In Fooled by Randomness, Nassim Nicholas Taleb invents ten thousand investment managers. Each year, every manager has an even chance of gaining or losing the same amount. One losing year removes a manager from the experiment. There is no skill in these rules. Yet after five years, we should expect roughly three hundred thirteen managers with an uninterrupted record of success.
Hundreds of potential stars, and we haven't given any of them talent. To judge a real winner, Taleb argues, we need the missing experiment around the record. We need the full group that tried, including its failures, and an explanation of how the results arose. A profitable outcome tells us what happened. It doesn't establish that the choice was sensible or the person unusually able. Repetition helps only when we understand what is being repeated.
A profile of one survivor might credit discipline, unusual habits, or an impressive education. Those qualities would make a satisfying explanation. They would also explain a result that our experiment produced without them. Choosing the winner afterward is very different from naming one manager beforehand and predicting five wins. More attempts give chance more opportunities to produce someone extraordinary.
Next year, each surviving manager still has an even chance of winning. The eliminated managers would have the same chance if allowed back in. A lucky streak provides no new advantage under unchanged, independent odds. It doesn't mean the winners are now owed a loss. And outside this deliberately skill-free experiment, bad luck can force a capable person out before their advantage has time to show. Even comparing two records can exaggerate a difference in ability when one person has been lucky and the other unlucky.
Recover the missing comparison
Real failures are harder to find. People close businesses, leave professions, or stop advertising. The successful musician remains visible; the unsuccessful musicians who tried just as hard disappear from the account. A fund's exceptional history may be the very reason its advertisement reached you. The selection happened before you began judging the evidence.
Historical tests can make this mistake quietly. Taleb describes a strategy that shifts money toward poorly performing investment managers. It looks excellent when tested on managers still operating at the time of the study. But that sample includes losers who recovered and excludes losers who failed completely. The test has already chosen the survivors. To repair it, start with everyone present at the beginning and keep the failures in the record.
A study of wealthy savers can select prosperous people during an unusually prosperous investment era. That selects twice. Diligent savers who held assets that collapsed are missing. Saving may help build wealth, but the selected winners cannot show that it guarantees their outcome.
None of this makes skill imaginary. A cook produces meal after meal, giving people repeated chances to observe competence. A chief executive makes relatively few large decisions in a changing environment. One enormous profit remains one outcome. Comparable repetitions make a modest underlying advantage easier to detect; the size of the prize cannot replace them.
Nor does winning repeatedly in one environment show how someone will fare in another. A trading style can fit a long rising market exceptionally well while remaining vulnerable to a reversal. Ranking traders at that moment rewards the fit to conditions so far. It cannot establish that competition has selected the people best prepared for whatever comes next.
When winning changes the odds
There is another reason a lucky start can become a long career. In Taleb's audition thought experiment, several actors can do the work. A small incidental preference helps one get the role. Being chosen makes that actor known, and being known attracts further attention. Later opportunities are now affected by the first success.
The advantage is real. But its persistence doesn't prove that the first choice identified someone vastly more talented. Unlike our imaginary managers, the actor isn't taking a fresh, independent test each time. Earlier results have changed what happens next. Selection changes which people we see; reinforcement changes their opportunities.
Products can acquire similar advantages. A program becomes more useful because other people already use it. Established habits can also make switching unattractive, even without technical superiority. One mechanism adds value through other users, while the other makes earlier choices harder to undo. Either can magnify a small initial lead into a huge difference in rewards.
Such growth need not be smooth. Accumulating connections can make one small addition produce a much larger change. Understanding that process still doesn't tell us exactly when a particular product or career will take off. The apparent tipping point may be obvious only afterward.
The reverse difficulty matters too. Learning can show little visible progress before understanding suddenly improves. Effort and reward need not arrive in matching installments. A stalled result can therefore be premature evidence of failure, although waiting longer guarantees nothing.
Persistent success deserves investigation. It may reveal ability, favorable conditions, or an advantage that early luck made durable. We can look for comparable trials and recover missing entrants. Sometimes we still cannot separate the explanations. Taleb leaves room for that uncomfortable answer. A record can be impressive while the credit remains uncertain.
Three things to keep
Judge a winning record against everyone who had a chance to produce it.
Repeated results reveal more when the trials and underlying conditions remain comparable.
Early luck can create lasting advantages without proving exceptional initial ability.