performance, a manager's returns measure the manager only if the market environment is a reliable instrument. In a period when one style was rewarded, the returns are largely measuring the period.

He pairs this with a claim about biology. Human beings did not evolve to reason about probability distributions; we evolved to detect patterns quickly and to feel losses far more sharply than equivalent gains. The consequence is that the errors in this book are not correctable by knowing about them. Taleb is explicit that he suffers from all of them himself and manages the problem with rules and physical distance from the screen rather than with insight, which is the most honest paragraph in the book.

5 · EchoesWhere the argument rhymes elsewhere

The book's structural claims travel well past markets. Three parallels test the argument on unfamiliar ground.

Clinical epidemiology reached the same conclusion first. Medicine spent centuries treating before-and-after observation as evidence, until regression to the mean, natural remission and placebo response were understood well enough to make the randomised controlled trial compulsory. A fund's track record is an uncontrolled before-and-after study on a sample of one patient. The parallel exposes something Taleb leaves implicit: the answer is not scepticism but design, and medicine solved its version by changing what counts as admissible evidence rather than by asking clinicians to be more humble.

Evolutionary biology recognises a similar pattern in genetic drift. Motoo Kimura's neutral theory showed that in populations of finite size, variants with no fitness advantage reach fixation routinely, purely through sampling. Look at a fixed trait and the temptation is to construct an adaptive story for it. Markets contain populations large enough and generations short enough for drift to dominate selection over any period a career spans, which reframes the survivor problem: the visible winners may not be lucky exceptions to selection but the ordinary output of a process where selection is weak.

Signal processing supplies the general law behind the dentist. Any measurement whose signal accumulates linearly with observation time while its noise accumulates with the square root will show a signal-to-noise ratio that improves only as the square root of the window. Taleb's table is a special case. Stated generally, it gives a design rule that reaches every dashboard in a company: sampling faster than your signal accumulates does not give you earlier information, it gives you more noise with a timestamp on it.

6 · The business ledgerWhat the record proved, and what it did not

Taleb wrote about traders because that was his floor. The mechanism is not about traders. Each case below turns on the same defect: a signal was read out of a series that was mostly noise, or a survivor was read as evidence about a population.

Neil Woodford is the cleanest natural experiment in modern asset management, because the record transferred and the returns did not. He built a formidable reputation over more than two decades at Invesco Perpetual, launched his own equity income fund in 2014, and raised £1.6 billion on day one from investors who were buying the track record. The fund peaked above £10 billion. Illiquid unquoted holdings and sustained outflows then forced its suspension in June 2019 with around £3.7 billion remaining, trapping roughly 300,000 mostly retail investors; the fund was wound up four months later, and in 2025 the FCA announced fines against Woodford and his firm.² The reputation was real. It was also, in Taleb's sense, a sample.

The systematic version of the same finding sits in S&P's Persistence Scorecard, and it is more damning than the anecdotes. Of the top-quartile US active domestic equity funds as of December 2020, not one remained in the top quartile across the following four years. Lower the bar to the top half and 4.21 per cent of US equity funds stayed there, which is worse than the 6.25 per cent that random reshuffling would produce.³ S&P's own framing is Taleb's argument in a sentence: skill would persist, and luck does not.

Move outside asset management and the same defect shows up in reporting cycles. Graham, Harvey and Rajgopal surveyed 401 financial executives and found that 78 per cent would sacrifice long-term economic value to deliver a smoother earnings path, and 55 per cent would decline a clearly positive-NPV project rather than miss the current quarter's consensus.⁴ Read that alongside Exhibit 2. The quarter is a sampling interval chosen for administrative convenience, its contents are heavily noise, and companies are destroying real value to make that noise look smooth. Taleb's dentist at least only harmed himself.

