Luck vs. Skill

Concept Analysis

Definition & Origins

The luck-versus-skill question is one of the most intellectually honest themes in Marks' memo corpus — and one of the rarest in investment management discourse, where self-promotion is the norm. Marks engages directly and repeatedly with the uncomfortable reality that short-term investment results are substantially driven by randomness, and that the ability to distinguish skilled investors from lucky ones — including oneself — is genuinely difficult.

The intellectual lineage draws on probability theory, Michael Mauboussin's work on skill and luck, and Nassim Taleb's analysis of randomness in financial markets. But Marks' contribution is the practitioner's perspective: the honesty to apply this framework to his own career and Oaktree's own track record.

The seed was planted long before any memo. Marks has written: "The first thing I remember learning at Wharton in 1963 was that the correctness of a decision can’t be judged from the outcome." That single classroom observation — absorbed at age seventeen, decades before Oaktree existed — is the trunk from which everything else in this concept grows. The later reading gave the idea its vocabulary: Taleb's Fooled by Randomness supplied the language of alternative histories, Mauboussin's Alpha and the Paradox of Skill supplied the framework for separating process from result, and Elroy Dimson's dictum that more things can happen than will happen supplied the probabilistic grounding. What Marks added was the willingness to aim the framework at himself — to write, in print, for his clients, a list of the ways demographic luck contributed to his own success.

Core Ideas

Good decisions can produce bad outcomes, and bad decisions can produce good outcomes. This is the core insight. In any probabilistic endeavor, the short-term correlation between decision quality and outcome quality is weak. A sound investment thesis can produce a loss if low-probability negative scenarios materialize. A reckless bet can produce a gain if high-probability failure scenarios fail to materialize. Judging the quality of decisions by their outcomes — rather than by the quality of the process and the probabilities — is the most common analytical error in evaluating investment performance.

Survivorship bias distorts our perception of skill. The investment managers with the best track records may be the most skilled, or they may be the ones whose particular risk exposure happened to match the market environments of their careers. A manager who made aggressive growth bets in 1995-2000 and again in 2010-2021 would have a spectacular record — not because they were skillful, but because they were lucky about when they were born and what strategies were fashionable during their peak career years.

Large sample sizes are required to distinguish skill from luck. In a field where outcomes are probabilistic and careers span 20-30 years, the statistical evidence for distinguishing skill from luck is weaker than most practitioners acknowledge. Marks cites research suggesting that even 20-year track records are insufficient to statistically confirm alpha at conventional significance levels.

Process is the observable; outcomes are random. Because we cannot directly observe the probability distributions underlying investment decisions, the best we can do is evaluate the quality of the process: Was the thesis well-reasoned? Were the risks understood? Was the position sizing appropriate for the conviction level and uncertainty? A good process that produced a bad outcome is being incorrectly evaluated if judged only by results.

Intellectual honesty about luck is a competitive advantage. Counterintuitively, the investor who clearly sees the role of luck in their success is better positioned for long-term performance. They avoid overconfidence (the principal cause of concentrated losses), they maintain appropriate humility about position sizing, and they focus on improving process rather than celebrating outcomes.

Being right once proves almost nothing. The investment business is full of people who became famous for a single dramatic call that events subsequently validated. Marks offers the case of Joe Granville, the technical analyst whose 1976 warning preceded a 26% two-year decline and won him lasting fame — but whose next accurate call did not arrive for twenty-four years. One correct forecast, however spectacular, is statistically indistinguishable from luck; the forecaster may have been right for the wrong reason, or simply been the one person in a crowd of guessers whose number came up. Marks' test is different: "the skillful investor is right more often, over a long period of time, than an assumption of randomness would permit." Only a long record of consistently sound decisions — not a highlight reel — licenses the conclusion that skill, rather than chance, is at work.

Practical Application

The Getting Lucky memo (2014): Marks walks through a series of investment scenarios to illustrate how indistinguishable good luck and good skill can look ex post — and how the appropriate response is to evaluate each through the lens of process, not outcome.

Portfolio management implication: If outcomes are significantly random, position sizing should be calibrated to conviction level and analytical confidence — not to recent performance. A position that has performed well is not automatically more likely to continue performing well; a position that has performed poorly is not automatically more likely to reverse. The process evaluation, not the recent outcome, should drive sizing decisions.

The Penn endowment case study (2012): In "Assessing Performance Records: A Case Study," Marks put himself on the examining table. He had joined the University of Pennsylvania's Investment Board at the end of a highly bullish period, maintained a cautious approach, and watched the Global Financial Crisis vindicate that caution. His uncomfortable question: what if he had gotten the job two years earlier? The same decisions would then have included a year of horrendous underperformance and omitted the year of vindication, and he would have been considered a very average chairman at best. The strategy was identical; only the start and end dates of the measurement window changed. His conclusion is the practical rule for anyone evaluating a manager: "be understanding when evaluating track records, and refuse to accept the results at first glance."

Common Misconceptions

Misconception 1: Great track records prove skill. All investment track records are a mixture of skill, the factor exposures of the era, and the realization of probabilistic outcomes. Decomposing these components — especially over short periods — is genuinely difficult and requires more analytical rigor than simply looking at returns.

Misconception 2: Admitting luck means denying skill. Marks explicitly acknowledges that Oaktree's performance reflects genuine skill. His argument is not that skill doesn't matter — it is that skill and luck are both contributors to outcomes, and intellectual honesty requires acknowledging both.

Misconception 3: Successful people make their own luck. "Success is never accidental," Jack Dorsey tweeted, and the sentiment — no accidents, just planning; no luck, only strategy — is the standard executive summary of every success story. Marks' response, borrowing Ed Smith's essay "In Defence of Luck," is that a great many things contribute to success: some are our own doing, while many others are beyond our control. Hard work, planning and persistence are essential for repeated success, but even the hardest workers and best decision makers will fail to succeed consistently without luck — the accidents of birth and genetics, chance meetings, fortuitous choices, and unforeseeable events that cause decisions to turn out right. The denial of luck is not humility's opposite by accident; as the sociologist Michael Young warned, people who believe their advancement comes entirely from their own merits tend to become insufferably smug.

Howard Marks' Own Words

Thought Evolution

Related Concepts

Key Memos

Mentioned In

Source: Chian.io — Howard Marks Knowledge Base