A run of successful investment outcomes is always welcome. When those outcomes are the product of random variance rather than genuine skill, the welcome they receive can be actively dangerous—because random success produces the same psychological effects as skill-based success while providing none of the actual predictive power that genuine skill would imply. The investor who has experienced a run of random successes will feel confident, will believe her approach is working, and will increase her commitment to the approach at exactly the moment when regression to the mean makes a run of poor outcomes most likely.
The mathematics of random success are instructive. In any population of investors making roughly comparable investment decisions, a significant fraction will experience above-average returns in any given period simply as a result of variance. After five years, a meaningful fraction will have consistently outperformed by chance alone—not because they possess any skill, but because the probability of extended runs of good luck, in a population of sufficient size, is not negligible. The investors in this lucky fraction are indistinguishable, on the basis of their five-year track record alone, from investors who possess genuine skill.
The reinforcement learning dynamic makes this particularly dangerous. Human brains are designed to extract patterns from experience and to reinforce the behaviours that have produced rewards. Random success is experienced as reward in exactly the same way as skill-based success; the reinforcement system does not distinguish between them. The investor whose successful trades are the product of random variance will have those trades reinforced in exactly the same way as if they were the product of genuine insight—meaning she will be more likely to repeat the approach, more likely to increase position sizes, and more likely to develop an analytical narrative around what the approach entails. The narrative is post-hoc rationalisation of random outcomes, but it feels like genuine pattern recognition.
The specific sequence in which luck arrives matters as much as its overall magnitude. The investor who experiences success early in her investment career, when she is most susceptible to updating her self-assessment and most likely to make the behavioural changes that successful outcomes encourage, is more damaged by lucky success than the investor who experiences the same run of luck later, when she has a more stable and calibrated self-assessment. Early success tends to produce large and lasting upward revisions to self-assessed skill that persist even after the luck reverses, because the early revision is anchored into the investor's self-concept in ways that later disconfirming evidence struggles to dislodge.
The corrective is the development of a statistical framework for evaluating one's own performance that explicitly accounts for variance. This means asking, of any run of successful outcomes: is this result consistent with what a skilled investor would produce, or is it also consistent with what a lucky but unskilled investor would produce over the same period? The honest answer, for most periods and most investors, is the latter. Holding this answer genuinely—allowing it to constrain the risk-taking that successful outcomes encourage—is the primary defense against the trap that random success sets.