Two organisations in the ledger got it right, and both did so by refusing to treat single outcomes as verdicts. Microsoft's experimentation platform found that only about a third of tested ideas improved the metric they were built to improve.⁵ That number is the point: without controlled measurement, two thirds of shipped features would have been attributed to the judgement of whoever proposed them. Amazon's Fire Phone produced a $170 million write-down in the third quarter of 2014, including $83 million of unsold handsets.⁶ The company treated it as one draw from a portfolio of bets rather than as evidence that its process was broken, and the same device organisation shipped Echo. Neither the write-down nor the Echo tells you much on its own; the portfolio does.

7 · DisciplinesFour things to do differently on Monday

The book offers temperament where a reader wants procedure. What follows is the procedural residue, opened out for anyone who wants to go further and compressed for anyone who wants to remember it.

8 · RelevanceWhy the book bites harder in 2026

Taleb wrote for a reader who checked a portfolio on a screen. Since then the observation frequency of the entire economy has risen. Live dashboards, weekly cohort reads, daily active user counts, real-time attribution and continuous performance feedback have pushed most organisations well past the point where the signal is visible in the window they are watching. The dentist has become the operating model.

Machine learning makes this sharper rather than safer. A model trained on a fluctuating series will find structure in it, because finding structure is what the method does, and it will report that structure with a confidence figure attached. Generative tools then supply a fluent narrative explanation for whatever the metric did last week. Taleb's narrative fallacy used to require a human to do the storytelling and to get tired. That constraint has been removed.

9 · StrainsWhere the argument gives way

The book invites engagement from the disciplines it borrows from rather than dismissal from outside them. Four lenses bring its real claims into focus.

Through the lens of decision science, Taleb diagnoses without prescribing. The outcome-bias literature that Baron and Hershey opened in 1988 supports him, and the work that followed built protocols he shows no interest in: reference-class forecasting, premortems, calibration training, structured decision journals. Taleb's remedy is stoic temperament and physical distance from the screen, which is a personal solution to a problem he has shown to be structural. The gap between his diagnosis and his prescription is the largest weakness in the book.

Through the lens of finance, the noise argument leans on assumptions his own later work rejects. The dentist calculation assumes returns that are independent, identically distributed and Gaussian with a constant drift, which is exactly the machinery The Black Swan calls a fraud. The conclusion survives under fatter tails, but the tidy table does not, and Taleb presents the table as though the arithmetic were the argument. There is also a stronger claim available to him than the one he makes. Cross-sectional studies using false discovery methods find that a small minority of managers do show genuine alpha, which is more precise and more damaging than saying performance is all luck.

Through the lens of organisational economics, the incentives outlive the insight. A manager paid on annual results with no clawback holds a call option on volatility, and the rational play is to sell hidden tail risk and collect the frequent small wins. This is a structural feature of compensation, not a failure of temperament, and no amount of reading Taleb changes it for the individual inside such a contract. The book's implied remedy, becoming a more sceptical person, does not touch the mechanism that manufactures lucky fools in the first place.

Through the lens of the philosophy of science, the framework is hard to falsify from the inside. If success is attributed to luck and failure to the tail, no observation can count against the thesis. Taleb leans on Popper throughout, which makes the omission conspicuous. The argument becomes properly testable only when it is turned into a measurement question of the kind Mauboussin and the persistence literature ask: given this population and this variance, how long a record would chance produce? Read that way it is a strong empirical claim, and it holds up. Read as Taleb often writes it, as a stance, it is unfalsifiable.

10 · Adjacent readingWhere to go next

11 · ClosingWhat the streak was worth

Return to the fifteenth year. In January 2006 the record was complete, the odds against it looked astronomical, and every profile treated the run as settled evidence. What nobody in that moment could see was the denominator: how many managers had been running, with what dispersion of styles, over how many periods, and how many perfect runs a population that size throws off on its own. Miller himself later said the streak combined skill and luck, which is the most accurate thing anyone said about it and the least quotable. The fifteenth year added one observation to a sample that was still far too small, and it felt like proof precisely because it was the longest.

If the last five years had gone the other way, would the strategy have been wrong, or only the outcome?

12 · Carry awayThe compressed reading

If you read this synopsis once and leave with only what these five items say, you have the essential reading.

13 · ReferencesSources for the figures cited